London represents the first time the Olympics is being live-streamed over the Internet and rich media coverage is so widely and effortlessly available on any screen. We no longer have to sit glued to our TVs and laptops, slaves to network broadcast schedules. Multi-screen viewing - and especially via high-resolution tablet and smartphone screens - has freed us from the constraints of time (primetime), place (our living room) and a single screen in the same way TiVo, DVRs and streaming services like Hulu have revolutionized TV consumption.

According to a recent survey of likely Olympics viewers by web-based surveys firm SurveyMonkey, television is still the primary viewing device for the London Games, however, nearly 60% of respondents identified their secondary viewing source as the computer and the smartphone as their next most popular screen. Mobile marketing technology firm Velti uncovered similar intentions in their pre-games survey: more than 40% of Americans were expected to live-stream the London Olympics to their smartphones and tablets, and 77% specifically planned to use their smartphones to live-stream coverage. And if my unscientific survey of colleagues, friends and family is any indication, even those Olympic viewers watching live-streamed or time-shifted content over TV also are likely to be searching for additional results, coverage and social media interactions via their tablets and smartphones at the same time. As Byte magazine points out, mobile devices (including tablets) are giving fans more control over how they watch sports.

The London Games highlight another by-product of our increasingly web-enabled, bandwidth-intensive viewing habits: traffic spikes are no longer confined to predictable peak periods. Similar to other recent world events that drove significant spikes in global Internet traffic and defied traditional definitions of 'peak hour' (World Cup football matches, the UK Royal Wedding and the 2011 Japanese tsunami come to mind), Vodafone reported a record for the amount of data carried over its mobile network in the moments following Tour de France winner Bradley Wiggins' gold medal for Britain in the Olympic cycling time trial. That event occurred on a Wednesday afternoon.

A cautionary tale for network operators not only to build out adequate capacity but also to ensure that capacity can efficiently absorb the more distributed and extreme data traffic peaks symptomatic of the broadband era. The next big (data) thing is guaranteed to consume it.

Despite lackluster economic conditions in most regions of the world, mobile broadband subscribers continue to consume more and more bandwidth with an ever-expanding range of bandwidth-intensive apps and devices. One can argue how steep the growth curve is, but the underlying cause isn't in dispute. As a result, network operator demand for capacity (in the form of radio spectrum and backhaul) is not abating either.

At the recent Small Cells World Summit industry conference in London, O2 UK's chief radio engineer Robert Joyce explained: "As we see it, with the increasing demand from tablets and smartphones the macrocell will not be able to cope. We can take the macro cell grid to eight times its current capacity and then we'll run out of spectrum." O2 UK claims they could run out of spectrum in their macrocell layer as soon as 2014 based on current projections. Similar issues are surfacing in Germany and elsewhere on both 3G and 4G (LTE) networks.

To hear O2 UK and other mobile operators tell it, their options for dealing with this spectrum capacity constraint include acquiring more spectrum (where feasible), re-farming 900/1800 MHz spectrum currently reserved for GSM 2G services to 3G mobile broadband (again, where feasible and approved by regulators), migrating to LTE (also only where feasible and approved by regulators), and adding network density in the form of small cell base stations.

Small cells - and especially non-femto outdoor small cells that bolt to street lights and other urban infrastructure - absolutely need to be part of the solution. Without this geographic and spatial expansion (deepening) of the RAN footprint, the network won't be able to deliver the necessary bits per hertz per square kilometer. But here's the rub: all these small cells still need to be backhauled. Fiber access, multi-point microwave and xDSL will only go so far in addressing the problem due to availability and cost. Until those problems are solved, network/RAN optimization will play a critical supporting role helping manage the underlying traffic volumes.

As we highlighted in a recent post, multi-device data plans remove a formidable hurdle to wider consumer adoption of tablets and other media-friendly mobile devices by creating a pool of shared data accessible to all of the subscriber's devices and all members of that "family". For operators, the introduction of shared/family data plans is expected to accelerate sales of mobile-enabled devices, increase the average data revenue per user, and ease the migration to LTE, premium data plans and tiered offerings as users become increasingly hooked on higher speed access to satisfy their multimedia viewing habits.

So what does all of this mean to mobile backhaul networks, where the capacity gap is most apparent and is impacting service quality for many subscribers? A report issued by Comscore in early June indicates, not surprisingly, that tablet users are nearly three times as likely to watch video compared to smartphone users. Given that tablets already generate approximately 5x the data traffic compared to smartphones, even a small uptick in cellular-enabled tablet users portends a significant jump in traffic volumes - and usage patterns - on mobile networks.

While most tablet connectivity is via WiFi today, more widespread availability of these flexible, multi-device, pooled data plans is sure to spur adoption of 3G/4G tablets. In order to meet the capacity challenges associated with this steady traffic growth while maximizing the revenue opportunities and service expectations of these multi-device plans, operators will need to continually refine their network optimization, data optimization and even signaling optimization strategies from the device all the way to the Core - and especially in the high-cost access network.

An interesting example of this trend cited by the author is Viki, an international video site that relies on an army of fans to provide free - and virtually instant - subtitled translations of popular shows and other video content into dozens of languages. By working with content owners and leveraging the power of the Internet and IP TV, Viki enables everything from head-end hits in Brazil to niche long-tail programming like Japanese Anime, originally offered only in their 'home' markets, to be delivered seamlessly anywhere. By crowd-sourcing translations, Viki has dramatically lowered the cost and speed to open new markets to curious non-native speakers. The result is a whole new level of content exposure and immersion not previously possible.

Reality TV is another example that comes to mind: the shows are cheap to produce and their human interest themes (some would say freak-show antics) have universal global appeal, making them natural candidates for wider distribution.

This 'torso' phenomenon of content distribution expanding beyond traditional borders will no doubt have far-reaching effects on traffic patterns in both fixed and mobile networks all around the world.

Further evidence of this migration from basic wholesale M2M connectivity toward feature-rich service enablement includes recent operator announcements around 'connected devices' and services, the industry trade group GSMA's Connected Life and new business models initiatives, and a burgeoning M2M ecosystem. Judging from recent mobile broadband history, this last factor is likely to have the most profound effect on the M2M industry's success.

Qualcomm recently announced that third-party vendors have released more than 100 cellular and connectivity solutions for this emerging Internet of Everything/M2M ecosystem based on its chipsets. Qualcomm has set their sights on enabling everything from entry-level applications such as smart meters and asset tracking to high-end applications requiring mobile broadband capabilities such as automotive infotainment and digital signage. Although most M2M applications are currently low bitrate, low profit margin endeavors for operators, and although automotive M2M applications have not generated significant traffic volumes or revenues to date for operators, 'connected car' infotainment stands out for its long-term potential to drive significant average revenue per device. Industry research and forecasting firm Machina Research expects the automotive sector to generate over $200 billion in M2M revenues by 2020, but they reckon infotainment alone will account for over 80% of the data traffic and 70% of mobile operator service revenues from this sector.

Why? The automakers are positioning infotainment as a premium add-on option, which is likely to spur demand among the early-adopter/influencer demographic who view seamless delivery of cloud-based content and services right into the passenger compartment as highly appealing; premium add-ons tend to become mainstream over time (remember when air-con, audio and sat-nav were expensive options on luxury cars?); infotainment requires HSPA+/LTE connectivity to work properly, and since most operators are already successfully making the transition from unlimited to volume-based service plans, they are well-positioned to reap any rewards from consumer uptake of this technology.

This shift up the M2M value chain in many ways mirrors operators' 3G network build-outs of a decade ago: an initial focus on basic 3G coverage and low-speed data with a limited number of clunky devices offered, evolving to an emphasis on network capacity expansion and true mobile broadband delivery to a wide range of smart devices and through a variety of service plans. And just like in the case of mobile broadband adoption, M2M's success in transitioning to the Internet of Everything will hinge in large part on the health of the surrounding ecosystem - and especially on the availability of innovative devices and services.

As these services mature and more rich media content finds its way out to connected devices (and especially into connected cars), mobile operators will have to maintain a tight focus on network traffic optimization as close to the network edge as possible to prevent these services from overwhelming network resources. If they are successful, we can look forward to the distraction of broadband connectivity from the dashboard while we're stuck in that traffic jam on the motorway!

Adaptive Video Streaming (AVS) stems from the efforts of content owners and distributors to ensure a consistent user quality of experience in environments where the content owner or distributor does not control the end-to-end network performance. In general, AVS tries to match the video quality level (encoded bit rate) to the available bandwidth of the network in near real-time. When high bandwidth is available from the network, a higher quality version of video is sent to the client; conversely when only lower bandwidth is available, a lower quality version (lower bitrate encoding) of the video is sent. This contrasts with how most Internet video is distributed today: a single quality level that is progressively downloaded but cannot be adjusted for changes in network bandwidth (a major problem with mobile networks in particular due to the variability in the air interface). This results in frequent video hangs and abandonments and can result in concentrated periods of network congestion that affect all users in that part of the mobile network.

How does Adaptive Video Streaming work?
There are several popular AVS implementations in the market today, each with its own particular features and clients. These variants come from major application vendors such as Microsoft (Silverlight), Adobe (Dynamic Streaming) and Apple (HTTP Live Streaming), and from popular content providers such as Netflix, ESPN, BBC and HULU to name just a few. In general, commercial AVS deployments to date have all implemented a variation of stream-switching which creates multiple versions of the same video file, encodes them at different bitrates, and splits them into multiple segments. The video client is informed of the different video rates available (and the respective segment names) when the user selects a video. The client then selects the video segment that corresponds to the bandwidth available to the client at that moment from the network. As this rate changes over time, the client dynamically selects subsequent segments that best match the available network rates. An example below shows the video bitrate and quality (resolution) levels available for a video from Akamai's HD Network, a commercial AVS-enabled CDN service.

As you can see, this video can be delivered at data rates ranging from 300 Kbps to 3.5 Mbps. Experimental results show that AVS reacts to changes in the network rather efficiently and can reliably prevent the viewing quality degradation frequently experienced with progressive downloads.

Does AVS help the mobile network?
As AVS becomes more widely deployed it will have a material - and mostly positive - impact on mobile networks, however some issues will persist:

• AVS dynamically adjusts OTT video contribution to congestion and provides relief during congestion. However, AVS is "greedy" and generally attempts to deliver the highest bandwidth possible to an individual user. In a mobile context this might result in continued congestion as AVS users compete with other users (including progressive download video users) for limited resources.
• AVS, as currently implemented, is focused on maximizing the experience of specific users watching specific videos from OTT video content owners and distributors. It has no knowledge of other network users, nor does it know the localized congestion state of the network. This can result in unfair and unbalanced network performance from a mobile customer perspective.
• AVS is primarily intended for wireline broadband delivery where overall bandwidth tends to be higher and bandwidth variation is much less than in a mobile network where bandwidth is often limited, is shared dynamically among many users, and is dramatically affected by mobility and RF variability in real-time.
• AVS may complicate or reduce the effectiveness of existing optimization approaches - such as caching and video optimization solutions - by creating multiple video encoding rates and a plethora of segment files that caches and optimization platforms may not be able to handle effectively.

Ideas for improving AVS in the mobile context
Careful coordination of AVS deployments between content owners and mobile network operators has the potential to significantly improve the user quality of experience, lower network costs, and increase service differentiation. Approaches that should be considered include:

• AVS-enabled caching and CDNs - the mobile operator deploys an AVS-enabled solution that caches OTT AVS content and uses operator policy and network state information to control what bitrate and quality levels are provided to subscribers.
• AVS and mobile operator policy coordination - the mobile operator provides additional policy information to the AVS server on the maximum bitrate and quality level permissible per device type, billing level, service type, and localized congestion state. Based on this information, the AVS server only advertises a subset of the total bitrate and quality options to the client. This could also be accomplished using a proxy application that intercepts and changes AVS settings based on mobile operator policies.
• Standardized AVS protocols and interactions to facilitate new services and business models - this process is underway in 3GPP and MPEG working groups through the Dynamic Adaptive Streaming over HTTP (DASH) standard (which is part of 3GPP Release 10).

Adaptive Video Streaming is a rapidly emerging technology that could become a remedy for the explosion of OTT mobile video or another thorn in operators' sides. Mobile operators must look at ways to leverage the advantages of AVS to improve network performance in the constrained and complex mobile environment. Solutions that can adapt to the changing content patterns created by AVS and can provide unique information about localized congestion will have the best long-term value for mobile operators.

For more detailed experimental results of AVS check out:

• http://www.cc.gatech.edu/~dovrolis/Papers/final-saamer-mmsys11.pdf
• http://c3lab.poliba.it/images/d/d8/Akamai_wima2010.pdf

From an end-to-end service quality perspective (i.e., from user device to content server in the cloud and back), it is hard not to overlook the dependency between the iPad's mobile broadband feature set (which also includes HSPA, HSPA+, CDMA EV-DO, and WiFi) and its cutting-edge retina graphic display technology. After all, filling all those pixels (3.1 million to be precise) and realizing the full ocular value of this and other soon-to-be-released high-resolution mobile displays requires significant bandwidth. While the moderately sight-impaired among us will no doubt appreciate the ability to read emails with ever more "razor-sharp" clarity, it is safe to assume that most of the data streaming to these high-resolution displays will come in the form of HD video. It is also safe to assume that once accustomed to the visual perfection offered by devices like the new iPad, users will develop, Darwinian style, an increasingly heightened sensitivity to quality impairments.

