Data Center Power Blog

Three Videos that Visualize Data Transactions

Erik Stabile — Mon, 20 May 2013 19:50:00 GMT

It's hard to visualize data on a macro scale. The amount of data transactions per day is really mind blowing. Here are three great videos that attempt to visualize different data transactions similar to those taking place in data centers, labs and super computers all over the world.

1/2 second of trading activity in Johnson & Johnson

Something interesting happens when you trade algorithmically. Things speed up to a staggering pace. Above we visualize high-frequency trading taking place during just half a second on a single stock.

A DDOS Cyber Attack on the VideoLAN downloads infrastructure

Distribute denial-of-service attacks flood servers until they can no longer accept new connections, causing the site hosted on that server to appear broken. Above, you can see what an attack like this looks like.

Hurricane Sandy Animation: Wind Data and Twitter Activity

When Hurricane Sandy struck the East Coast, there was no shortage of social commentary. Here we see how twitter reacted in synch with wind activity.

Data Center High Density Issues, Low Power Servers-No Rats Nest

Julie Brown — Fri, 17 May 2013 23:13:00 GMT

Server Technology responds to data center power customer demand with new 48 outlet product family of power distribution units.

Server Technology has just introduced a new family of 48-outlet switched power distribution units (PDUs) to meet high density needs and power more servers in data center racks, and to help clean up the “rats nest” of clumps of cords, improve airflow and heat dissipation and allow data centers.

The launch of these 48-outlet products is in response to customer demands. Not only did we respond to customer requests for a bank of outlets that are useful to put a lot of servers in the rack, but we utilized our years of power expertise to deliver something more – something innovative that gives customers more than just what they asked for. The new PDUs match each U space in the rack, allowing them to match exactly to the servers. This works perfectly with customers that want to fill their racks and need one PDU to power everything in the rack.

The family includes three 48C13 outlet products for customers that will populate the rack with one type of server – the STV-4301, STV-4302 and STV-4303. There are three families for three different power levels:

- Single-phase 30 amp

- 208, three-phase 30 amp

- 400 volt

“The product that we’re most excited about is the single-phase 30 amp,” said Steve Hammond, Server Technology product manager. “One of the trends we’re seeing in the industry is that more customers are using these 1U, low-power servers. They might draw only 50 to 100 watts. What that means is that these customers can power this entire rack of equipment with only a single-phase 30 amp PDU going into the cabinet. Typically PDU design does not have this many outlets on a single phase 30 amp drop. This product solves the issue and provides customers with the high outlet count they need to fully populate the rack.”

Hammond compares this to putting LED lights in your house that don’t use as much electricity. It’s the same thing with these customers and their servers; they’re saving on energy and improving efficiency, and these 48-outlets help them power and control their servers and low-power server installations.

The products are targeted toward customers like Web 2.0 companies. These companies, like Facebook or Google, usually run a specific application on a set server design in a uniformed fashion. This is in contrast to Server Tech’s alternating phase technology, which can manage different power demands for different types of equipment and/or applications in the data center.

Not only are these products useful for customers, but data center managers can benefit from the PIPS (per inlet power sensing) measurement to make efficiency calculations needed to determine PUE. They also feature locking receptacle design, which is a standard feature in all Server Technology products and allows users to clip locking cords into receptacles and prevents losing power to the server.

With these products, customers have uniform outlet spacing to better match the U-Space in the rack, network and LED displays in the center of the PDU to better accommodate the PDU whether traditional or inverted mount in and shallow depth to better accommodate narrower racks.

Data Centers - Done the American Way.

Robert Faulkner — Wed, 15 May 2013 23:03:00 GMT

I recently read the headline, "Data centers evade consolidation plans", and it triggered a philosophical question in my mind. Often in the data center world, just as with every other aspect of life, there are leaders and followers of ideas. Data center consolidation is an idea that many bright minds have considered carefully and determined to be the best move forward for their organization. Then, there are those that hear that consolidation is good and so, they do it.

