But first a little history …

Early Commercial Internet History
In the early 1990s the commercial Internet boom started.  Many folks may not be aware, but while the concept of the Internet itself is quite old (reaching back into the 70s) it wasn’t commercially viable to use the original Internet.  For one, it wasn’t yet very mature, but for another, it was largely an artifact of the U.S. government via a grant to the National Science Foundation (NSF).  The NSF maintained a nationwide backbone called the ‘NSFNET‘ which connected all of the major regional university networks[1]:

• BARRnet: Bay Area Regional Research NETwork (San Francisco Bay Area)
• SURAnet: Southern Universities Research Association NETwork (South Eastern U.S.)
• NEARnet: New England Academic and Research NETwork (New England)
• CERFnet: California Educational and Research Federation NETwork (San Diego)
• NYSERNET: New York State Education and Research NETwork (Greater New York)

Use of the NSFNET backbone required abiding by the NSFNET Acceptable Use Policy (AUP), which said specifically stated: no commercial traffic.  This then was the de facto Internet until the early 90s when small startup Internet Service Providers (ISPs) started building out their own commercial networks. In an effort to make these networks friendlier to business, ISPs began interconnecting directly and trying to work around the NSFnet.  The government encouraged this and began actively shutting down the NSFnet, which was officially decommissioned in 1995.

Internet Hype & Bust
From the early 90s until the late 90s it was predominantly these small ISPs (and a few large ones) who helped hype and build out connectivity to end users and bring the Internet to ‘the masses’.  Large telecom players were largely absent with a few exceptions[2].  In the very late 90s two key things happened:

1. The telcos woke up and used Digital Subscriber Line (DSL) to further drive Internet adoption[3]
2. A wave of consolidation hit the Internet industry with ISPs and datacenter businesses being absorbed by bigger players, largely telcos

This trend further accelerated when the initial Internet hype imploded in 2000, leaving many of the major startup ISP players (some had grown quite large) as smoking craters because they had relied heavily on the first wave of web startups to provide revenue.  Do you remember Exodus Communications?  You should.  It had a peak market cap of 32B, but eventually did a smoking crater impression.

When you look around today you’ll notice that while there are smaller ISPs it’s nowhere near the plethora of players that existed in the early to mid 90s.  The commercial Internet backbone and datacenter business is dominated by cable companies, telcos, and datacenter businesses.

Cloud Futures
I’m sure you noticed the similarities to the trends today with cloud computing.  We’ve barely come out of the initial hype cycle (or perhaps we’re still deep in it) and there are quite a few small players in the space.  More importantly, there are many small players looking to enter.  At some point forcing functions will cause a major culling of the hosting and cloud computing industries, at least at the Infrastructure-as-a-Service (IaaS) layer, which is most susceptible to commodity pricing pressure.

I suspect the forcing functions will be twofold:

1. The telcos and large players will ‘wake up’ and use commercial products like VMware’s vCloud, Citrix Cloud Center (C3), VMops, or EUCALYPTUS to drive wide adoption
2. Serious consolidation will take place

Fights for market share will drive pricing pressure which will drive consolidation.  During the consolidation process large and small businesses alike will see both opportunity and failure.  If it’s anything like the original commercial Internet consolidation there will be a ‘great culling’ of all of the current and new players.  Only the strong or savvy will survive.

How to Survive
If you’re a cloud computing provider, you need to be a Service Cloud, a Commodity Cloud, or a Focused Cloud.  If you’re a large player, you’ll wind up as one of the first two where you can play at scale.  If you are a small player now is the time to look for an area of focus where you can drive value and create sound business fundamentals.  This will position you best for acquisition or, at least, for surviving the culling.

This slipped by me, but looks like GoGrid made the famous Gartner ‘magic quadrant’ for Web Hosting and Hosted Cloud Infrastructure Services (On Demand).[1]  You can see the quadrant here to the right.

It’s great that GoGrid is listed in the visionary quadrant.  When I joined GoGrid in the office of the CTO, one of my explicit goals was to increase the view of GoGrid as a thought leader and visionary in the cloud computing space.  It gratifies me to see the recognition by Gartner.

For those of you who may have missed it, I left GoGrid recently and am currently an independent Cloud Strategist.