Caught as they are in a conundrum between promoting robust subscriber devices and fending off the OTT content providers that bring these devices to life, mobile operators understand more than anyone that service quality depends on many links in a complex and often fragile chain of inter-dependent delivery elements that includes networking equipment, bandwidth management technologies, billing systems, physical communications facilities - and especially the wireless air interface. Delays, breakdowns or inefficiencies in the flow of content along that chain can easily impact the quality of an operator's service offering and, thus, their customer standing and competitive differentiation.

When it comes to performance improvements, LTE offers increased spectral efficiency and reduced end-to-end latency. Clearly, this improved use of available spectrum represents an important performance boost, and the progress demonstrated in LTE deployments to date has been compelling. But bolstered by early stage successes and self-serving "breakthrough year" promulgations, the chorus of LTE hype frequently overlooks or diminishes the importance of other factors affecting performance, not least of which is the backhaul portion of the network. Simply put, LTE does not solve the mobile backhaul problem. To the contrary, it can aggravate capacity constraints due to its significantly faster air interface and, depending on the spectrum used, the high number of base stations that need to be deployed to provide adequate coverage. With this increased network throughput comes additional stress on the backhaul. Pulling fiber to all these cell sites will remain impractical for most mobile operators due to simple economics and other technical challenges (e.g., availability) associated with fiber deployment. In most cases, operators will look to leverage existing backhaul assets when rolling out LTE, resulting in many instances where multi-generation traffic (3G and 4G) will share the same backhaul facilities. This will only compound traffic congestion in mobile access networks and further raise the specter of end-to-end service degradation. As a result, backhaul planning and dimensioning strategies have become an even more critical priority for operators investing in LTE. Fortunately, the mobile broadband ecosystem has produced a number of traffic optimization solutions that enable mobile operators to address data-driven congestion pain points, including RAN-based solutions that facilitate seamless migration from 3G to 4G and beyond.

It is only a matter of time before Apple unveils its LTE-enabled iPhone. With iPhone market volumes significantly higher than the iPad, this event will mark yet another industry milestone, engendering even greater levels of service quality expectations among devout iPhone users looking to push the boundaries of the mobile broadband experience. Operators and Darwin take note.

Mobile World Congress saw a flurry of announcements around 'connected device' (aka M2M) operator initiatives and customer applications. Unlike previous events, and in contrast to the hot-topic-of-the-moment herd marketing that tends to prevail at this major industry event, there was refreshing evidence of mature M2M projects underway and something actually happening in the space after an extended adolescence. Vendors, service providers and operators are actively collaborating and executing on real, revenue-generating services. In terms of the Gartner Group's technology Hype Cycle, M2M appears officially to have progressed from the 'Peak of Inflated Expectations' through the 'Trough of Disillusionment' and onto the 'Plateau of Productivity'.

Jasper Wireless, an M2M platform provider, claims 180 million connected devices were active as of the end of 2011 and projects steady 30% growth in the sector. The GSMA, along with Machina Research, released a study during the show projecting the connected devices market will be worth $4.5 trillion by 2020. Ericsson's CEO used his keynote speech to reiterate that they project over 50 billion connected devices in the same time frame. According to the GSMA and Machina, connected autos, mobile healthcare, connected homes/buildings, pay-as-you-drive auto usage/insurance and smart metering all rank as top revenue-generating applications.

Several operator announcements stood out, lending credibility to these projections. Vodafone Germany announced a deal with BMW to outfit all new BMWs with a mobile data connection and a SIM card. Beyond the obvious benefit of providing instant emergency services in the event of an accident, this collaboration has the potential to open up a wide range of 'connected life' applications - including in-car entertainment. Qualcomm's mobile health business unit, Qualcomm Life, is collaborating with Orange's Business Services M2M operation to support Qualcomm's 2Net remote medical device monitoring service across Europe. Similarly, Vodafone is teaming up with US-based Boston Scientific, a major medical devices manufacturer, to develop remote health monitoring solutions. And the list goes on.

As we have said before, the connected devices, network and services ecosystem is clearly coming together and building business momentum: operators are learning how to make their networks more M2M-compatible to exploit this emerging (and potentially massive) mobile data opportunity. And it is encouraging to see the operator community moving from simply delivering wholesale connectivity to third party M2M providers to recognizing the long-term opportunity in delivering fully-managed, cloud-based services across wide regional and international footprints.

Two operator challenges going forward will be raising ARPD, or Average Revenue Per connected Device, in what has historically been a very low revenue service arena, and dealing with the slow but steady ramp in data volumes as these services evolve and start to compete with smartphones, tablets and other consumer devices for mobile broadband network capacity (think in-car entertainment on the go). Addressing the first challenge will require process optimization (i.e., fully automated and customer-controlled M2M service delivery). And as the services evolve and more rich media content finds its way out to these devices, operators will have to devise new charging schemes to capture additional revenue and offset higher costs.

Addressing the second challenge also will require network traffic optimization. What today represents hundreds of thousands of short, low bitrate data sessions that don't really tax the network (transported mostly over operators' 2G networks) could quickly evolve into billions of 3G/4G data connections representing a capacity-consuming mix of short and long, low bitrate and high bitrate network sessions. A significant percentage of these connections may be transporting very similar content too - a good reason to optimize at the content layer and as close to the network edge as possible to prevent these services from overwhelming network resources. And this is before taking into consideration the dramatically increased signaling loads that will result from billions of additional connected devices on the network.

Wi-Fi offload is one of several tools mobile operators are using to help manage growing data traffic volumes in their macro 3G/4G networks. Recent industry initiatives are expected to streamline the current cumbersome discovery and connection process for both on-net and roaming users. And the Wi-Fi industry also has set clear price/performance standards that the 'small cells' industry (vendors of femtocells, picocells, microcells and metrocells) will struggle to match for the foreseeable future. These factors all contribute to a favorable return on investment outlook for Wi-Fi, however a number of operational and business issues must be overcome before it can become a truly 'frictionless' option for users and a significant data off-ramp from the macro network. Even then, evolving broadband device and usage patterns will prevent this form of offload from solving the macro network data problem entirely. And an over-reliance on Wi-Fi to expand service coverage and contain data delivery costs could have the perverse effect of lowering customer satisfaction and impacting operator revenues.

Wi-Fi has the advantages of being fast, well-suited to today's cloud-based 'always on' Internet content, generally simple to operate, relatively cheap to deploy and relies on free unlicensed spectrum. And it is especially well suited to the bursty, high-bandwidth video streaming that is coming to dominate fixed and mobile broadband networks. Most 3G/4G operators are already pursing Wi-Fi-based macro offload strategies in one form or another to exploit these advantages; KDDI, Softbank and China Mobile alone are pledging to add millions of hotspots over the next few years. In contrast to Wi-Fi access, which involves providing access in places beyond the reach of the macro network or improving in-building coverage, Wi-Fi offload is all about cost containment. The main objective in deploying the latter is to divert broadband data off the capacity-constrained macro network wherever and whenever possible in order to regain control over a rapidly growing component of operational expense: the cost per gigabyte served.

Building a carrier-class Wi-Fi network that can effectively perform this offload function without impacting the user experience requires:

• 'Semi-automatic' hotspot discovery, connectivity, user authentication and billing (just like cellular networks are today, but with the distinction that users be given an explicit choice to connect to a personal, corporate or free Wi-Fi access point vs. automatically connecting them to their wireless providers' preferred Wi-Fi offering)
• Stable and secure service delivery so users don't know the difference between Wi-Fi and cellular connectivity
• Broad coverage footprints in urban areas or locations where significant volumes of mobile data are consumed
• Expanded roaming agreements and footprints to ensure fluid coverage as users move around - and across different operator networks
• Ease of integration not only with the existing macro and Core networks (for policy, billing, etc.) but also with future small cell network extensions (i.e., transparent session handoffs between the Wi-Fi and cellular domains - especially when the user is mobile)

But the lack of common discovery and user authentication schemes, coupled with limited coverage, renders Wi-Fi an inadequate substitute for cellular service in the near-term. Various industry initiatives - including the IEEE's 802.11u, 3GPP's Access Network Discovery and Selection Function (ANDSF), the Wi-Fi Alliance's HotSpot 2.0, the NetGen Hotspot (NGH) championed by the Wireless Broadband Alliance and others - seek to standardize hotspot enhancements, add device-side intelligence and make overall network integration more seamless to address these general shortcomings. And unlike previous failed attempts at providing a fully integrated Wi-Fi-based alternative such as UMA, Wi-Fi offload is not intended to replicate existing 3G/4G voice and messaging services in a way that requires specialized devices.

Hotspot enhancements include broadcasting additional service discovery attributes beyond just the SSID and extending range using lower frequency spectrum. For example, 802.11u has the potential to greatly expand service coverage anywhere low frequency (470-698 MHz) spectrum is available (which is unfortunately not the case in many regions today). On the device side, SIM-based authentication is already in use by some operators and promises to simplify Wi-Fi access, improve security and more tightly integrate it with Core network policy functions. Establishing a common model for integrating Wi-Fi and cellular networks and supporting seamless roaming are proving to be more challenging issues. Even an extensive Wi-Fi build-out combined with 3rd party provider roaming agreements cannot guarantee fluid coverage as users move around. Without the opportunity to relieve the strain on the macro network in all locations where the heaviest usage takes place, operators cannot make the most efficient use of this alternate resource.

A recent report by Analysys Mason comes to a similar conclusion: operators cannot strictly rely on offload to outdoor Wi-Fi because almost 95% of tablet and 70% of smartphone usage occurs indoors. All you have to do is look at Apple's 2011 "mobile PC" shipments, nearly 80 percent of which were iPads (and of those, the vast majority were Wi-Fi only), to understand why. Evolving device and usage patterns will continue to move the broadband target for operators. Perhaps more importantly, the flip side of an over-reliance on Wi-Fi access and offload poses another risk for operators: lost revenue when subscribers don't exceed their cellular data caps because they can shift their consumption exclusively to unmetered Wi-Fi hotspots¹ once they hit their cellular limit; users spurning their 3G dongles altogether because they know Wi-Fi is widely available (the mix of tablets vs. PCs in Apple's recent shipment numbers could portend further dongle erosion too).

The struggle then becomes how to recapture these revenues in a way that doesn't alienate users and increase churn rates. The reality is users don't care about offload; they care about convenience, speed, cost and ubiquitous access, to say nothing of the quality of the viewing experience. It remains to be seen whether carrier-grade enhancements and operational refinements to Wi-Fi will eventually convince subscribers to view Wi-Fi as a service and all that implies (ubiquitous, fast and reliable connectivity; but not free). Or will they stubbornly cling to considering it a convenience ('good enough' availability, speed and reliability; should be free)?

Simply deploying Wi-Fi as a data management strategy is not a panacea for mobile broadband. Wi-Fi has to evolve into a true service before users will spend enough time on it - and therefore consume enough gigabytes of data on it (vs. the macro network) - for operators to reap the full offload benefits. Nevertheless, as Sycamore and others in the industry have pointed out, Wi-Fi must necessarily be part of a comprehensive data strategy encompassing content and application-level traffic optimization, network capacity enhancements, policy controls, tiered data plans and partnering with device makers and content providers in order to achieve the necessary operational efficiencies underpinning profitable and sustainable services.

¹AT&T's 30,000+ Wi-Fi hotspots in the USA allow unmetered access for subscribers.

As we've highlighted in previous posts, the rise of mobile video traffic and concentration of the videos among fewer content sources and distribution points, represents an opportunity for optimization technologies that exploit repeat traffic patterns in the RAN where congestion poses the greatest risk to positive subscriber experience. While video content concentration trends may vary market-to-market, the impact on capacity planning in an operator's network is the same.

Arieso, a location-based services infrastructure vendor, recently published a smartphone market study with several interesting insights into evolving user behavior as smartphone penetration ramps up in developed markets. Arieso thoroughly cataloged how and why average smartphone users are steadily increasing their data usage. The smartphone statistics in their press release were widely covered by Bloomberg and other media outlets, but one of the most revealing insights from their study was largely ignored: the amount of 3G dongle/modem traffic being generated compared to their iPhone 3GS benchmark. The reported usage numbers – reflecting increases since 2010 – tell the story:

Uplink data volumes:

• 3G Modems (various): 2654%
• HTC Desire S: 323%
• iPhone 4S: 320%

Downlink data volumes:

• 3G Modems (various): 2432%
• iPhone 4S: 276%
• Samsung Galaxy S: 199%

As the industry well knows, dongle-equipped laptops are consuming the elephant's share of mobile broadband traffic volume – Siri or not! Arieso's own data indicates this consumption approaches 10x the downlink volume of the iPhone 4S; the greediest 1% of users (over 65% of which are dongle users) now consume over half of downlink data volume. Yet the 276% increase among iPhone 4S users was regarded as the 'big news item' by most media outlets that ran the story. 3G dongles may not inspire headlines as readily as the latest smartphones, but for network operators this class of device represents a disproportionately clear and present problem.