Consolidation of data centers is typically meant to reduce costs. And, who wouldn't want to reduce costs? But there is a long-term side effect to consolidation - lost innovation. Certainly, there are occasions in which those in charge of a data center are just being obstinate when rebuffing the consolidation plans of the organization, but it should be considered whether it is just ego or an innovative mind at work. An example might help: A corporation with 3 data centers run by 3 different data center managers are costing the company $3M a year in operating costs. The brightest minds in the company have the numbers crunched and find that they can consolidate to a single data center with cloud DR for $2M per year. "No brainer!" they say. Over the next 3 years a consolidation plan takes place and 25% turn-over in management takes place as well. Now, here we are 3 years later and operation costs are found to be $2.2M per year; only 10% high, everything worked out well. Or, so they say. Now there is one data center manager who was the "top" of the original three. This manager knows the best way to go and isn't afraid to let everyone know that. Three years go on and nothing changes. Operating costs continue to rise to $2.5M per year and that manager now retires. Upon hiring the new manager, new ideas enter the company - others do things differently than they did six years ago when the project for consolidation started. Operating costs begin to decrease over the next three years to $2.0M do to these different practices. What just happened?!

Consider an alternative: Consolidation doesn't happen. The three data center managers are put in competition to reduce costs, with every year the best ideas being enforced between the three data centers. Three years into this competition, the operating cost is now $2.2M per year (same as with consolidation), and the competition continues. Another three years pass and the operating cost is now $2.0M (same as after the new hire). Which is better?

Numerically, they are identical; philosophically, they are much different. Which of the two methods has more control and which of the two methods are likely to result in further reduction (or less increase) in costs over time?

I don't suppose to know the right answer for each and every company or organization, but I do suppose that the average of all organizations would benefit from internal competition just as the aggregate organization benefits by competition with others.

Of course, I could be wrong. what do you think?

Blow Hard! The power of wind brings opportunity to Iowa

Marc Cram — Fri, 10 May 2013 16:11:00 GMT

See the number behind the graph

A recent article on Iowa’s largest energy company, MidAmerican Energy, announced plans to invest $1.9 billion by the end of 2015 in bringing up 656 turbines to generate electric power. With the promise of providing nearly 40 percent of the energy used by customers, MidAmerican thinks they will be able to lower electric rates “eventually.”[1]

Doing some quick math gets you to a cost per turbine of about $2.9M, and a generating capacity of roughly 1600 kW/turbine. The good news is that depending on what MidAmerican charges per kWH, they can recover the acquisition cost of a turbine in as few as 2 years. That is if the wind blows continuously at 8 m/sec, the turbine has no downtime or maintenance (labor costs), there are no taxes on the installation, and MidAmerican can bill out at $0.12/kWH. But rarely does wind power provide good base load (the regions continuous energy demand) generation capacity.

Like solar, wind is instead an “intermediate load plant” used to “fill the gap between base load and peaker plants. Both are intermittent by nature, as output fluctuates with weather patterns. Wind and solar cannot be relied upon to meet constant supply needs, nor can they be immediately called upon to meet peak demands.” [2]

With all of the new datacenter projects in Iowa, there is a tremendous increase in the potential base load energy requirements. Microsoft and Google already operate there, Google planning an expansion, Facebook is on the way. Let us hope that these data centers take advantage of all of the energy saving opportunities at their disposal to minimize their base energy requirements. Free air cooling, hyperscale compute architecture using low power servers, solid state drives, and turning off those servers that are underutilized or not in use are all tools in the datacenter manager’s arsenal to minimize their footprint on the power grid. Iowa will make an outstanding operating laboratory for learning about managing the power demands of a datacenter. Taking advantage of the capabilities of the various intelligent power management technologies available from Server Technology such as Per Outlet Power Sensing, load shedding, and remotely managed outlets can make a big difference to the overall base load of the compute infrastructure.

Contact your Server Technology authorized reseller today for details on how your datacenter can become more efficient and reduce its base load power requirements through power management.