At the moment my client list is short, but impressive:

• VMware (#1 Virtualization system; working with their new Cloud business unit)
• GoGrid / ServePath (#2 or #3 IaaS provider in U.S. depending on how you count)
• EngineYard (#1 Ruby on Rails PaaS provider)
• Cloud Central (first IaaS provider in Australia; pre-launch)

Find out how I can help you with your cloud strategy in my one page briefing (PDF).

Unaccounted for in all of this is where the real dollars in today’s market are.  The reality is that they are largely provided by VPS customers.  Possibly as much as 50% of today’s cloud computing market is a single customer buying a single server and leaving it on indefinitely.

One has to wonder what might happen if large ‘hosters’ became aware of this (they are) and how current cloud providers might capitalize on it in the short term (they will).

VPS Hosting

Virtual Private Server (VPS) hosting is a subset of the larger web and dedicated hosting market.  VPS can be thought of as an interim step between dedicated (physical) server hosting and Infrastructure-as-a-Service (IaaS) on-demand hosting.  Essentially you could order a virtual server at a lower price point than a full physical server.  Virtual servers, but not on-demand.  Instead you used the older hosting model of:

1. talk to a sales person or fill out an online order form
2. get a quote
3. sign a contract
4. wait for your virtual server to be provisioned (usually 1-2 days)

I wasn’t able to find detailed numbers on the size of VPS hosting, but I did find some numbers for the overall hosting market.  It was 12.3B in 2007 and predicted by T1R to be 24.4B by 2010.  In the years directly prior to 2008, when cloud computing took off, VPS was gaining very steady traction into the hosting market.  It’s hard to say how much of that 12.3B was VPS, but by all reports it was a fast growing segment.

Add to this that I have heard that large hosting providers like Peer1 and The Planet are seeing a significant impact to their VPS businesses.  Enough to cause them to start moving sooner rather than later on formulating their cloud computing plans.  There are already whispers that both of these folks are building up specialized cloud teams and working on formulating their cloud strategies.

Given that the pricing of many cloud providers is more expensive that VPS from a typical provider, why is this happening?  Simply put, customers are showing the hosting market that they value self-service enough to pay a premium for it.  Self-service means: on-demand, pay-as-you-go, and use-only-what-you-need.

Amazon a Near Term Threat?

Given that it’s common knowledge that most cloud providers are seeing a bulk of their customers from the VPS segment of hosting and given that Amazon is the 800 lb. Behemoth of cloud computing, they must have a plan.  Right?  One can only speculate, but given that VPS plans are typically at a lower price point and providing a smaller footprint server (i.e. less RAM, less disk) you would think they could accelerate capture of this market by providing smaller server sizes at a lower price point.  This is almost certainly in the cards.

Another point that potentially backs up this thinking is the recent release of Amazon’s reserved instance pricing, which allows you to pay for your server a year in advance in a single lump sum and receive a steep discount in return (~70%).  You’ll notice that Amazon announced reserved instances the day after Rackspace announced and launched their Cloud Servers offering.  As you might expect, Rackspace, a savvy and experienced hosting provider, priced their offering very aggressively and catered towards the low end VPS market from day one.

Amazon Instance Size Changes in the Cards?

The smallest current server available in the Amazon Elastic Compute Cloud (EC2) is a ’small’ instance that provides 1.7GB RAM, 160GB disk space, and a single core for $0.10/hr.  Presumably Amazon would either cut that in half (e.g. 758MB RAM) or possibly even smaller (e.g. 512MB or 256MB RAM).  These offerings would compete directly against both VPS and Rackspace’s Cloud Server pricing.

If Amazon follows their current model and simply slice in half, thirds, or quarters their small instances, pricing might look like:

• 768MB: $.05/hr, $.015/hr for a reserved instance
• 512MB: $.03/hr, $.009/hr for a reserved instance
• 256MB: $.015/hr $.0045/hr for a reserved instance

This is $30, $22, and $11 per month at full price or $10, $6.50, and $3.25 per month for reserved instances respectively.  Obviously this would have a tremendous impact on the VPS segment of the hosting market.

Summary

There is an imminent threat to hosting companies of any kind that provide VPS services today.  Larger clouds, including the AWS behemoth and Rackspace, are already launching and preparing to launch price competitive services to capture the VPS market.  The threat is serious in that not only do these clouds provide value based on pricing, but they also provide the traditional values of cloud computing around self-service.  This is a double threat against the VPS market and combined there will likely be a rapid exodus from hosting providers that don’t have their own cloud strategy.