Value propositions for vendor solutions that optimize mobile services typically fall into three categories:

• Operational efficiency and savings – by optimizing application, data, voice or video content these solutions reduce the need for investments in mobile core or radio access networks, thus reducing CapEx and OpEx
• New Service Creation – by providing more control and granular visibility, optimization solutions can enable new services that operators can sell or differentiate their offerings
• Improved Customer Experience – by making networks more efficient and avoiding congestion and poor application performance, optimization solutions can improve how customers perceive network performance, thereby reducing churn and increasing customer satisfaction with their operator

The following diagram illustrates one potential model for understanding the relationships among customers, users, applications, service providers and mobile networks as they pertain to measuring and assessing quality.

It is evident that technical measures of network Quality of Service (QoS) represent only one of several inputs to the overall "customer experience" equation. Optimization solutions enhance and improve the traditional QoS performance of existing network deployments. These solutions affect the performance of mobile applications by, among other things, improving throughput, lowering latency and jitter, reducing dropped packets, reducing connection times, increasing coverage, and improving data call success rates.

The user experiences these benefits through the lens of applications they are using on their smartphone, tablet or other mobile device. The measures of Quality of User Experience (QoUE) vary widely by application and device type and tend to be subjective. The most formalized examples of QoUE measures are based on Mean Opinion Score (MOS) tests from the voice and video world. MOS attempts to quantify the user's perception of the quality of a voice call or a video. QoUE can also be broadened to other human factor metrics about the applications themselves (e.g., HTML page abandonment, bad links, shopping cart abandonment, slow search results, etc.) which are beyond the control or influence of the mobile network itself.

One important distinction is that just improving the quality of the user viewing experience and application performance does not guarantee the overall customer experience will improve in all cases. In the above model, the customer is separate from the user. In the case of a consumer these might be the same entity (i.e., you!), but it could also be an enterprise purchasing services from a mobile operator on behalf of its employees. The metrics for Quality of Customer Experience (QoCE) are much different than QoS or QoUE. QoCE focuses on the business level interactions with the service provider (e.g., service mix, pricing, billing, and support) and is measured through metrics such as churn rate, ARPU, and call rates into service centers.

So given the complex and indirect relationship between optimization solutions and a holistic customer experience assessment, what are some of the best ways to measure and assess the value of optimization solutions from a QoUE perspective? First, these solutions should continue to measure traditional QoS metrics (throughput, latency, etc.), but also expand to include metrics that focus on measuring congestion and the avoidance of congestion at critical points in the network. Congestion is the most direct cause of poor network and application behavior – and by association QoUE. Optimization solutions should not only improve network performance but also provide a multi-layered measurement of congestion onset, congestion duration, and congestion impact and then report important metrics such as:

• Minutes per day above a congestion threshold
• Congested minutes avoided by using optimization solutions
• % congestion avoidance

What really matters from a customer experience perspective is to more closely tie optimization effects to QoUE improvement. Sycamore's IQstream solution provides a strong set of optimization capabilities tied with specific KPIs focused on measuring and reporting congestion avoidance in the performance- and cost-sensitive access portion of the mobile network. Armed with these KPIs, mobile operators can directly link the value of content optimization with improvements in their subscribers' QoUE.

For more information about this topic, please see Kalevi Kilkki's "Quality of Experience in Communications Ecosystem" article.

But what was different this year was evidence of the growing impact the tablet revolution is having in our lives - and specifically our shopping lives. We've discussed tablet trends in previous posts, but recent data from Compuware APM adds yet a new dimension to the picture, underlining the significant influence these new media devices are having on the retail environment despite relatively low penetration rates (11% in the U.S. according to Pew Research).

The fact that we are still in the very early stages of the tablet revolution was confirmed by comScore data released last month: traffic from tablets still represents a small percentage of overall traffic in the U.S., yet its rate of growth (from little more than 20% to nearly 30% in the last four months alone) is convincing evidence of a steep trajectory.

These trends are clearly not lost on advertisers, who are already beginning to see (according to a September 2011 Rhythm NewMedia report) increased effectiveness across multiple segments from in-stream mobile video advertising vs. traditional online video ads, with an 87% completion rate on average for mobile in-stream ads compared to 59-81% completion rates for online ads. As tablets drive increased consumption of premium video content (and especially long- versus short-form video), advertisers are also reaping the benefits of shifts in consumer behavior that include a greater willingness to sit through video ads in exchange for free content.

Retail shopping is only one of many segments in our society and culture undergoing dramatic change as a result of the adoption of increasingly sophisticated mobile devices and tablets in both consumer markets and the enterprise space. This is encouraging news for mobile operators looking for any and all ways to drive more data traffic over their networks. But along with this opportunity comes the obligation to continually deliver an optimal user experience - even in the depths of a U.S. mall on Black Friday.

My friend was visiting family at the time of the concert, but his extensive Springsteen Twitter network quickly alerted him to a bitly link to a live concert feed from the venue. Within seconds, he was connected remotely to the Ustream-hosted video feed (www.ustream.com). Excited about the concert and wishing other family members could enjoy it as well, he then deftly shifted from watching on his iPhone to watching on his relative's 3G-equipped laptop connected to an HDTV screen and stereo system. For the next few hours, everyone was able to rock to a mobile broadband-enabled concert experience.

While short-form, on-demand video streaming continues to be the dominant form of Internet video consumption, the live-stream video market is in the early stages of an upward trajectory. This content form, not surprisingly, reflects a myriad of market forces driving a seemingly infinite range of consumer interests and commercial prerequisites. For example, while online auto enthusiasts partook of the live HD-streamed Internet launch of Chevrolet's all-new Malibu earlier this year (an industry "first" according to the GM press release), a much larger group of Internet video consumers (by an order of magnitude) intently watched live, around-the-clock video coverage of a family of bald eagles in Decorah, Iowa (the most-watched live Ustream channel at approximately 100 million views during May, 2011). Regardless of whether you consider the big, mainstream events (e.g., 72 million total YouTube live streams for this year's Royal Wedding) or the small events like my friend's concert experience, they all add up to more demand for network capacity. Given that a significant proportion of this type of video is now originating from (and often streaming to) mobile devices, this trend portends additional sources of congestion in mobile networks - and especially at the network access edge.

Among the market forces driving this live-streaming phenomena are several user-generated aggregation sites including Ustream, Livestream, Justin.tv and YouTube. All the top live entertainment, news and sports sites offer live-streamed content too: Sky TV's Sky Go broadcasts of English football matches; BBC's iPlayer-based live streaming of newscasts; and ESPN3's live-streaming of basketball, soccer and baseball games. According to their recent corporate blog post, Justin.tv pushes at peak usage about 250 Gbps of video (the equivalent of 23,936 standard DVDs per hour) to over 30 million viewers. They also claim their iPhone and Android applications have been downloaded a collective 4.5 million times. The convergence of these broadcast platforms with live-streaming and social media (e.g., Ustream's Facebook application that enables live streaming from your Facebook page) will contribute to the further evolution of data traffic patterns and put yet further demands on mobile network capacity.

As the live-stream market picks up, mobile operators need to prepare for this dynamic shift in traffic patterns and subscriber viewing habits (remember: my friend's Springsteen video stream was delivered from that nearby 3G base-station for approximately two hours!). Mobile broadband optimization in the Core and Access networks will be critical to an operator's ability not only to deliver but also to capitalize on consumers' growing desire for live-stream content. If operators hope to more effectively monetize their networks - particularly as it relates to their highest-value subscribers - then traffic optimization and better conservation of network resources will be critical to delivering the differentiated services, consistent service quality and capacity needed to attract new subscribers and mitigate churn along the "Thunder Road" of mobile broadband.

The tablet's 'connect on the go' attributes - easy portability and near-instantaneous boot up thanks to lightweight, smartphone-derived operating systems (OS) - coupled with a rich visual experience and natural gesture-based inputs have quickly made tablets the default gadget for casual browsing at home and the local coffee shop, and a versatile platform for the office. It has also become an essential travel accessory for many business and pleasure users, replacing the ubiquitous laptop at a fraction of the weight and space. No doubt the degree of personalization possible - i.e., the ability to load your tablet with any number of clever and engaging apps to suit your personal/professional interests and needs - is a big part of this appeal. Multi-screen TV/video experiences such as HBO-GO, ESPN3, Sky-Go, Netflix and many others are ideal for the tablet form-factor too.

Both Forrester Research and ComScore recently observed that tablet users spend more time browsing the web and are more inclined to spend money online compared to smartphone and laptop/PC users. The Economist reckons this is because tablet owners tend to be more affluent, and the immersive tablet experience creates greater buying impulses. ComScore's research affirms this point (at least in the US): as of August 2011, nearly half of tablet owners belonged to households earning $100K or more. This demographic represents a lucrative opportunity for mobile operators, advertisers, content providers and others serving the demand for personal connectivity.

Enterprises are aggressively adopting tablets for several reasons including the wide range of business applications for gesture-based technology, signature capture, image generation and markup, etc. Another compelling attribute is the ability to use off-the-shelf hardware with a consumer (read: rapidly falling) cost curve. But probably the most important attribute is the desire of enterprise users to have the same light, portable and versatile platform at work as they do at home. Apple's CEO Tim Cook claims over 90% of the Fortune 500 are testing or deploying iPads now, as are a number of major hospitals around the US. On the content side, HTML5-based platforms like Pressly, Zite and Flipboard are helping redefine online publishing by enabling content owners to quickly convert their publications into tablet-friendly multimedia websites supporting the analytics, full screen ad insertion, e-commerce, mapping and other essential functions associated with traditional websites. During their iPad app launch this week, Facebook took time to extoll the cross-device value of HTML5 to their mobile platform strategy. These developments underscore the influence tablets are having across numerous industries and the growing importance of these devices to enterprise business models.

Another self-reinforcing tablet trend is the staggering breadth of available apps and the rate of proliferation of new ones for the two most popular smartphone/tablet OSs - Apple's iOS and Google's Android. Convenient online app stores, low prices, 'freemium' apps with in-app purchase options and seamless downloads have turned these purchases into a frictionless process and therefore a more spontaneous (and frequent) activity for consumers. Behind this vibrant market is a healthy ecosystem of hardware, OS, application and service delivery developers, many coming from the mobility industry, who have turned the traditional personal computing platform and OS development models upside down and ushered in an era of much more agile development, faster product launches and faster updates. The spider chart from market intelligence firm Asymco cleverly illustrates how this new approach is compressing time-to-market at the same time it is helping expand the total market opportunity.

So what opportunities and challenges do these developments present to mobile operators in particular? Although the majority of tablets sold to date have been Wi-Fi only (notwithstanding Apple's original 3G-enabled iPad launch in 2010), if mobile broadband has taught our industry one thing it is that mobility is a big value-add for consumers. The freedom of high-speed Internet access virtually anywhere without the need for a Wi-Fi hotspot is compelling to many users. But mobile operators will have to continually improve their 3G/4G service coverage, capacity, quality - and capabilities - in order to attract new users and significantly expand the population of connected devices on their networks.

Mobile cloud computing is one capability operators are pursing to broaden their offerings and deliver integrated 'smart pipe' services for all types of connected devices. Tablets and smartphones - really apps in general - are becoming heavily dependent on cloud-based services. Music, video streaming, IP videoconferencing, social media, social gaming and personal storage are all benefitting from (and driving) this shift to the cloud. And this dependency will only increase in the future. Letting people (and businesses) store their content in the cloud and retrieve it from anywhere and any device is a liberating and empowering proposition. A number of mobile operators including SingTel with their "intelligent utility" private cloud platform, Telstra with their Connected Clouds concept for enterprise customers that spans fixed and mobile, and T-Mobile parent Deutsche Telecom, are already aggressively expanding their mobile cloud capabilities specifically to capitalize on this accelerating demand in the enterprise.

From an operator perspective, two medium-term hurdles to widespread adoption of connected tablets are capturing a larger piece of the sales channel and bringing down device costs. Recent analysis by ABI Research revealed that operators only represented 13% of all tablets shipped during the first half of the year as customers presumably opted for a visit to their local brick-and-mortar outlet to touch and feel these new devices before making a purchase commitment. As ABI points out, operators need to quickly translate their success marketing smartphones and all manner of wireless accessories into more effective in-store and online tablet campaigns if they expect to capture a larger piece of this pie.

In the same research report, ABI Research estimated only 36% of tablets sold in North America included built-in cellular connections through the first quarter of 2011. Early adopter consumers and enterprise users no doubt comprised the bulk of these sales. The good news is Android-based, 3G-equipped models are appearing on the market every month at retail prices well under $500. And Apple's strategy of value-pricing earlier product generations as they launch successors presages a 3G iPad under $500 before long. Most importantly, in contrast to the PC and mobile industry's failed attempt a few years ago to embed 3G radios in low-end netbooks and laptops, the tablet provides everything that was missing back then: a superior interactive experience due to greater computing power and graphics - and compelling apps! These dynamics will likely spur significant numbers of people and firms to make the mobile tablet plunge over the next few years. Juniper Research estimates the balance will shift substantially by 2016 when models with 3G connectivity are expected to comprise over half of all tablet sales worldwide.