References

See the number behind the graph above

http://en.wikipedia.org/wiki/Base_load_power_plant

http://www.energyvortex.com/energydictionary/baseload__base_load__baseload_demand.html

[1] http://news.ca.msn.com/top-stories/iowas-biggest-economic-project-ever-dollar19b-wind-turbine-plan

[2] http://www.area-alliance.org/documents/base%20load%20power.pdf

Data Center Uptime: Why 0.1% Makes a Difference

Xiaoteng Ma — Mon, 29 Apr 2013 16:52:00 GMT

I finally decided to upgrade to a PlayStation 3 last summer after six years of quality video game time with my PS2. I was amazed at the awesome new features the PS3 offered, but what I liked the most were the cool things you can download from the PlayStation Store, which included games, apps and even movies. However, I have noticed a few instances where I could not get on the network to download things because of server downtime. Although it didn’t bother me very much, but it did bring up a good question: What might be some potential consequences if a company like Google experienced server downtime like this?

The Telecommunications Industry Association (TIA) is an organization accredited by the American National Standards Institute (ANSI) to develop industry standards for different information/communication technology products, which includes data centers. In 2005, the ANSI/TIA-942 was published, which created four different tiers for data centers. There are various differences between different tiered data centers, but for this blog post, we’ll just go briefly go over the availability, which is reflected in the table below. As you can see, the percentage of uptime between a Tier 1 and a Tier 4 data center appear to have a very minimal difference of 0.324%. However, when you take into consideration the number of minutes in a year, a Tier 1 data center will have 28.44 more hours of downtime than a Tier 4.

Type	% Availability	Downtime (min)	Downtime (hrs)
Tier 1	99.671%	1,730.37	28.84
Tier 2	99.741%	1,362.21	22.70
Tier 3	99.982%	94.67	1.58
Tier 4	99.995%	26.30	0.44

When selecting a tier to fit your business needs, some important questions to ask might be “Can my organization afford this many hours of downtime?” and “What are the potential consequences for this amount of downtime?” For example, large companies in the banking and healthcare industries will not likely opt for a Tier 1 data center due to the potential implications of prolonged downtime. It is imperative for data center administrators to be able to ask the right questions and select the most cost-effective option that fits organizational/industry business needs.

What are your thoughts? Please feel free to share in the comments.

Facebook's New $1 Billion Data Center

Erik Stabile — Mon, 22 Apr 2013 17:32:00 GMT

When I think of Iowa, I think of three things, John Deere, Hawkeyes and soy beans. But now it might be time to add one more item to that list—data centers. According to USA Today, Facebook has announced a plan for $1 billion data center in Altoona Iowa—the most advanced data center in the world. The data center, which should bring in an undisclosed amount of jobs to the region, is going to be 1.4 million square feet. Undoubtedly it’s going to open the door for new jobs and economic growth. During a time when states are furiously competing for opportunities for economic development, why did Facebook choose Iowa?

For starters, Iowa is one of the most affordable states to open a new data center, according to a ranking published by the The Boyd Company in 2008. The ranking compared cost of land, labor, property taxes and power. But Facebook seemed to have an added interest in Altoona’s modified water ordinances, location to a large energy substation, available fiber optics, and local to major interstate highways.

It’s uncertain what kind of tax incentives, if any, Iowa will provide. But Facebook hopes to receive tax breaks for wind energy production, which would require a legislative action, according to the Des Moines Registrar.

Facebook is known for touting their energy efficiency. Learn about some of their energy saving techniques in this video:

Image Flickr CC 2.0: IntelFreePress

Data Center Energy Efficiency from Server Tech, RF Code and Nlyte

Julie Brown — Fri, 19 Apr 2013 16:42:00 GMT

“Wildly successful,” “Very informative,” “Valuable, easy to understand content,” and “Interactive” were some of the attendee comments to date from the "Data Center Energy & Operational Efficiency 2013" event series, presented by Server Technology, RF Code and Nlyte. Hot topics discussed and debated by attendees and presenters included capacity planning, energy planning and management, and cost savings. Four events still remain in the series:

. April 25, 8 am-noon, Reston, VA: CoreSite Data Center

. May 2, 8 am-noon, Chicago, IL, Steadfast Data Center

. May 23, 8 am-noon, New York City, NY, The Player's Club

. June 6, 8 am-noon, Raleigh/Durham, NC, Location TBA

The series addresses data center challenges from both an operational and financial perspective, including capacity planning, full utilization of facilities, increased power costs, lower power ability, environmental monitoring, proactive power management and predicting future power needs.