There were a few particular presentations I wanted to point you to however.  There are some very important tectonic shifts at play now highlighted in these presentations.  One in particular, the notion of highly automated infrastructure and being able to describe infrastructure as code, is a particular hobby horse of mine.  My startup, CloudScale Networks, was one of the first to talk about ‘programmable infrastructure’, ‘infrastructure as code’, and ’self-aware infrastructure.’[1]

Check these out when you can!  Most of the presenters and participants are friends of mine and I can vouch that they are very clued on this subject matter.

• Infrastructure as Code
• Configuration Management Panel
• 10+ Deploys Per Day: Dev and Ops Cooperation

In fact, they help in two ways: one by allowing an application to process more data using horizontal scalability (aka ‘elasticity’) and two by reducing load on the primary relational database and hence allowing you to go longer before ‘sharding‘.

Chris Wensel is the man when it comes to understanding Hadoop and he recently gave a couple of talks introducing Hadoop.  Here is one of them:

Happiness in Business

If you can’t be ‘best‘ or ‘cheapest‘, that only leaves being ‘first’ (see Pt. 1: Service Clouds and Pt. 2: Commodity Clouds).  Since Amazon Web Services (AWS) clinched the ‘first’ and ‘best’ titles for the general marketplace, your best bet is to pick a subset of the market to focus on.  Focused clouds find a sweet spot and exploit it.  This is really Business 101 for Startups.  A diagram I saw recently by Ben Caddell brought this into focus and provides a very simple to understand reminder for those of us who may have forgotten (see right).

Let’s look at some of today’s focused clouds.  I’ll mostly talk to Infrastructure-as-a-Service (IaaS), but also touch on Platform-as-a-Service (PaaS) and Software-as-a-Service (SaaS) briefly.

Horizontally-Focused Clouds
By ‘horizontal’, people usually mean a longitudinal slice of the general market focusing on either a stakeholder (e.g. QA, IT, business management) or a business size (e.g. large enterprise, small/medium enterprise (SME), small/medium business (SMB), startups, or individuals).  A horizontal focus, by definition crosses multiple verticals  (e.g. financial services, health, etc. — see below).

We have some interesting examples of these available to us today.  I’ve picked just three to highlight my point: SkyTap, Terremark, and EngineYard.

SkyTap
Perhaps my personal favorite is SkyTap.  SkyTap focuses tightly on providing a unique experience for those in Quality Assurance (QA).  They allow a rich workflow experience that greatly facilitates deploying and saving the state of multi-server applications.  A QA person can find a bug that affects multiple servers in a complex application and literally save the entire system for reuse or re-play by the affected developer at any time.  Combined with easy replication of multi-server environments and other great features designed for this segment only, SkyTap, even though technically an Infrastructure-as-a-Service (IaaS) play is generally under the radar when folks talk about infrastructure clouds.

Terremark
A relatively new entrant into IaaS, Terremark is making it’s mark by focusing on the enterprise.  In fact, their offering is called simply The Enterprise Cloud, showing where they plan to focus.  Terremark uses VMware, which hasn’t had a lot of traction in the public clouds to date.  Presumably this is because they plan to offer some of the more advanced enterprise-class VMware features like HA and DRS.  From my sources at VMware I’ve heard that the Terremark cloud product is quite good and they have developed quite a bit of secret sauce on top of VMware. [1]

Regardless, by picking an area of the market that has been under-served by the heavyweights I think they have a good opportunity.

EngineYard
It’s quite a bit easier, as you move from Infrastructure to Platforms and Software to differentiate and focus on a particular target market.  EngineYard (and their close cousin Heroku who I have mentioned before) focuses on providing a fully managed and automated Ruby-on-Rails (RoR) stack to web startups.  This has already distinguished them amongst the platform crowd and allowed them to ramp up a very respectable business in less than 2 years time.