But perhaps the biggest remaining hurdle to wider consumer adoption is resistance to having separate 3G/4G subscriptions for their smartphones, home PCs - and now tablets. A recent operator survey conducted by Strategy Analytics suggests that consumers with multiple devices are hungry for a single data plan to connect all of their devices. They cite Rogers Wireless, Bell Mobility, Telefonica Spain and Proximus Belgium as examples of service providers already offering multi-device plans. In the US, AT&T has widely hinted at offering multi-device 'family' plans that would draw from a pool of shared data; Verizon Wireless has announced their intention to launch similar multi-device plans. Given the expected growth of data traffic from tablets, what better way for operators to create stickier customers than by offering all-in-one plans?

Bringing all these trends and market dynamics into perspective, tablets are quickly assuming a role as an essential but supplementary multimedia consumption device rather than a direct smartphone or laptop replacement. If all the stars continue to align, these connected tablets promise to drive another surge of rich-media broadband traffic over mobile networks in coming years and may even eclipse smartphone and laptop users in terms of bandwidth consumption. So tablets represent a significant business opportunity and challenge for service providers. Mobile operators with robust networks, innovative enabling services, compelling device offerings and attractive multi-device service plans - anchored by a comprehensive data optimization strategy - will be in a good position to reap the bounty.

We are all guilty of attention-deficit surfing behavior: randomly bouncing among videos on popular social networking, news, and entertainment websites, rarely patient enough to view entire clips. Since the video server and client device tend to deliver and buffer the video well ahead of the viewing rate (also called the encoding rate) of the video itself, video abandonment can result in a waste of precious bandwidth and significant amounts of data dropped along the way.

Our traffic studies of live HSPA+ networks in 2011 confirm the dramatic effects of this user behavior. The following charts show the behavior of users streaming video sessions from a single base station. The analysis in this case is based on Flash video, the dominant video encoding format today.

The distribution chart above reveals the following:

• Only ~35% of videos are watched to completion (far right vertical line)
• 50% of videos are viewed for less than 40% of their total length
• 20% of videos are viewed for less than 10% of their total length

These trends point to wasted bandwidth when the encoding rates are compared to the transport rates for these videos in the chart below:

• For this sample of videos, the range of encoding rates is rather small (~500 Kbps to 750 Kbps), while the transport rate varies greatly (~500 Kbps to over 3 Mbps).
• Only a small percentage of video sessions match encoding rate to transport rate (<20%)
• The median difference between the encoding and transport rates is ~500 Kbps, which implies that every abandoned second of a video represents 500 Kbits or 62.5 Kbytes of wasted capacity.

Further analysis shows that an average video session wastes almost 1 MB because of poor matching between the encoding rate and the transport rate and early abandonment of the session. This significant inefficiency leads to higher levels of congestion and a degraded user experience.

The industry and operators have recognized the need to better match the transport and encoding rates and "over-buffering" of user videos - especially in times of high network use and congestion. Two approaches are currently being implemented:

1. Video Pacing - This involves the use of various techniques to better match transport rates to the encoding rates of streaming videos. By more closely managing the user buffer depth, less instantaneous transport bandwidth is required and fewer bytes are wasted if a user abandons the video. Video pacing techniques include buffering and rate-shaping in the network, split-session video servers, and TCP session optimization to rate shape the streams. These techniques require specific knowledge of the video meta-data, client and server capabilities, and network conditions in order to prevent negative user impact (for instance, choppy or stalled video resulting from buffers not deep enough to accommodate network variability).
   
   
2. Adaptive Streaming - This is separate and distinct from video pacing in that it attempts to provide the best user experience (highest video quality with lowest stalling/stopping of videos) by adaptively changing the encoding rate based on the available transport rate to the client. This approach is being actively pursued by the dominant video player vendors (Microsoft Silverlight, Adobe Flash, and Apple QuickTime) and standards bodies (HTML5). Major Internet video sources such as Netflix.com and Hulu.com already support adaptive streaming. In its most basic form, the adaptive streaming video client monitors its buffers to determine if the available network bandwidth (transport rate) is sufficient to support the video stream encoding rate. If the transport rate is low (i.e., the buffer is draining too quickly), the client requests a lower quality stream from the server to better match the rates. Correspondingly, if the transport rate is high (buffer filling too quickly), the client requests a higher quality stream.

In some respects, these two schemes are in conflict. Video pacing takes the network perspective and attempts to optimize the network resources needed to deliver a video stream whereas adaptive streaming takes the user perspective and attempts to deliver the highest bit-rate quality for a given network condition. Nevertheless, the two techniques can be used together to provide the best overall user experience for a collection of users (not just a single user) in the face of constrained network bandwidth. Video pacing and adaptive streaming represent just a couple of the many tools and techniques operators can use to better manage the unpredictable data behavior in their networks. Without content and flow optimization at critical points in the network, congestion and its resulting user impact will occur more frequently and lead to expensive over-building of the network and costly subscriber churn.

HSPA is the first mobile broadband technology to credibly compete against fixed broadband, however many subscribers still rely on a cable or DSL connection for TV and even for primary Internet access due to coverage, bandwidth, or service limitations. Vodafone's announcement implies they are confident their LTE network - which already serves 4 million businesses and households across Germany - will overcome any perceived shortcomings of HSPA and enable them to compete directly against 'fixed' services. Vodafone's instincts are right. And no doubt their experience in the large, developed German market will influence their plans in their other key markets. In a sense, Vodafone is betting fixed-mobile substitution (for all services) will finally become a reality.

And until Vodafone's announcement, most major mobile operators considered complementary fixed services as an essential bundling strategy to attract new subscribers and make existing ones more sticky. As recently as mid-2008, when it acquired the remaining 26.4% of German DSL provider ARCOR it didn't already own, Vodafone Group was aggressively pursuing this fixed-mobile integration strategy as a way to bolster its HSPA offerings. Vodafone Germany's latest move is partly attributable to the excessive fees they must pay for unbundled local loop (DSL) access, however CEO Vittorio Colao's team has been steadily divesting less profitable, non-strategic assets for the past two years, and they apparently now lump the German DSL investment into this category. This could presage a broader industry shift away from accepted doctrine about the value of fixed-mobile integration efforts.

Vodafone's decision also underscores how mobile data traffic continues to drive operator revenues and business decisions, and how, as more operators roll out and expand LTE services, many of the data traffic management challenges they face today with HSPA - most notably in the UTRAN - will persist and even amplify with LTE.

According to a recent report from telecoms regulator Ofcom providing data and analysis on the UK communications sector, 60% of teenagers and 37% of adults describe themselves as "highly addicted" to their smartphones, whose multi-purpose use is gradually cannibalizing other activities including using a PC to access the Internet, reading newspapers and books, and even watching TV.

Another online survey conducted by the Guardian echoes the findings in the Ofcom report, with 58.7% of respondents claiming that they take their smartphone everywhere they go. Everywhere. (Emphasis theirs, not mine!) And the Pew Research Center added another perspective on mobile broadband dependence with recent market data that indicates two-thirds of U.S. smartphone users sleep with their phones right next to their bed.

With smartphone and tablet adoption rates poised to grow significantly and mobile applications permeating every facet of our lives, the number of users (mobile junkies?) who develop a high level of intimacy and attachment to their mobile devices will also likely increase. This addiction not only portends heightened demand for mobile data, but also heightened service expectations as popular mobile applications, in particular video, become an increasingly important part of users' online experience.

What is notably absent from their corporate press release and website is any indication whether this new software suite will support *3G* applications. As we've seen from the smartphone revolution, signaling is only one part of the problem. As the number and diversity of active M2M devices and applications increases in coming years, what was a manageable trickle of 2G/EDGE data connections is likely to become a torrent of repetitive 3G/4G data -- especially in the health and security verticals as these M2M applications become more sophisticated and consume progressively more data bytes per transaction over time. Effective optimization schemes will be essential to manage this steadily growing traffic type in the signaling plane and in the data plane without impacting the user experience for 'human' subscribers sharing these networks.

An announcement buried in the past week’s industry news highlights a mobile application destined to accelerate a shift in user behavior and have a long-term impact on operators’ mobile broadband capacity planning. Ustream, the popular (and free) user-generated video streaming service, just launched their Android app. Ustream’s existing Viewer app provides a slick interface to watch both live and stored content on PCs, laptops and Apple devices; their Broadcaster app enables equally impressive live video uploads. The newsworthy bit about this latest product release – and what will undoubtedly drive a surge in bandwidth-consuming mobile video streaming – is that upload support is finally available for Android tablets. Ustream is clearly aiming at a big market: Google revealed during their Q2 2011 earnings call on July 14 that Android activations are up to 550,000 per day, with more than 135 million total activations to date, making it one of the most rapidly adopted technologies in history.

Expanding upload support to Android tablets will greatly expand the ranks of visual artists, couch potatoes, soccer moms and myriad others playing with their tablets, and as a result generate millions more hours of video streaming over mobile networks. The rub for operators, as the veteran telecom consultant Tom Nolle of CIMI Corp recently observed, is that the slower users move (e.g., watching or uploading video via tablets), the more likely they are to use more network capacity. Nolle accurately describes stationary tablet usage as more akin to using a bandwidth-hungry 3G-equipped PC or laptop, whereas when fully mobile a tablet functions more like an oversized smartphone that consumes bandwidth less ravenously. The iPad and competing products have proven over 100 million times to date that tablets combine the best of both worlds for users: a rich viewing experience with versatility and full mobility. It will be up to operators to manage the capacity planning implications of all this new traffic along with evolving usage patterns on their radio, backhaul and core networks.

In the past few years 'cloud' computing and 'cloud-based' services have quietly become integral to the success of the mobile ecosystem, underpinning the growth of mobile broadband and mobile apps in particular. Not only operators but also major device and applications vendors and content providers have been pursuing various cloud-based strategies to facilitate more flexible delivery of services and enhance their network capabilities. Services like Netflix, Facebook, Pandora, Google's YouTube, etc., have all leveraged the cloud so users can have instant, 'on the go' access to content.

A new wave of cloud-based innovations promises even greater possibilities for users - and opportunities for operators. AT&T, for example, recently announced it will invest $1 billion to boost its cloud, mobility and network solutions to enterprise customers - largely in reaction to the explosion of embedded wireless devices and smartphones. Australia's Optus, in cooperation with Google, just launched a bundled cloud solution based on Google Apps and targeted at small business customers that allows existing Optus HSPA mobile broadband customers to access these apps from their mobile devices without incurring any data charges. And now Apple is trying to bring the cloud to the masses and convince them to store their music, images and video in the "iCloud". Apple's service strategy is similar to that of online storage offerings from Amazon (Cloud Player), Google (Google Music) and others: make digital storage and retrieval ubiquitous, seamless and less device-dependent so users perceive it as a convenience service (a utility) and let go of the physical, localized, manually intensive and limited portability digital media paradigm of today.

Cloud-based storage and delivery will be necessary to achieve this goal, but it remains to be seen whether and how quickly consumers can be incentivized to change their usage patterns and entrust "their" content to the cloud - especially in light of several recent high-profile cyber attacks on Google, Sony and other media companies. So until these ownership and security issues are addressed, content providers' and operators' paths to monetization of these subscription-based services will remain… well… a bit cloudy.

But two things are quite certain:

• Delivering the infrastructure required on the ground and 'in the cloud' to enable these services on a fraction of the 50 billion connected devices projected by 2020 will be more complex and challenging than we think*.
• Ensuring a consistent, high-quality mobile experience will require innovative traffic optimization strategies, as capacity expansion alone will not be enough to address the data deluge these cloud-based services promise to create.

On the bright side, if British mobile subscriber behavior is any indication of global trends, devices and applications that offer fast, "on the move" internet access will continue to drive increased consumption of rich media for business and pleasure. And that will prove beneficial for the entire mobile ecosystem.

*But this shouldn't be too complex, right? After all, RIM launched one of the original cloud-based mobile applications years ago - the Blackberry - and has successfully grown it to support hundreds of millions of devices worldwide.

I can't say it came as a surprise that data usage for the top percentile represented even more startling growth rates, given the characteristics and behavior patterns of elephant users we've seen in our traffic studies. According to Nielsen, smartphone data usage for the top 1 percent (99th percentile) grew 155 percent from 1.8 GB in Q1 2010 to over 4.6 GB in Q1 2011.

And the mobile broadband beat goes on.

The chart below illustrates how quickly HSPA has grown, far outstripping the growth of GSM services in the 1990s and the predicted adoption of LTE in this decade.

Nielsen's recent snapshot from the U.S. market suggests that we are already well down the path toward the 2015 estimate of 776 MB/month ATPU for smartphone subscribers that Informa Telecoms & Media projected back in November 2010. With fast-paced innovation in mobile devices and subscribers' seemingly insatiable demand for rich media, in particular video, there is every reason to believe that the ATPU in 2015, in North America and worldwide, may in fact exceed industry projections, including the estimates many operators are currently using for business modeling and network dimensioning exercises.