“In the past year many Fortune-100 companies have suffered huge financial impacts over service disruptions resulting from the lack of robust visibility and capacity management in their data centers. Other organizations have suffered asset loss and security breaches and the resulting fines ranging into the millions of dollars. We read about these highly visible failures all the time, and they all stem from poor visibility and management of the data center as a single system,” the companies explain on the event site.

On their agenda for 2013, data center professionals are thinking about transactional costs for the delivery of services, existing exposures, cost reduction, audit control, financial and regulatory compliance, infrastructure security and power consumption. The seminar series will help these professionals manage the data center to meet their different business needs.

Registration is still open. The Reston event, happening Thursday, April 25, is taking place at CoreSite, an expanding data center (to 200,000 sq ft) and a great example of the data center boom in the U.S. today. The Chicago event will be at Steadfast Data Center, known as the World's Largest Data Center. Chicago is a demonstration of the conversion of one of the great old 'loop' buildings into one of the most energy innovative Data Centers in the U.S.

The Last Mile and “the energy monster of cloud computing”

Marc Cram — Fri, 12 Apr 2013 15:50:00 GMT

A recent article by Spandas Lui of ZDNet titled “Wireless is the ‘energy monster’ of cloud computing” quotes Dr Kerry Hinton of the Centre for Energy Efficient Telecommunications (CEET) as saying that “By 2015, the energy consumption of datacenters will be a drop in the ocean compared to wireless networks in delivering cloud services. The problem is, we’re all accessing cloud services – things like webmail, social networking, and virtual applications – over wireless networks.”

The CEET report is cited as saying “that wireless access network technologies including wi-fi and 4G LTE will account for 90 percent of total wireless cloud energy consumption by 2015, while datacenters will make up 9 percent.”

“The CEET said that the global telecommunications system will consume about 2 percent of the world's energy, and that figure could jump to 10 percent by 2020 if the IT industry remains idle.”

"We often think of bandwidth as the barrier to the way online services evolve and improve," Dr Hinton said. "The very real message here is that the real bottleneck looming sooner than we think may be energy."

One way to fight this problem is to forgo some of the mobile luxuries we are accustomed to when accessing cloud services, but Dr Hinton knows that this is unlikely to happen.

"Another solution might be to increase the way network resources are shared among users, but more likely, we'll need a radical improvement in the efficiency of the technologies themselves, and potentially, a fundamental change to the way data is managed across the global network," he said.

http://goo.gl/bGHlC

As the world’s population continues to take up smartphones and tablets for their internet access tool of choice, how do we collectively reduce the power consumption of all the various wireless access technologies?

The simplest way is to turn off assets that are not in use. Utilizing remotely switched power distribution can facilitate this.
Deploy higher bandwidth radios that can share more connections with the same amount of power.
Re-aim the cell tower antennas during off peak hours to cover broader areas, and then turn off those towers that are not needed. Bring the turned off towers back on line as network loads increase.
Measure the electrical load on the home gateway and turn it off when not in use, or at least enable a low power state that can quickly restore bandwidth when needed.
Develop more sensitive receivers with better signal to noise ratios that do not require as strong of a signal in order to communicate.
Continue to work on new means of data compression
Rely more on photonics for transmission of data; switching and signal processing can remain electronic for now

For more information, take a look at the following link: http://transmoder.com/2012/01/26/is-the-internet-green-interview-with-dr-kerry-hinton/

Creative Commons image: Flickr User plastAnka

Google Glass while driving - Not in WV - Responsibility vs Liberty

Robert Faulkner — Thu, 11 Apr 2013 22:16:00 GMT

There is pending legislation in the state of West Virginia to ban "operating a motor vehicle using a wearable computer with a head-mounted display." This is obviously directed at Google Glass. The intent is to reduce distracted driving, but what is the evidence that this will actually do so?