Vertically-Focused Clouds
If you can’t go horizontal, go vertical.  A vertical focus is an industry focus, be it financial services, health, construction, high-tech, life sciences, energy, or other.  A vertical focus tends to be more solutions-oriented.  When you put together a package that focuses on a single industry it is rarely transferrable, without major changes, to another industry.  However, this kind of focus can be very beneficial for a smaller cloud trying to make a mark early.  This also means it can be rather hard to build a vertical infrastructure cloud.  An example might be someone building a cloud that was highly secure and HIPAA compliant for the medical industry.  Or one that focused on PCI compliance for financial services companies.

Outside of infrastructure, many Software-as-a-Service (SaaS) businesses focus tightly on a given industry.  I don’t know of any current IaaS clouds who are vertically focused and the list of SaaS providers who are vertically focused is too long to list.  A couple of brief examples:

athenahealth
athenahealth provides doctor and patient management services online.

BankServ
BankServ provides online payment processing specifically for financial institutions.

Focus, Focus, Focus

As you can see, if the general market already has dominant players who are ‘first’, ‘best’, and ‘cheapest’, then picking a subset of the market that is not currently served and being ‘first’ there is a great strategy for any new cloud.  In the final part of this series I’ll talk about the particular importance of focus for those players currently in the general market who need to compete on value, not price to survive.  Ultimately, the best way to make money is to help your customers.  Don’t help them on price.  Provide value instead.

[1] If Terremark wants a full review, perhaps they could give me a trial account? <hint>

Enter … commodity clouds.

Infrastructure Commodity Clouds
Recently large providers like AT&T announced their entry into the Infrastructure-as-a-Service (IaaS) cloud computing market and some bigcos with a vision, like British Telecom  (BT), even tried to move early building upon 3tera’s technology to deliver on-demand hosting services.  Even Sun started to get in on the act with their Sun Cloud.  One could argue they led the whole charge with their early on-demand grid service, but by all accounts it wasn’t very successful.[1]

What’s common about these new entrants to the cloud computing space is that generally they are not technology & innovation businesses themselves with the obvious exception of Sun.  In particular, they don’t have track records in innovating in the large scale on-demand infrastructure and datacenter space.  But … they have name brand recognition, buying power, and the ability to price themselves competitively against large incumbents like Amazon and Google.

Without the ability to innovate their clouds the primary tool of differentiation for these entrants is pricing.  While I’m sure they would argue with me about this, the fact remains that we’re unlikely to see the next AppEngine, SimpleDB, or similar from most commodity clouds.[2]

An ecosystem play is just not in the cards.

Platform Commodity Clouds
We haven’t seen a Platform-as-a-Service (PaaS) commodity cloud and it’s unlikely we will in the short term given the relative technological complexity of building a scalable shared multi-tenant infrastructure that can host arbitrary code.  If it were easy, then AppEngine and Salesforce/Force.com would have a whole lot of competitors right now.

Bottom line is that PaaS doesn’t play as a commodity cloud.

Bringing it Back
Again, just a reminder that there are three ways to be a market leader: first, best, and cheapest.  Commodity clouds are focused on being cheapest or, at least, that’s primarily where they can play in terms of value.  This doesn’t necessarily mean that the list pricing will necessarily be cheapest.  Companies like AT&T and BT, at least, will deliver cloud computing as part of their product portfolio which means that businesses that pick them likely will do it because their overall costs or ROI are lower.  They will see cost savings either through discounted pricing when combined with other services or through non-obvious areas like avoiding bandwidth charges.  If you’re in an AT&T datacenter, have AT&T connectivity, then data transfer between ‘the cloud’ and your office will be ‘free’.

What do you want to be?  Best, first, or cheapest?  If service clouds are best, commodity clouds are cheapest, that only leaves being first.  If you can’t be first to the general market (Amazon wrapped that one up) then you need to be first to a smaller focused market.  More on that in Pt. 3.

—-

[1] Sun doesn’t have a reputation for delivering on-demand services as I’m sure even they would admit.

[2] You’ll notice I haven’t mentioned RackSpace.  The jury is still out there.  It seems clear that RackSpace has service cloud ambitions, but it’s unknown whether they can execute.  I’m going to lump them in the commodity cloud bucket for now until more time elapses.  Also their current pricing for the Cloud Servers product seems to reinforce the notion that they are largely a commodity play.