In another industry view issued this week, Cisco's mid-decade scenario predicts an unquestionably important milestone: the cross-over point at which wireless IP traffic exceeds wireline IP traffic. According to Cisco, that tipping point will take place in 2015, with wireless traffic edging wireline traffic by an estimated 54% to 46% respectively.

Coupled with the disproportionate impact of elephant users on mobile data consumption, which we've discussed in previous blog posts, these trends reinforce the need for innovative traffic optimization strategies, as capacity expansion alone will not solve the fundamental challenges operators face as congestion increasingly threatens network performance, service quality, and profitability.

Several recent announcements and industry developments point to the growing importance of machine-to-machine (M2M) and similar 'connected devices' to mobile service providers' business models.

At operators such as Verizon Wireless, AT&T Mobility, Sprint, FT/Orange, DT/T-Mobile and Vodafone, "other connections", "connected devices" and "emerging devices", which include M2M and telematics-type autonomous networked devices, now account for millions of their total connections. There is no single industry-accepted definition for this non-retail category, so every operator accounts for it differently but it is clear that this segment is growing rapidly. In Q1 2011, AT&T alone added between 1.3-1.6 million 'emerging devices' and counts 12 million on its network. The operator has certified more than 1,000 wireless specialty consumer and M2M devices - such as eReaders, digital photo frames, personal navigation devices, and home security monitoring and smart grid devices - for use on its wireless network.

While many of these devices do not communicate over 3G networks and therefore don't constitute mobile broadband connections per se, the general trend of more and varied autonomous connected devices using mobile networks augurs a time (quite soon) when these devices will start consuming significant broadband-level bandwidth. Operators are taking this service opportunity very seriously and putting organizations and platforms in place to take advantage of this trend.

France Telecom (Orange) and Deutsche Telecom (T-Mobile) recently agreed to enhance cross-border M2M integration and cooperation among their subsidiaries and affiliates to encourage growth in this mobile sector.

And in Orange's "Conquests 2015" strategic plan announced in mid-2010, the operator set an objective to sell 10 million SIM cards into the M2M sector by 2015.

In the past year, Vodafone has leveraged their global footprint to significantly ramp up their M2M offerings for consumer electronics, automotive telematics, smart home/metering, security and healthcare applications, just to name a few. Sprint, which, like most other major operators, has a dedicated M2M business unit, claims its nascent M2M division is generating tidy ARPU of US$5-$10 per month. According to Geoff Martin, Sprint's Manager of Platforms who also leads their new M2M Collaboration Center, M2M services currently generate a wide but highly profitable range of revenue figures: digital signage, for example, can create ARPU of around US$150, whereas smart grids may generate less than US$1. The attractive profitability derives from a combination of lower subsidies compared to traditional handsets and smartphones and little to no customer care costs.

The proliferation of these specialized connected devices - combined with the increased scale, reach and IP service capabilities implied by cloud-based computing - are further evidence of the shift from voice-centric to data-centric networks and the ongoing surge in data traffic volumes they portend.

For mobile operators, this means a steady (some would say relentless) increase in function-specific applications driving concentrated traffic to users. Think of mobile apps like News and Sports content channels optimized for specific videos, streams, and images. Consider mobile games that require network-based interactions between players. Think of social networking apps that make it easy to upload and share pictures and videos with friends. Imagine mapping, navigation, and location applications that download common content to every user within a geographic location. These and many more examples will drive significant mobile broadband traffic to hundreds of millions of 'app-enabled' users in the coming years.

Reducing churn in the competitive mobile broadband market remains a strategic imperative for network operators as subscribers increasingly opt to consume rich media content via smartphones and 3G-enabled tablets. As we've highlighted in previous posts, heightened user expectations carried over from 'fixed' broadband pose a particular challenge for mobile broadband service providers, calibrating subscribers' views of their providers and becoming one of the more important underlying triggers for churn. Changing behavior and expectations of mobile users are aptly represented in the chart below, which is based on a study that monitored actual end-users and compared their mobile vs. other browsing habits. Mobile users expect the same (or very nearly the same!) time to load a web page or video over mobile connections as they do over fixed lines. The days when mobile operators could get away with providing a "good enough" user experience are long over (think back to our old feature phones and frustrating WAP-based browsers): users expect the viewing experience to be the same whether at home on DSL, riding public transit, or sitting at the local coffee shop on their 3G devices.

While subscriber dissatisfaction comes in multiple flavors, in today's mobile broadband marketplace Quality of Experience (QoE) is a competitive advantage for operators with the network coverage and capacity to deliver a consistently high-quality, always-available data service. To do so, however, necessitates maintaining a high quality viewing/streaming/downloading experience for all subscribers at all times, including in areas and during periods of high traffic demand. As events over the past week have reminded us, "flash" traffic congestion can be triggered both in expected and unexpected ways. In either case, these events can have dramatic consequences for capacity constrained backhaul networks, both mobile and fixed, and particularly as a result of elevated video viewing.

According to a recent mobile broadband study by Acision, QoE issues are significant, with 79% of consumers polled across a range of geographies encountering issues during the viewing experience (frequent stalling was reported as being the most unacceptable in the case of viewing videos).

ROI and subscriber retention remain key decision-making criteria for operators as they attempt to crack the video performance nut before subscribers get frustrated, vote with their feet and switch providers. Successfully combating this churn and delivering a compelling (sticky) mobile data experience calls for aggressive and ongoing network optimization, and especially in the areas of radio access and backhaul optimization. Attacking the root of the congestion problem - a flood of rich media content being downloaded and uploaded at the network edge - favors content optimization strategies directly in the RAN where alleviating peak traffic loads can have significant and immediate benefit on subscribers' QoE.

As we have described several times before in this blog, 'elephant' mobile data sessions (typically generated by users with data cards, dongles and tablets) will continue to be the real drivers of mobile broadband traffic growth. And as STL Partners points out in a recent assessment, tablets could prove to be the real network congestion poison. Despite the emerging ubiquity of smartphones and other smart devices, the fact remains that a small percentage of users (the elephants) generate the vast majority of network traffic by volume relative to all other users. So why then do we keep hearing industry chatter about femtocells as a viable data offload strategy?

Putting aside for a moment the concerns about the economics of consumer femtocells (purchase costs, spectral interference mitigation costs, tech support costs and even regulatory costs - e.g., considered base stations in Taiwan), their principal benefit is purported to be enhancing voice and data coverage in areas with poor network reception. But they rely on a fixed broadband connection in order to do so. Which brings me to my point: when mobile elephants are 'in the wild' (nomadic), this data load must be carried by a combination of the radio network and public/private WiFi hotspots, so femtocells provide no benefit in this case. And when elephants are at home they are most likely offloading to WiFi and the fixed broadband network already anyway. So femtocells simply don't address the elephant in the family room - or anywhere for that matter.

Whether femtocells can be effective for HSPA/LTE data offload is unclear and yet to be proven. And because elephant behavior can have a significant impact on network performance and service quality, especially during peak periods, optimizing these elephant sessions is essential to the health of operator business models. A combination of network-based offload and optimization techniques represents a far more effective approach for operators to address the broadband data growth problem than relying on femtocells.

One of the more interesting topics surrounding the growth of mobile broadband traffic is the effect social networking has in driving traffic patterns and content concentration in mobile networks - and more specifically the impact on video traffic. There is considerable research underway in both industry and academic arenas to better understand how highly connected networks of individual users drive viewing behavior and popularity of content. A deeper understanding of these trends is important in order to effectively dimension the capacity needs of broadband networks but also the back-end data centers and storage networks that enable broadband content delivery. In addition, these insights are of high interest to content developers and advertisers. For mobile operators in particular, this understanding can assist in developing more adaptive networks that can better support these rapid and random "popularity flash" events, also called "social cascades" in social networking circles.

What is clear is that social networking sites are a dominant and growing source of video traffic. YouTube (ranked #1) and Facebook (ranked #4)¹ are two key examples of popular social networking sites experiencing significant video uploading and video sharing. Recent analysis reveals the effect that highly connected, highly regarded users have on the popularity (as measured by number of views) of video content. Studies conducted on large data sets from popular sites have concluded that:

• Video sharing sites have a high overlap of common content. Duplicated videos or common subsets are produced/uploaded across multiple networking groups with the overall overlap being as high as 25% on a given day.
• Unlike traditional websites that rely on search engines, social links represent the primary method users find content on these social networking sites. Having someone you have a relationship with and trust recommend content makes it much more likely you will view that content.
• The number of social links a user has directly correlates with the popularity of content that is posted/recommended to the user's community. For example, in one study, users with 1,000 or more followers/friends routinely drove 10,000 - 1,000,000 views of a video. This effect accelerates geometrically as a progression of overlapping social network groups; first order followers/friends connect the content to other networks, and so on.
• More influential users have a disproportionate amplifying effect on popularity. If a content owner/advertiser can identify these highly connected and influential users they will see a direct effect in popularity of their content.
• The number of social links associated with a video far outweighs (greater than 10x) its user quality metrics or comments (like/dislikes, "thumbs up/down"...). Just being uploaded or referenced by a highly connected group will drive popularity independent of quality metrics.
• This popularity progression can vary widely. In some cases it can be characterized by long, sustained growth as organic sharing occurs slowly over time. In other cases, the growth and subsequent decay can be very fast as the content is shared rapidly across multiple networks and the social network site "most watched" list picks up the demand for the video, further accelerating its popularity.

All of these studies and analysis point to the importance of social networking and online communities in selecting what content is going to be popular on the Internet at any given time.

So what are the implications for mobile broadband operators?

The social cascade phenomenon of fewer content objects being viewed by more people - coupled with the fact that this content is increasingly composed of larger, bandwidth-intensive video objects - means operators must continue to add capacity to their networks or the multimedia experience we have come to expect from our 'smart' mobile devices will suffer. Solutions like IQstream Adaptive Content Optimization (ACO) provide operators with compelling ways to mitigate the effect of popular content flash events and content concentration, as well as help rebalance their capacity planning efforts.

Below are some related references:

1. Characteristics and evolution of content popularity and user relations in social networks
2. Impact of Social Network Structure on Content Propagation: A Study using YouTube Data
3. Automatic Video Tagging using Content Redundancy
4. Predicting the popularity of online content

¹comScore Video Metrix: Top U.S. Online Video Properties by Video Content Views Ranked by Unique Video Viewers, February 2011

These innovations could go a long way toward bridging the coverage gap between femtocells and macro cells and 'greening' power-hungry mobile broadband networks. But what's missing from all the press coverage is the fact that these cloud-based approaches still require backhaul capacity: the virtual cloud requires physical transport infrastructure to connect to its terrestrial network elements. And any scheme to dynamically shift baseband capacity to offset congestion activity presumes adequate backhaul capacity already exists at the affected sites. But if backhaul capacity were plentiful everywhere we'd never experience congestion in the first place (dream on). Backhaul is a major drag on most operators' earnings per bit precisely because it is in short supply. This constraint will only become more acute as unabated consumer demand for rich media connectivity ratchets up mobile data rates. So until the backhaul problem is solved, these clever, compact and efficient alternatives to the RAN status quo will likely be limited to greenfield network deployments, campus environments, and tactical capacity extensions off existing macro sites with sufficient backhaul capacity.

While it is difficult at this stage to predict the impact of mobile video advertising on traffic volumes in the radio access network, one thing we do know is that display frequency is a fundamental component in advertising formulas, particularly in the case of agencies developing video campaigns for the Fortune 500, many of whom have already jumped into this arena. In mobile networks, the ad frequency factor translates into the dissemination of repeat ad content, much of it now professionally produced. For some of us, this was painfully evident during the February media blitz promoting the documentary "Justin Bieber: Never Say Never." I can't remember the exact number of video trailers for this movie I was exposed to while catching up on some favorite TV shows via my iPad apps, but it was enough to be the envy of the most seasoned CIA interrogator.

Embedded pre-roll video formats are favored by video advertisers since mobile viewers tend to be more captive audiences than their multi-tasking online counterparts, resulting in higher retention rates. And while the ads themselves tend to be short clips and the ratio of video advertising content to programming remains low compared to traditional media channels, these trends nevertheless portend a steady increase in mobile data traffic and further traffic concentration, as we highlighted in a previous post.

What is not widely discussed or analyzed is the growing concentration of video traffic into fewer distribution points (video oriented websites, mobile video applications…) and into a smaller number of highly popular videos. It is true that the ever-growing list of video-oriented websites and do-it-yourself video uploading services creates a "long-tail" of video content. For example, over 20 million videos are uploaded to Facebook every month. While this "long tail" is likely to grow longer and comprise more diverse content, what is also occurring is that the traffic volume is concentrating in a smaller percentage of the most-popular videos that are viewed over and over. This trend in concentration of video content is evident in some compelling statistics, including most recently the U.S. Super Bowl Championship on February 6th, 2011, which generated over 20 million viewings of the Volkswagen: Darth Vader advertisement on YouTube in just 6 days. This included over 2.8M viewings from mobile devices. In addition, the top requested BBC iPlayer show in 2010 ("Dr. Who" episode 1) was streamed 2.2 million times, while Netflix in the U.S. represents 20% of the primetime internet traffic on any given day.

In addition to these industry statistics, our investigations into mobile broadband traffic patterns illustrate this trend clearly.