Matt Peckham over at Time Tech has a nice article on this very topic, but I want to touch on something more fundamental. Is technology making us more or less free? And, where does personal liberty end and public responsibility begin?

Ok, I'm not going to be able to really answer these questions here, but it should be food for thought for any technophile or politically minded individual. The idea of banning distracted driving is really the topic at hand. Few disagree with the idea that distracted driving is bad, but why is one type of distracted driving necessary of legislation but others are not? Sure, it is more dangerous to text while driving than to not text while driving. It is also more dangerous to have children in the car than to not have children in the car. Certainly, the former sounds reasonable to ban, but the latter would be ridiculous. What about phone calls while driving? The West Virginia bill requires the driver to use a hands-free device for the conversation but not for the dialing. It seems to me the dialing is more distracting than simply holding the phone to my ear. Certainly it isn't driving with one hand that is the problem. I seem to find myself at a loss more often than not when trying to understand these differentiations. Perhaps that should be expected. Legislation is more often than not based on subjective opinion above demonstrable fact.

That brings us to Google Glass and the ban pending in WV. You can't have data on the safety of driving with it before the technology is available. How does the legislature know it won't actually reduce driving accidents and fatalities? If it or other technology reduces the need to touch anything but the steering wheel, isn't it possibly better for safety? I guess the theory is ban first, ask questions later. I'll leave the more cynical possibilities for another day.

Now, back to the first question: Is technology making us more or less free? Many articles and books have provided a wealth of information and opinion on this topic, but I'll throw in my two cents. Technology itself makes us more free mostly through easier access to knowledge, though it takes an ever more discerning mind to distinguish truth from fiction in this age. The thing that threatens that freedom is the fear of technology and the limitations of our human mind in handling change. "Wearable computers with head-mounted displays" are a simple example of this. The technology has the potential to provide huge amounts of data to the wearer - road markers, night vision, weather conditions, traffic information, feeds from cameras, etc. We just don't know how much is reasonable for the individual to take in without disrupting the task at hand - driving safely to a destination. I often think about the fact that we readily drive 65+ miles per hour in heavy traffic paying attention to road signs, debris, GPS, the radio, kids in the back seat, and other drivers and relatively rarely get into an accident. Would the drivers of 100 years ago have been scared out of their wits to be plunked down onto these modern roads? I think so. Can we add one more bit of information or 100 more? No one knows, including the WV legislature.

As usual, I leave you with more questions than answers. What do you think?...

A few words about replacing fuses and resetting circuit breakers

Bruce Auclair — Mon, 08 Apr 2013 20:59:00 GMT

As a Systems Engineer who spends a fair amount of time responding to inquiries received through our Technical Support department I often hear "The PDU blew a fuse. I replaced it but it blew again. Something is wrong with the PDU. Please send me a new one." Would that we could, but the PDU is doing what it should. Following are some notes about why fuses blow and circuit breakers trip, and how to respond to re-energize the outlet branch safely and effectively.

The first thing to note is that the cabinet-level PDU doesn't blow its own fuses or trip it own circuit breakers. The PDU itself does not create a load on the protected branches of outlets. Rather, it is the devices connected to the PDU's outlets that create the load and are responsible for conditions that can cause an over-current protection device (OCPD) in the PDU to react. It's a simple premise, but understanding this is crucial to reacting constructively.

Fuses blow and circuit breakers trip for one reason: Current through the OCPD has exceeded its maximum current-carrying capacity. Two types of faults exist: over-current and short-circuit. Over-current faults are the aggregate load of the attached equipment exceeding the ampacity of the OCPD and are avoidable through proper loading of the branch. Short-circuit faults are the result of a faulty component, usually a power supply gone bad, and are difficult to avoid. In either case, the OCPD is doing its duty when it opens to de-energize the branch and prevent an otherwise dangerous condition.

Steps for re-energizing the outlet branch after an OCPD interruption:

1) Disconnect all power cords from the outlets that were de-energized by the OCPD. This is often overlooked, but the issue is this: If the fault still exists the OCPD will react again when the branch is re-energized.