Some folks think that there is no such thing as a ‘private’ cloud.  By definition all clouds are public.  Unfortunately, some times the confusion is around whether a cloud exists internally or externally rather than if it’s public or private.  Yet, even in the case where this confusion doesn’t exist, the notion of a private cloud is sound.  What it means is that there is value in cloud technologies being deployed behind the firewall in a corporate environment.  Why is this?

It’s because cloud computing provides it’s primary value through self-service.

Enterprise IT, Utilization Rates, Hoarding and the Procurement Nightmare
Today’s average enterprise IT shop has arcane interfaces for getting help, usually through some kind of help desk ticketing system.  I once saw the ticketing systems for Microsoft.  There were at least 4 or 5 different systems and it was never clear which one to start with.  This is a very common issue.

Combined with procurement times that some times reach into the 6 month range, any request to get your own IT infrastructure can become a show stopper for your department’s project.  Even virtual machine requests can take days to weeks as there is usually no self-service method and if there was, no usage metrics or charge back mechanisms exist.

Typically this leads to hoarding where departments and managers purchase far more than they need to make sure resources are available when needed.  Is it any wonder that the recent McKinsey cloud computing report predicts that even with virtualization enterprise datacenters can’t exceed 39% utilization?

Internal or Private, It Matters
The CIO must learn how to drive new efficiencies into corporate datacenters while showing how IT can provide direct business value.  The primary strategy, despite McKinsey’s bleak assessment, is to embrace the notion of self-service, preferably in the form of internal cloud computing resources.

This is the future.  This is the way to make IT the corporate star instead of the corporate cost center.

UPDATE:  Had a great response.  Closed the survey.  Here were the results:

1.  Do you prefer articles that are shorter, longer, or a mix?

• 8% - Short is sweet
• 17% - I’ve got time and focus, give me details
• 75% - Mix it up so I don’t know what to expect

2. Do you enjoy reading more technical content? Business-oriented?  Or a mix?

• 33% - I prefer technical content
• 25% - I like it high level and business-focused
• 42% - Give me everything you’ve got

3. Do you consume this blog’s content move via browser or an RSS feed?

• 26% - Browser
• 74% - RSS Feed

Needless to say I found this very interesting.  It could be that a certain kind of respondent self-selected, but there is a decided leaning towards folks who want to read more involved technical articles in their RSS readers.

Thanks to everyone that responded.  It is extremely helpful to have your input.

Service Clouds
I’m borrowing this notion from Alistair Croll (linkedin, twitter, blog), because it’s smart, can be bent to my needs, and makes the most sense in the end game.  By end game, I mean that eventually, most clouds will probably be service clouds.  Why?  Because service clouds offer the best opportunity to expand your market.  Let me explain.

Service clouds shy away from the notion that you must be either Infrastructure-as-a-Service (IaaS), Platform-as-a-Service (PaaS), or Software-as-a-Service (SaaS).  They provide a slice of on-demand services that include all of these as shown in this diagram:

The best examples of service clouds are Amazon, Google, and Microsoft.  All three either provide, or are in the process of providing, a complete ecosystem that includes elements of infrastructure, platform, and software (applications).  Service clouds are also ‘cloud platforms‘ (not to be confused with PaaS).  Platform in the same sense that an operating system is a ‘platform’.

Service clouds are the future.  All cloud computing providers will eventually converge towards this model … for the most part.

Market Evolution
I can’t locate the original blog posting, but I once read that there are essentially three ways to be #1 in a given market:

1. first
2. best
3. cheapest

I think everyone understands this at least on an intuitive level, which means that most cloud players are going to try and stake out a territory as either first, best, or cheapest.  If we assume for now that service clouds are the ‘best’ way to deliver cloud services, then surely it’s important to understand what ‘first’ and ‘cheapest’ might mean.  I’ll take a stab and say these are probably ‘targeted clouds’ and ‘commodity clouds’ respectively.

Targeted clouds are usually smaller cloud players who realize that since the Amazon’s and Google’s of the world are already ‘first’ to the general market they better be first to a specific market segment.

Commodity clouds are larger businesses that may not be able to compete with Amazon or a Google on innovation, but command enough buying power to shrink wrap solutions using best-of-breed software and price everyone else out of the market.  Think ‘utility computing’.

We’ll cover both targeted and commodity clouds in Pt. 2 & Pt. 3, and then wrap up some predictions in the final installment.