We believe there are several market factors driving the video concentration trend in mobile broadband, including:

• Rapid adoption of high-definition video screens on smartphone and mobile tablet devices
• More professional videos from traditional media outlets targeting mobile users - higher quality, highly promoted
• Long form content - Movies, TV shows, sports
• Short form content - Webisodes, news clips, music videos
• Online Advertising - highly repeated, short clips, embedded in other content
• Promotion of popular videos (rapid feedback loops)
• "Top Video" lists from sites like YouTube, BBC, CNN, Yahoo, Netflix, Amazon, and LOVEFilm
• Video-specific applications for devices that concentrate content (Netflix application on iPad, BBC iPlayer application…)
• Social Networks promote video sharing that creates "Flash" events with multiple streams of the same video. The Volkswagen: Darth Vader video was referred by Facebook users over 4M times.

These important trends suggest there may be ways of exploiting video traffic concentration to better design and optimize mobile broadband networks. Technologies such as video caching, content distribution networks (CDNs) and adaptive content optimization will become more effective at reducing the effect of video growth on network congestion and customer experience. In addition, mobile operators need to evolve their networks and services to better support high quality, high demand video services by developing smarter distribution networks.

Top 10 All Time Free iPhone Apps:

• Facebook
• Pandora
• Google Mobile App
• Shazam
• Movies by Flixster
• The Weather Channel
• Google Earth
• Bump
• Skype
• Paper Toss

Not surprisingly, mobile devices garnered a significant share of attention at the recent Consumer Electronics Show (CES 2011). Tablets alone captured a top spot on nearly every pundit's wrap-up report, putting Apple's iPad on notice and paving the way for broader market penetration as attractive price points make tablets accessible to a more diverse set of consumers around the world. Given the bandwidth required for subscriber applications (think video) enabled by these powerful mobile devices, the tablet trend also promises to put more backhaul links under stress. The bandwidth requirements of first generation M2M applications - which predominantly run over 2G connections - are trivial in comparison.

But that could soon change as operators begin to consider the possibilities, such as those evidenced in OnStar's announcement of its mobile-equipped rearview mirror and a variety of LTE-based prototype applications and services, including home monitoring, electronic user guides and even forms of video chat. Although a distant drum beat at the moment, innovations in M2M mobile computing technologies and applications have the ability to slowly but steadily drive up data traffic volumes across the last mile. And fueled by entrepreneurial innovation and the promise of efficiency gains across a wide variety of industries, these applications are expected to proliferate in the future. They'll be smarter, and more autonomous. And they will, slowly but surely, start to compete for mobile broadband network capacity.

First, a few points that appear to be the current reality of DPI:

• DPI is a technology, not a product (despite how it is typically marketed). It is a core technology for classifying and manipulating traffic, but requires much more to be useful in real production networks.
• DPI is a technology that can be centralized in the mobile operator core, distributed in the network or embedded into other network elements.
• DPI helps enable four broad categories of operator activities
  • Operational Management
  • Service Creation
  • Optimization
  • Security Services (intrusion detection for instance)

DPI performs two main tasks on user traffic:

Classifies - DPI looks 'deep' into the underlying packets (past the normal IP headers) to identify users, applications, flows and anomalies. It also maintains complex statistics on the traffic it sees.

Acts - Based on configuration or dynamic policy, DPI alters the behavior of the traffic, for example:

• Dropping certain traffic
• Shaping the traffic
• Prioritizing certain traffic over others
• Re-directing certain traffic

These simple tasks (in concept - not execution) enable the operator to perform many potentially useful activities including creating new services (e.g., premium service classes and application-specific classes of service); developing more sophisticated billing models (e.g., capped plans through rate limiting); thwarting denial of service attacks; reducing congestion through shaping, etc. All of these new capabilities require significant integration and development with the carriers' billing, policy, network operations and security systems. This is the principal reason why most operators are moving slowly with DPI while they assess its business value compared to the technical and operational costs. Additional factors that have limited the wide deployment of DPI include:

• Regulatory - many telecom regulators around the world are reacting to privacy and fairness concerns voiced by subscribers and advocacy groups: should fixed and mobile operators be able to directly inspect and act on the traffic flowing between users and the Internet applications and content they are accessing? In the US, for example, the focus is on "Net Neutrality"; content and application providers want all Internet access to be neutral so no service is unfairly restricted or blocked. In a creative twist, ISPs in Australia have introduced 'shaping' to control excessive downloading from certain sites instead of imposing excess usage charges. But in most cases, control doesn't come without controversy.
• Scale and Performance - DPI is a computationally intensive process. As mobile broadband traffic volumes have exploded, first generation, centralized DPI appliances have not been able to keep up with traffic demands. As a consequence, more nodes must be deployed in a distributed architecture, which increases complexity and cost.
• Accuracy - the range and complexity of Internet- and cloud-based applications and services are increasing every day. Every one of these new applications potentially creates a new classification rule for DPI, so keeping these systems up to date and accurate is a real challenge.

So DPI is an important tool for mobile operators but it is not the "killer" solution to all their broadband service delivery challenges. DPI will eventually become an enabling technology embedded in various elements distributed throughout the network. The key architectural questions for mobile operators regarding DPI are; what role DPI should play and how to combine it with service creation, policy, billing and content optimization solutions to achieve a scalable, cost-effective and flexible service delivery platform.

For additional perspective, check out Dean Bubley's recent blog post at Disruptive Analysis.

3 UK recently announced that it is not only scrapping its current 1 GB/month data limit, but also officially allowing subscribers to tether their laptop or PC to a handset for mobile broadband access. Numerous operators around the globe either continue to offer unlimited data usage plans or have quietly reintroduced them in recent months despite a general industry migration away from unlimited, flat-rate data plans.

We're not surprised by this reversal and predict unlimited offerings will be hard for operators to kill. And that could be a good thing for the industry over the long run. As long as aggressive operators like 3 UK see opportunities to grab market share through flat-rate plans, they will pressure local competitors to follow. This is good for subscribers of course - especially those consuming large quantities of data each month - and ensures healthy competition. And it attracts a steady stream of fresh, value-conscious customers to the flat-rate operator's door. And while these price wars and market dynamics (a non-linear relationship between data traffic growth and revenue growth) also put acute pressure on flat-rate operators to drive down their cost per bit, this relentless focus on efficiency usually leads to innovative solutions to persistent problems that eventually benefit the industry as a whole.

We couldn't agree more with this conclusion. With steady improvements in mobile network performance and dramatically increasing smartphone penetration, users will naturally gravitate away from traditional, low-bandwidth forms of communication toward the high-bandwidth, rich-media, instant access afforded by mobile apps in order to satisfy their need to connect and interact. Viewed another way, increased availability of network, bandwidth and devices drives application innovation, which in turn drives new cycles of innovation and, of course, demand for more bandwidth. Mobile operators, in particular, have the opportunity and the challenge of fulfilling our seemingly insatiable demand for bandwidth.

• 365 HSPA operators in 137 countries
• 361 million HSPA connections worldwide, more than double from a year ago
• 15 million new HSPA connections a month
• 1 billion HSPA mobile broadband connections by 2012
• 120 million HSPA connections in Asia Pacific
• 90% growth in HSPA in Asia Pacific over the last 12 months

The Asia Pacific region unquestionably boasts a vibrant mobile market, with world-renowned mobile broadband networks, leading-edge services, and a sizeable subscriber base hungry for the latest and greatest in mobile innovation. And as we know, it is also one of the most diverse economic, geographic and cultural regions in the world. In mobile communications, this diversity plays out in the market landscape, which includes highly saturated urban markets in developed economies, rapidly growing and dynamic emerging markets, and under-served rural markets where many believe the next wave of growth will occur. It also plays out in the region's telecom infrastructure that mobile operators rely on to transport mobile data traffic, and which includes everything from fiber-rich urban centers to remote satellite-dependent mountain villages.

The terrestrial mediums for mobile backhaul in Asia include copper, fiber and microwave, the latter used extensively by competitive providers as an alternative to costly leased-line services from incumbent operators. If we all had our druthers, fiber would be the natural option given its high capacity and scalability. Yet although a number of industry reports have projected the steady rise of fiber-based mobile backhaul connections around the world, the reality is that fiber is simply not an option for many backhaul spans linking cell sites to aggregation sites (e.g., RNCs). This is particularly true in Asia.

While there are obvious geographic constraints in Asia that limit fiber connectivity (e.g., island archipelagoes), the economic challenges of fiber-based backhaul remain formidable. The cost and complexity equation of fiber must factor in, among other things, right-of-way negotiations, installation (imagine trenching across less than hospitable terrain such as jungles, deserts, and mountains!), site reconstruction, and time-consuming deployment schedules.

In an effort to effectively manage network costs in the face of rising mobile data traffic demands, operators need to fully leverage (or "sweat") existing network assets wherever and whenever possible. In the case of Asian mobile broadband operators, this means continuing to leverage their existing copper and microwave-based backhaul infrastructure for several more years at least while slowly adopting fiber – if and where it is economically feasible to do so. This presents an opportunity for solutions that can optimize mobile data traffic and create virtual capacity across existing mobile backhaul infrastructure.

The argument against FD is that this flood of signaling messages leads to the poor network performance many users experience. In fact, this is simply not true. First of all, most smartphones use FD, not just iPhones. Our analysis of actual operator signaling traffic shows that Android phones use an even more aggressive 3-second FD timer compared to the iPhone's 5.2 seconds. But it turns out FD doesn't really add to the network load; the state change caused by FD would have happened anyway, just several seconds later. So if we can't blame Steve Jobs, who should we blame? Try Mark Zuckerberg! Smartphones have become very useful (and well used) devices due to the incredible variety of applications (like Facebook) developed for them. Users often pause to view what they have downloaded and many of these applications stay connected with their servers (using polling or push technologies) even when the user is not actively viewing the device. Email is a good example of this. The pauses and poll times are usually long enough that the mobile device will enter the idle state. So every time the user or app wants to reconnect to the server, the mobile device must make a new radio connection, with all of the signaling that connection entails, and for what may turn out to be a relatively short time. Fast Dormancy doesn't cause this behavior, but it certainly leads to increased network signaling. So what is an operator to do?

One way to limit the amount of signaling is to use the Cell-PCH state. Cell-PCH is a low-energy state compared to Cell-DCH, requiring only 1/100th as much power, and unlike IDLE state, it retains the network connections. So when the user (or app) wishes to interact with the server again, the network needs very little signaling to return to the active Cell-DCH state. This has the added benefit of speeding up the connection, as the user will see only a few hundred millisecond delay compared to the multi-second delay that would be incurred when transitioning from Idle mode. There is one drawback, however, to using Cell-PCH with Fast Dormancy: the current proprietary use of RCRIs for FD forces the device to Idle mode, effectively blocking the ability to go to Cell-PCH.

Luckily, the 3GPP has come to the rescue. Release 8 defines a clever, simple solution that returns control to the network operator while still allowing the mobile device to manage its energy use efficiently. The solution is to add a cause code to the RCRI message. Now, when the mobile device wishes to move to a lower-energy state, it sends the RCRI with the cause code "Mobile device requested PS data session end" and waits for the network to tell it which state to enter. Also defined is a new timer, T323, which tells the mobile device how long it should wait before trying to release the connection. For more information, the GSM Association has published an excellent overview of the Release 8 Fast Dormancy feature and recommendations for how operators can best deploy it. With this enhancement, control is returned to the network operator, who can now balance the need to limit signaling with the requirements for better battery life and improved user experience.

Back to the original concern - network congestion. Is signaling the cause? Well, not really. The capability to support network signaling is a limited resource like everything else and intelligent management of this traffic is important. But the real cause of network congestion is simply the large quantities of data consumed by the applications enabled by smartphones. Finding ways to manage this flood of data is the real way for the operator to improve their network's overall performance.

If you're not familiar with QR codes, they are essentially square, 2D barcode-like images that employ various encoding techniques to deliver a simple, instant message which can be interpreted by any wireless device equipped with a camera and a simple optical reader. NTT Docomo pioneered the use of QR codes as far back as 2005 after borrowing the technology from the Japanese manufacturing sector. Today, Japan is the largest user of QR codes in the world (QR codes are the de facto standard for Japanese mobile phones); QR codes have become a ubiquitous presence at Japanese train stations, in retail locations, on outdoor billboards and even on sidewalks.

Over the past two years, QR codes have finally caught on in Australia, the US and Europe, largely due to the rapid growth of smartphones. They are being used in a range of applications to convey snippets of information quickly to mobile users, especially public service messages and advertising, and are showing up on everything from airline boarding passes to packaged foods, magazines, direct mail pieces, public spaces and on-air (TV shows). A recent article in the New York Times highlighted their use in Albany, NY train stations as a marketing experiment in part to gauge rider awareness of new media.

A reader application on the phone scans the image, interprets the service type and executes an associated instruction (e.g., "email this address" or "SMS this number"). Think of QR codes as the natural evolution of SMS short codes, providing a more convenient and automated link between 'static' media (e.g., a billboard) and the dynamic web (e.g., "come visit our website to learn more"). Today, most QR codes only encode 40-60 characters of information to accommodate the optical capabilities of average phones, but as smartphone camera capabilities improve - and encoding techniques become more standardized - 300-500 characters or more of encoded data will become the industry norm, allowing advertisers and others to deliver a rich vein of information to curious consumers in a snapshot.