2) If the PDU uses fuses for the branch protection, de-energize the entire PDU before proceeding. The only exception to this is for Server Technology's -48VDC products which are equipped with hot-swappable fuses that are rated as a disconnect switch. If the PDU is equipped with circuit breakers, there is no need de-energize the entire PDU.

3) If the PDU is equipped with fuses, note that the outlet branch may be protected by a pair of fuses. If two fuses are present per branch, always replace both fuses even if a continuity test with a common multi-meter indicates that one of them is not fully open. Replace both fuses, and always replace using new fuses that are the identical make and model as before. Re-energize the PDU after replacing the fuse(s).

If the PDU is equipped with circuit breakers, locate the circuit breaker that had opened. See the photo above for reference. To reset, press the rocker button until it latches and the face of the circuit breaker is flat.

4) If the PDU is a non-switched type, the outlets will be energized immediately upon re-energizing the PDU after replacing fuses, or after resetting the circuit breaker. If the PDU is a Switched type, wait an adequate amount of time to allow the PDU to intialize the ports and sequence the outlets to an On state. Continue only after verifying that all outlets are live.

5) Reconnect the power cords one-at-a-time. Wait a few minutes between each to allow time for the device to boot and the load settle before moving on to the next cord. If a fault with a single device still exists and the OCPD interrupts the fault again, you will need to return to step #1 above but you will have identified the culprit. If all devices are successfully attached but the OCPD reacts and de-energizes the branch a short time later, the branch is most likely over-subscribed and some loads need to be either removed or re-distributed to another branch.

What you need to know: High Voltage Applications in the Data Center

Calvin Nicholson — Mon, 08 Apr 2013 20:14:00 GMT

By Calvin Nicholson, Server Technology, Inc.

High Voltage Applications / Power, Efficiency and Safety in the Green Data Center

No matter how you look at it servers run more efficiently at higher voltages. You can look at studies by The Green Grid, efficiency information from the servers themselves or other studies like the one below from 7 x 24; running servers at higher voltages makes sense.

Marcoux and Sumrell, 7x24 Exchange, “277V Power Supplies”

You can argue all day about what the real efficiency values are but shown below are some numbers from one of the many studies on this topic. The key goal is to increase efficiency and reduce operating costs that ripple through the organization in additional savings.

While there has been an explosion of higher voltages being used in the United State for greater efficiency, it’s 230 or 240V solutions that are saving the day with power distribution solutions. These solutiosn are off the shelf and available today.

For most organizations 277V power at the server is not a viable solution. Cost, custom power supplies, legacy gear, 277V power availability and a number of other reasons make this a less than viable solution for most users.

So How do you get 277 V from 480 V?

3 Phase 480 power is distributed as single phase power at 277V as 480 / 1.732 (the square root of 3) = 277V. I won’t bother to explain why this math work—it just does (at least for most of world using Wye power).

In the real world most organizations are still dealing with legacy single cord devices and devices that still operate at 120V. So how can they even consider solutions like 277V power distribution.

A case for 400V solutions within the data center:

Higher voltages have brought more attention to the interrupting ratings required as these values have increased. All 30A or greater Cabinet PDUs (CDU) are required to have over current protection devices (OCPDs) per agency requirements. No matter where it is in the system all overcurrent protection devices must have interrupting ratings equal to or greater than the available fault currents at their line side terminals per NEC 110.9 and OSHA 1910.303 standards. OCPD devices using fuses typically have interrupting ratings in the 100,000 A or more for 600V AC or less. With 415 VAC or greater voltages it is not uncommon to have 50 kA or greater short circuit current available at a RPP or server cabinet busway plug-in. Even the CDU can have high fault currents. A careful review of what type of OCPD and the ratings is critical for high fault current installations.

Note: Interrupting rating is the maximum short-circuit current that an overcurrent protective device can safely interrupt under standard test conditions. IR is an abbreviation for the term interrupting rating. Interrupting capacity with an abbreviation of IC is an older synonymous term carried over from years past. The National Electrical Code, UL Standards, and markings on fuses and circuit breakers now use the term interrupting rating and markings such as "IR 200KA" or "200kA IR." The term AIC or KAIC, such as in "200k AIC," is no longer used for product markings nor in the NEC or UL Standards.

Data Center Power: Best Practices for Cost Savings

Bob Parente — Tue, 19 Mar 2013 18:25:00 GMT

The cost impact on where and how your IT Equipment is deployed requires analysis of power availability, infrastructure, and rack power density. However, with a close analysis of the end rack / IT equipment groupings some alternative rack based power distribution options could make sense and provide significant cost savings. For example, if you are utilizing a co-lo provider and are being charged by the quantity of power feeds and circuit protection devices from the RPP (Remote Power Panel), a significant savings may be realized by simply moving to an in-rack based power distribution scheme.

Using an example of quantity three to six racks with a fully redundant total power requirement of approximately 17kW to 21kW maximum, benefits may be realized with the delivery of two higher power 3-phase feeds to the group of racks and perform the power distribution within the racks themselves. This reduces the number of RPP circuit protection devices required and the number power feeds to the racks.

Once the primary ‘A’ side power feed and secondary ‘B’ side power feed are brought to the group of racks, a power distribution unit within the rack can be used for further distribution such as shown in the above diagram.

When evaluating power density and redundancy needs for your IT equipment, a thorough analysis of the various physical equipment distributions within groupings of racks may present alternative power distribution schemes that provide cost savings.

Click...Click...Boom! - Unreasonable Expectations for Mobile Apps?

Robert Faulkner — Thu, 14 Mar 2013 00:00:00 GMT

Over at TechCrunch, Sarah Perez has an interesting article on the low tolerance for buggy mobile apps. According to research, only 16% of users will try a failing app more than twice.

This certainly has big implications for app developers which might be frustrated at the possibility that the first impression is the only one they will be able to give. If you are like me and the other typical user of mobile apps, you would probably agree that the time and money invested in deciding on the app is what drives the behavior for discarding that app. Take a mobile game for example: I click to open the store, quickly browse the top free games, and make a selection based on the name and picture. This means I have invested maybe five minutes and $0. Thus, if it has even the slightest annoying problem, it's gone.

In comparison, to buy the mobile phone or tablet in the first place, I probably would spend at least 8 hours on reviews and comparison shopping before plunking down several hundred dollars. I can't imagine anyone simply discarding their device if it had a problem on first use! So, it becomes obvious that it isn't an unreasonable expectation that causes the mobile app to be discarded, but the fact that it has little to no value built into its acquisition that does so.

Competition is also a huge factor. Apps, like that for Facebook, have very little competition. Users have to put up with any problems because of their goal of keeping in touch. This doesn't apply to killing time with a game because the goal of killing time is met even by the process of getting and discarding poorly executed apps.

I'm sure there are marketing departments around the world trying to understand and manipulate the psychological aspects of this phenomenon, and I would be very interested in the results. But, it should be considered that even though consumers don't act this way with hardware, they do act this way with other types of consumables like food, hotels, and places of amusement. The return customer is key in most aspects of the technical and non-technical side of consumerism, but I'm not sure that has been much of a factor yet in the mobile app world. That makes me think that the third prong of this phenomenon, along with low investment and high competition, is that there is very little sense of loyalty to most app "brands".

I suppose that as major app publishers consolidate to reduce competition and bring incentive for loyalty, consumer forgiveness will rise and then the mobile app industry will finally be mature.

What do you think?...

Moore’s Law and Murphy’s Law in Full Effect in your Data Center

Calvin Nicholson — Mon, 25 Feb 2013 21:34:00 GMT

by Calvin Nicholson Senior Director and FW and SW Development

Moore's law is a rule of thumb in the history of computing hardware whereby the number of transistors that can be placed inexpensively on an integrated circuit doubles approximately every two years.

Murphy's law is an adage or epigram that is typically stated as: "Anything that can go wrong will go wrong".