Where things get really interesting is personalized use of QR codes, since anyone with a smartphone can generate them on the fly using a simple code generator app. Personal 'branding', geo-tagging, fashion, stealth social networking, political activism and minor acts of civil disobedience are just a few of the numerous recent examples where QR codes are being used to great effect.

So why, in our view, are QR codes likely to have a significant impact on subscriber usage patterns and network traffic volumes? Because mobile devices are increasingly becoming the first (and most frequent) interface users go to for new content and services consumption, and QR codes create yet another reason for users to interact with their devices and the Internet in a tightly coupled real-time data interaction. The primary 'event' of the user scanning the code with their wireless device leads to the secondary effect of surfing to the website and consuming additional content. Advertisers love this because it increases the level of consumer interaction with their brand and messages. Operators, for their part, usually latch on to anything that drives more data usage - especially if they can charge for it - and QR codes appear to be a great way to generate more data traffic.

FIGURE: The example QR code below, created by the BBC for one of their promotions, decodes as "http://bbc.co.uk/programmes"

Mobile operators are especially impacted by this video surfing behavior as upwards of 60 - 70% of mobile broadband traffic is video. Every abandoned video wastes valuable radio network bandwidth because video servers and players "pre-buffer" content well ahead of the actual playing rate, so the network must allocate this capacity whether it is used or not. In fixed broadband networks this isn't such a big deal, but in mobile networks the cost per megabit delivered (over the air) is proportionately much higher. Every abandoned video may result in 30% or more wasted transport bandwidth. As we explained in a previous post on video optimization, mobile operators have several tools to alleviate this problem - including RAN-based content optimization. Long-term usage trends may also help here as more commercial entertainment content (TV shows, webisodes, movies, music...) moves to the web, average viewing times increase, and abandonment rates decrease.

The summit was held in Turkey, a Eurasia country that has witnessed significant uptake in mobile broadband subscribers since the introduction of commercial 3G services in 2009. Turkey also currently ranks fourth worldwide in the number of Facebook users and boasts one of the largest and most engaged online audiences in Europe, two facts that portend further challenges for mobile operators as fixed-line broadband usage patterns and subscriber QoE expectations continue to migrate to 3G networks.

What was clear from many operator discussions and presentations at this event was that mobile backhaul remains a key challenge in delivering profitable, high-quality mobile broadband services end-to-end. As one operator highlighted during the presentation sessions, mobile backhaul is still, in many ways, the "Elephant in the Room," albeit one that is getting more recognition these days. In Turkey, as in other highly competitive mobile broadband markets, managing backhaul costs while delivering reliable, high-bandwidth data services will be key success factors in improving operator service delivery economics and subscriber satisfaction.

If you missed us at the Eurasia event, we'll be addressing the same challenges at the upcoming 'Bandwidth and Backhaul' conference in Kuala Lumpur, Malaysia (Nov. 30 - Dec. 1, 2010). We hope to see you there.

In previous posts we discussed the underlying effects of more mobile video traffic and surveyed general optimization techniques mobile broadband operators are using to manage this deluge of rich media traffic in the networks. In this post, we'll focus specifically on IP video optimization. Numerous approaches exist to more efficiently carry video over a mobile network, including:

• Broadcast and Multicast Video over Mobile - This approach uses mobile spectrum as a video broadcast platform. The motivation is to save RF and network bandwidth by broadcasting TV and video flows rather than spectrally inefficient unicasting. Qualcomm's Media FLO Service and Digital Video Broadcast - Handheld (DVB-H) are two examples of this technique. But neither technology has achieved widespread deployment (or subscriber acceptance) due to the need for dedicated spectrum, radio and network upgrades and, most importantly, a lack of compelling devices. In addition, the total number of broadcast sessions is limited to a few channels of pre-determined TV. And as we all know, trying to predict what videos users will want to view and when they will want to view them is an impossible task.
  
  
• Video Transcoding - This technology uses various compression algorithms, encoding schemes, and rating mechanisms to modify the source video files and streams to match the display capabilities of the device. Transcoding adjusts the video resolution, frame-rate, and compression ratios to reduce the payload required to send and display the video. This technology can improve network bandwidth utilization and render video suitable for feature phones, but it also reduces the video's fidelity and requires significant processing and storage capabilities in the Mobile Core. The rapid growth of powerful, video-capable smartphones and smart devices on mobile networks is diminishing interest in this technology.
  
  
• Video Pacing - This technology smooths out the delivery of the video download to more closely match the playback speed of the user's device. This can result in a two-fold improvement in network efficiency: first, by allowing more concurrent video sessions on the same network bandwidth, and second, by reducing the quantity of actual video bytes carried over the network if the user abandons the download mid-stream (a very common occurrence).

Video Pacing Example: By more closely matching the download speed to the playback speed - peaks are lowered, more users can be serviced and less bandwidth is wasted.

• Adaptive Video Streaming - This approach maximizes the number of video users on the network by adjusting the video quality and delivery rate 'on the fly' depending on congestion conditions and available network bandwidth. In some cases, this requires a specialized video viewer application on the user device.
  
  
• Video Caching - Basically a subset of core-based web caching techniques described in a previous post, this technology involves serving complete video files (but not streaming video!) from a platform in the operator's Mobile Core. The primary benefit is a reduction in the Internet interconnection capacity required.

A rapidly growing number of user devices on the network are now capable of delivering a full fidelity video experience, so any attempt to control data traffic that also reduces video viewing quality will be instantly noticed by users. This implies the number of transcoded video sessions, adapted video streams, and cached videos that can be delivered over the network at any given moment (let alone during peak periods) will be constrained by practical service and operational considerations.

Video optimization is one of several tools operators can deploy along with other optimization techniques such as RAN-based data reduction and service/policy management to more efficiently manage the explosion of video traffic on their mobile broadband networks. Any technique that impacts the quality of the content delivered to mobile broadband subscribers, however, runs the risk of losing favor with mobile operators over time, given subscribers' rapidly increasing expectations for a high-quality, full-fidelity viewing experience.

This is further evidence that it's all about optimizing rich media content - and especially in the RAN - given rich media's disproportionate impact on mobile broadband service delivery costs (backhaul bandwidth) and user experience (localized busy hour congestion). The rest of the data traffic is essentially noise on the link.

The quality of the radio signal is a key factor influencing subscriber data rates. If a high-quality signal is present, it's more likely a lower block error rate is present on the air interface and therefore higher data rates are possible. In a weak or degraded radio link scenario, either packet delay or frame loss can be introduced as the radio network protocols attempt to deliver data traffic in 'noisy' or high-loss conditions. Given this constantly changing network environment, a broad range of MBB connection rates and round-trip times are typically observed.

One correlation seen across multiple MBB networks is that between long duration flows and increased bandwidth connection rate. One possible explanation for this pattern is that users with low bandwidth connections avoid bulky data transfers since they don't meet quality expectations or would take too much time to complete. Conversely, those users who are enjoying high bandwidth connections may be more likely to prefer larger data transfers (e.g. file transfer, rich web content or streaming video) to either minimize the transfer time required or maximize browsing performance and/or content quality. In this latter case, both time- and quality-related tendencies are likely contributing to the longer flows seen on these higher bandwidth connections.

In contrast, short flows consistently reflect limited connection rates due to normal TCP behavior. TCP slow-start (TCP's congestion control algorithm) has a tendency to penalize short duration connections since they can never reach the available bandwidth rate. In MBB networks, this effect is more pronounced because of the increased magnitude and variance of round-trip time (RTT) delay introduced by the mobile environment. The increased RTT times in the MBB network will restrict short TCP flow rates more than on a wireline network, which can contribute to sluggish performance - especially when retrieving large numbers of small objects (such as those found on a complex WWW page, for instance).

So what can we glean from this differing user behavior on the MBB network? It's clear the mobile network is a diverse environment representing a broad array of content types, usage patterns, and network performance levels. The mobile network is more challenging to fully utilize than the wireline network since the more frequently encountered and wider variations in rate and round trip time delay due to radio effects are interpreted by content originators as network congestion (with a corresponding reduction in data sending rate).

It is also clear that the growing popularity of video content and increased consumption of it on 'smart' mobile devices is driving dramatic changes in content mix, usage patterns, and traffic volumes on mobile networks. Embedded video clips on web landing pages will also drive a shift upwards in traffic volumes as these types of sites will require more bandwidth to initially load.

Analysis of the interrelationships between byte volume, connection duration, and connection rate indicates that:

• Large byte volume subscriber sessions have a tendency to consume a large share of available bandwidth. While this pattern is desirable if bandwidth is available (to avoid possible future congestion), it can be detrimental during busy hour periods where many users are competing for limited resources.
• Small byte volume sessions suffer from the effects of TCP slow-start and high round trip times (RTT), which both limit the maximum achievable throughput.
• Very short duration sessions can be contained within a single burst, and many sessions of this type are not WWW traffic related. These sessions typically only contain small volume flows.
• Short duration sessions (between two seconds and fifteen minutes) are diverse and numerous with a wide range of application usage observed including WWW.
• Long duration sessions are primarily WWW traffic that typically enjoy higher connection rates. While not 100% correlated, long sessions + high connection rates = large volume consumption.

So what does it all mean? No doubt, the MBB traffic mix and subscriber usage patterns will continue to evolve as users discover new applications and content, devices become more sophisticated, and network capabilities improve. All of these connection trends will drive higher bandwidth expectations on the network.

As one of the sponsors of this webinar, Sycamore invites you to view an archive of the broadcast and share your comments with us.

A number of recent industry announcements highlight specifically how rapid innovation in broadband content and applications (along with the services evolving to enable/deliver them) will in our view profoundly influence subscriber usage patterns and network traffic volumes going forward.

First, in early September Google announced their Instant search feature that delivers results before you've finished typing. Google claims it can reduce the average search time by 2-5 seconds. Pundits are referring to it as the Internet on fast forward and the Internet on steroids. While Google Instant will no doubt improve our web surfing productivity, it may also have a more subtle but profound impact longer term. Marissa Mayer, Google's VP of Search Products and User Experience, says she expects consumer usage patterns to dramatically shift not just in terms of faster searches, but also in the form of more searches and an increased level of engagement with the content. The implication is that as we become more efficient (faster) users we will also tend to consume more content. If this is true, expect a rush by application developers to "instantify" their offerings.

Second, Qik, a service that lets you record and share video from your mobile device directly to the web, just published user statistics for the first time. Although they only have 3.5 million users, they claim they are adding nearly 500K users each month, their user base has grown six fold in the last year, and through partnerships their application will be embedded in tens of millions of 3G/4G devices in the coming year. Other service/application players, including live streaming firms Ustream and Justin.TV, have jumped into this space and launched similar applications for the iPhone and Android. Increasing uplink speeds on mobile networks and more capable devices are no doubt helping spark consumer interest in this form of video consumption. Conclusion: what goes up into the network as a single video stream ultimately comes back down as a broadcast of many streams over time, imposing a multiplicative effect on operator data traffic volumes.

Third, Asymco, a Finnish application developer and industry research firm, recently published a report indicating that iPhone/iOS users are downloading applications from Apple's App Store at more than twice the daily volume that iTunes songs are being downloaded. The firm says download rates for both are still growing but application download rates are growing much faster and they predict the total number of application downloads will overtake iTunes downloads by year's end (interesting stat: it only took 2.2 years to reach 6.3 billion App Store downloads versus around 5 years for the iTunes Music Store). Apple only has about 15% of the global smartphone market by unit sales today, but its outsized success among consumers has spurred strong competitive responses from Google (Android), RIM, Symbian, and other smart OS vendors. As further validation of this trend, Nokia just announced this week that their Ovi Store has 130 million active users globally, services 2 million downloads each day, and 70% of these downloads are now applications rather than music.

Lastly, mobile operator interest in m-payments (also known as mobile payments and micropayments) may finally be getting traction after numerous false starts. Dutch operator KPN has allied with the local subsidiaries of T-Mobile and Vodafone, and regional banks ABN Amro, ING, and Rabobank to collaborate on a unified m-payments system for the Netherlands. Their aim is to enable m-payment transactions using Near Field Communication (NFC) technology and embed the supporting payment software in a secure part of the SIM card. Similar joint initiatives are underway in the UK, US, Japan, Korea, and China. Assuming this technology becomes a seamless part of the mobile device, billing system, and banking network, and if consumers can be convinced of its utility, operators could see a dramatic uptake in these machine-to-machine type data transactions over their networks.

So what are we to make of all this? The popularity of 'smart' devices and a plethora of compelling applications - combined with subscribers' growing demand for new and instantly available content to consume and share anywhere and anytime - are encouraging users to interact with their smart devices more often and stay connected longer. It appears to us that the net effect will be frequently changing usage patterns and ever greater data traffic volumes in the network.

What Mobile Core Caching doesn't solve is backhaul congestion at the edge of a mobile operator's network caused by rapidly increasing data traffic. This is a result of how the technology works and where it is located in the mobile network. Web or Content or Media Caches (all similar) sit in the IP Core between the mobile data network and the Internet. The cache intercepts requests made from mobile subscriber devices and, regardless of whether the content is cached or not, sends every byte over the mobile data network, including through the RAN, to the subscriber. If that same request is made 100 times, every byte must be sent 100 times. The good news is that interconnection bandwidth is dramatically reduced with this approach and subscriber performance is improved.

So as network planners look at how best to optimize their network for mobile broadband traffic, they need to consider multiple technologies and approaches in order to address the different operational challenges mobile broadband triggers throughout their network. Mobile Core Caching is one technology that helps with the interconnection problem, but other solutions like RAN Content Optimization are needed to help solve backhaul congestion challenges.

Couple this trend with the growing popularity of powerful smartphones, smart tablets, and USB dongle-equipped laptops, and it's not difficult to imagine the perfect storm that's brewing for many backhaul networks.

• World First: Mobile Broadband Boost Puts Australia on Top of the World
• This cutting-edge service is enabled by new device technology designed to support HSPA+ Dual Channel networks and peak network downlink speeds of 42 Mbps.

Short duration flows are always considered 'mice' since the narrow time period (less than or equal to two seconds) only allows the transfer of a limited number of packets, even at high connection rates. 'Medium' duration flows are the most frequent type observed and represent a diverse collection of large and small byte volume sessions (driven by subscriber application demands and mobile network conditions). There is an expected correlation between some long duration sessions and elephant flows (since it takes time to transfer large byte volumes) but mice sessions can also be long duration (that is, low total byte volume but lasting more than fifteen minutes). A small number of very long session durations (several hours or sometimes even days long) were also observed on multiple mobile networks. These long duration sessions can indicate a 'stationary, always-on' MBB user or perhaps a small office site connection. Connection rate is also a contributing factor to session duration which we'll explore further in a future blog post.

One key distinction between the different session duration categories is their 'burstiness'. A burst is defined as subscriber session traffic flow (or flows) separated by three seconds of idle time. As expected, short duration flows are always seen as a single burst since they can never exceed the defined three-second idle burst threshold. Medium duration flows typically are observed as one to two bursts although about 25% of the medium flows are seen with greater than five bursts (multiple bursts likely indicate interactive user activity). Long duration flows are the most 'bursty', with a wide range of session burst count values seen across different MBB networks (long flow median burst count values range from four to fifteen bursts). One sharp contrast observed within the long session category: many long duration sessions (up to 25% in some networks) occur as a single burst, reflecting a long, steady (and likely high byte volume) transfer such as FTP or HTTP file download activity.

Exploring subscriber session durations provides a slightly different perspective to help understand MBB subscribers and their usage patterns. The insight gained from analysis of session durations is that different session duration patterns can emerge from a single subscriber application. In fact, the two most ubiquitous network application types (WWW and email) include session durations across all three categories, with the 'medium duration' session being the most frequently observed for both application types.

The mobile broadband traffic studies we conducted include results that consistently demonstrate power law popularity distribution (i.e., the 80:20 rule) characteristics. Most network planners are familiar with the concept of 'Elephant and Mice' user behavior that goes back to the early days of the Internet: a very small percentage of users (the elephants) generate the vast majority of network traffic by volume relative to all other users (the mice).

Analysis from one of the HSPA Node Bs observed during the traffic studies demonstrated this elephant/mice phenomenon, with 5% of mobile broadband users consuming 80% of the total traffic over a one week test period.

In another multi-site study, the following elephant behavior in the RAN (downlink traffic from the RNC to Node Bs) was observed:

• Disproportionate elephant sessions observed at all sites in the study
• Elephant sessions tended to be lengthy compared to mice sessions (median elephant session length = approximately 1.7 hours; median mice session length = approximately 34 seconds)
• Elephant sessions use multiple applications (median port count = 10)
• Elephant sessions follow daily network load usage patterns
  • Differences in weekend versus weekday patterns
  • Represent key driver of peak bandwidth loads during peak periods

Because elephant behavior can have a significant impact on network performance and service quality during peak periods, optimizing these elephant sessions can provide significant value to the operator.

I expect mobile broadband content mix and usage patterns will continue to evolve as new and different types of devices become available, network capabilities expand, new subscribers join the network, and content and applications proliferate. In future postings I'll explore the content mix and usage patterns in more detail.

There is general understanding that the relationship between IP session control, most notably TCP, and mobile networks is a love and hate relationship. TCP does an admirable job of managing congestion, efficiently servicing the subscriber, and reacting to changes in the network on the scale of the whole worldwide Internet. TCP works best in low latency, low jitter, low congestion, and low packet loss environments. However, all of these characteristics are highly variable in the mobile network depending on weather, user device power levels, interference (buildings, vehicles, hills...), network loading, number of subscribers, and types of applications being accessed by subscribers, all of which can severely impact TCP performance. What mobile session optimization products attempt is to "tweak" the session control protocols and algorithms to make TCP applications think that they are operating in a pristine network environment, promising improvements in the subscriber's user experience, including:

• Higher effective throughput (faster downloads, streaming video, higher performance gaming...)
• Faster session setup
• Less user impact from changing network conditions

In deploying session optimization solutions, operators hope to achieve:

• Better user experience, less churn, less customer satisfaction issues
• Reduced air interface "chatter" by reducing the ping-pong interactions of TCP clients and servers
• More efficient use of radio and network bandwidth

So it is clear that session optimization is one tool to help improve customer experience and to make mobile networks more efficient. However, several implications need to be considered:

• Session optimization solutions generally require client software to be installed on the mobile devices. This is an operational and customer challenge which limits the number of devices that actually support session optimization.
• Session optimization requires dedicated resources in the mobile core that must scale to 10,000s to 100,000s of concurrent sessions with attendant cost, performance, reliability, and operational impacts to the mobile network.
• Since session optimization "tweaks" traditional session control, there are conditions whereby the optimization can increase the congestion and degrade user experience when there is very high traffic or unusual network conditions.
• Session optimization minimally reduces the amount of traffic, so it does not directly impact the constraint of limited radio and backhaul bandwidth, and therefore does not change the economics of delivering mobile broadband.

To better control traffic growth, mobile operators need to explore a range of optimization solutions and focus on those that provide the best results both in performance and economics. Session optimization is just one of the potential technologies along with RAN content optimization and others.

As I see it, when mobile search increases, mobile content is sure to follow.

When technologies fall prey to the voracious hype machine, it's easy to lose sight of practical market realities. LTE is a good example. In the mobile broadband ecosystem, LTE hype has been working overtime for several years, in many ways overshadowing the significant and ongoing roll-out of 3G HSPA services around the world.

One of the reasons behind the continuing LTE hype cycle is the very real issue that the explosion of mobile data has exposed – network congestion and its associated impact on subscriber Quality of Experience and service delivery economics. Infrastructure vendors would love nothing more than to sell another major upgrade cycle of "next generation" equipment.

In the long run, LTE/4G will offer compelling opportunities for operators to redefine mobile communications and run their overall networks more efficiently. But will LTE solve the backhaul bottleneck that is already at the epicenter of the mobile broadband dilemma? And what happens when higher speed air interfaces compound the issue of congestion in the RAN? The truth is – LTE does not fundamentally solve this problem.

Content-based optimization in the RAN is a different way of tackling the bottleneck challenge facing operators today and will become even more relevant as the industry evolves toward LTE.

Chart Source: Allot Communications

Bandwidth options for mobile backhaul fall into three major categories:

• Add capacity to the existing backhaul network – this is the quickest and easiest to accomplish operationally. However, it also can be the most expensive as it does not fundamentally change the economics. An example of this approach is to add additional E1/T1 leased lines or PDH microwave capacity to a cell site.
• Upgrade to packet-based backhaul – this approach requires significant planning, time and capital investment. However, it can provide a significant leap in capacity while lowering the recurring operational cost per Mbps of incremental bandwidth. Examples of this approach include upgrading from E1/T1 copper circuits to fiber fed, Carrier Ethernet services or upgrading from TDM-based microwave to Packet/Ethernet-based microwave solutions. Both of these examples require major network, equipment, right-of-way agreements, spectrum leasing, and operational upgrades that can incur significant time and cost.
• Offload and Divert certain backhaul traffic – this approach offloads mobile data traffic to lower cost transmission alternatives. Examples of this approach include xDSL offload at the base station and Femtocells and Wi-Fi hotspots that can transport the data over much cheaper fixed broadband connections. These approaches can be effective but have cost, operations, spectrum planning, policy, and SLA implications.

While effective in some cases, these approaches are typically expensive, reactive, and ultimately may not solve the underlying challenges: exploding use of data services; the unpredictable nature of when and where congestion will occur; and the cost and time to deploy new transport. Our view is that the best solution is backhaul capacity upgrades pursued in conjunction with RAN-based content optimization. This form of content optimization:

• Slows the impact of data growth in the RAN, making network planning less reactive and investment more manageable
• Enables capacity upgrades to be deferred by freeing up and increasing the "virtual" capacity of existing backhaul links
• Acts as a "shock absorber" in the network and buffers users from the adverse effects of short duration events and peak traffic congestion

With each successive (and ever shortening!) product cycle, the device industry has come up with new and exciting 'connected' gadgets to delight consumers and address business needs. 3G dongles, 3G-enabled laptops and netbooks, tablets (a category resuscitated by Apple's iconic iPad), e-books, and smartphones equipped with 5 megapixel cameras and HD video recorders are just a few recent examples.

This market-driven frenzy of innovation has created new winners and new losers with almost every cycle. But unlike the history of the PC industry, where only a few players have long dominated the hardware and software value chain, waves of creative destruction in the mobile industry so far have led to robust competition and the promise of renewal (case in point: Motorola's recent hit Droid X handset). And with sky-is-the-limit web content and hundreds of thousands of clever, relevant, and user-friendly applications available to subscribers in an open and integrated way (e.g., app stores), the appeal and adoption of these connected devices has soared.

The irony of this success is that we are quickly reaching a transition point where the hardware matters less and how the applications, software, and services are integrated matters much more. Just ask the average smart phone subscriber on the street: many don't care that they own a "smart" phone as long as it looks and performs well and are keener on how the technology seamlessly enhances their work, life, and play.

In one sense mobile technology has become fashion. Operators realize this of course and use it to their advantage in the form of sleek retail outlets and expansive choice in devices, service plans, and subsidies to retain them.

This diversity of choice and "better/faster/cheaper now!" market dynamic is fundamentally good for subscribers and the mobile ecosystem because it fosters competition and ensures sustained innovation among operators, device manufacturers, content providers, application developers, and infrastructure vendors.

But behind the scenes, operators must do the hard work of regular network upgrades to manage the explosion of data traffic and meet heightened subscriber expectations for 'wired' broadband performance at a reasonable cost while mobile.

In future posts we'll explore facets of these innovation cycles, and in particular, how content is impacting this evolving ecosystem.

Although there is a wide range of technologies, products, and solutions that fall into the mobile optimization category, they all attempt to tackle one or more of the following areas of pain and opportunity:

• Reduce traffic transiting the mobile network
• Improve the customer experience
• Classify and control subscriber sessions

A useful way to break down the Mobile Broadband Optimization landscape is to categorize solutions in terms of (1) which of the above areas they address, and (2) where in the network they are deployed and have their strongest effect.

The following diagram shows how a mobile network is constructed and the typical locations of optimization solutions. As you can see, solutions land in three main locations:

1. Client Device-based solutions

Consisting of software that provides user-based optimization including local caching of content and session optimization to reduce latency and improve connection reliability. Often times these solutions work in conjunction with Mobile Core-based solutions.

2. Mobile Core-based solutions

These are typically deployed in data centers or in the network core, and include:

1. Caching to reduce Internet connection bandwidth
2. Session Optimization to optimize specific applications such as video or user TCP sessions
3. Deep Packet Inspection and Policy Enforcement to control and enforce policy
4. Video/Audio Transcoding to reduce the traffic sent to users by lowering the size, quality, and frame rates of traffic
5. Internet Offload to shift mobile broadband to lower cost, best-effort internet connections

3. RAN-based solutions

Solutions focused on the Radio Access Network (RAN) include:

1. RAN Content Optimization – a new category that specifically targets the most congested and cost-sensitive part of the network: the RAN or backhaul. This category specifically reduces the traffic traveling across the RAN by looking into the user content flows and optimally delivering the content across the RAN. This has the overall effect of reducing congestion, lowering transmission costs, and improving the customer experience during peak traffic periods.
2. Alternative RAN Transport – not technically Mobile Broadband Optimization solutions but they can provide higher bandwidth and some cost savings to operators. These alternatives include Femtocells, Wifi offload, packet-based Microwave, packet Transport, and Pseudo-wire approaches.

In upcoming blog contributions, we will explore aspects of this landscape in more detail and look at the benefits, limitations, and challenges of various approaches. If you have questions or comments please click on the feedback button above.

Before we get started, and with a respectful nod to our legal guidelines, I'd like to remind our readers that the comments or opinions expressed on this Blog are those of the individual contributors only, and do not necessarily represent the views of Sycamore Networks, Inc. or its management. All information is provided on an as-is basis.