The demand for higher and higher computing capacity continues to drive up cabinet densities and therefore; the power required to operate these systems has greatly increased as well. Of course data center floor space is also some of the most expensive property per square feet to build that there is. These and other factors have led to three phase power commonly being brought down to the cabinet level with voltage requirements in the 400 V / 480 V range and with current densities at 60 or even 100 A to meet these greater compute demands. There have been articles written that estimate that Moore’s law can continue for the next 600 years http://arxiv.org/abs/astro-ph/0404510 and it will be interesting to see just how long it continues to apply. Of course higher voltages also have the additional benefit of allowing servers to operate more efficiently which reduces costs. These costs are becoming more critical as power costs continue to increase and power availability in many areas becomes scarce.

Data centers assume Murphy’s law is as true as Moore’s law and plan for redundancy. Hundreds of hours are spent designing critical facilities around LEED standards and determining the appropriate Tier level required for data center facilities. Redundant systems, including power reduces downtime and how facilities support often unplanned increases in demand. Unfortunately as data centers increase density and run at higher temperatures, to reduce operation costs, the room for error also greatly decreases. Often a loss in cooling now only allows for a short 60-90 second window before temperatures can rise to dangerous levels for many devices. Thus power and environmental monitoring with device control will continue to be the “new” law as data centers move forward. As The Green Grid coined the term long ago “You cannot improve what you are not measuring”. Whether you're looking to improve your PUE or ensure that failed systems do not result in down time or damaged equipment.

One simple way to ensure equipment is not damaged when temperatures increase, a UPS goes on battery or a current load exceeds a threshold is to implement a feature like Smart Load Shedding. Smart Load Shedding allows for non-critical devices, by priority, to be automatically shut off during these conditions ensuring system and device integrity. I can hear most of you saying “automatically shut off WHAT?” not a chance in hell am I going to allow any device to automatically shut off anything in my data center and I don’t blame you. For the right situations such as remote locations, secure locations that are difficult to access and other installations this is a solution that makes perfect sense for your environment.

View the load shedding diagarm

The Benefits of 240/415V Power Distribution in North America

Bob Parente — Fri, 22 Feb 2013 17:14:00 GMT

North American data centers continue to implement higher voltage power distribution to the end equipment AC power supplies at an increasing rate. Bringing higher AC voltages to the data center rack equipment power supply modules will maximize device energy efficiency.

AC power supplied equipment will operate more efficiently using the highest supported voltage available.

Provisioning for increased power demand using minimal power feeds into the rack requires increased voltage, current, or both. Using an example of a rack with a power requirement for up to 17kW of continuous current redundant power, we can determine how many power feeds to the rack (and upstream Remote Power Panel circuit protection points) would be necessary based on the power delivery system parameters (voltage, amperage, and phase). Referencing the table below, we can see the higher the voltage and / or amperage, the greater the power availability using less conductors (power feeds into the rack).

Number of Power Feed Required Comparison to provide safety rated 17 kW

The most common method of rack power delivery in the data center is to provide independent appropriately rated A & B power sources. The A & B input feeds load share with a maximum of 40% capacity on A & 40% capacity on B. This maintains an 80% maximum load (NEC Safety rated requirement) should any one side be required to carry the entire load.

Following outlines the two power distribution methods that would provide up to 17kW of continuous current redundant power with full power redundancy:

208V, 60A 3-Phase [208V x 60A x 1.732 x .8 = 17.3 kW, per power feed (Volts x Amps x square root of 3 due to line-to-line load with 2 of the 3 phases x 80% maximum safety rating per NEC].

NOTES: two3-phase power feeds (A & B) to the rack with full redundancy. This would provide a means to maintain an 80% safety rated maximum load of up to 17.3 kW should any one side (A or B) be required to carry the entire load.

240V 30A 3-Phase [240V x 30A x 3 x .8 = 17.3 kW per power feed: Volts x Amps x 3 due to line-to-neutral load with 2 of the 3 phases x 80% maximum safety rating per NEC].

NOTES: two 3-phase power feeds (A & B) to the rack with full redundancy. Increasing the power by means of increased voltage without increasing the amperage (240/415V @30A) allows use of rack based power strips utilizing smaller more pliable power inlet cords with reduced size connectors as compared to that of a higher amperage rated (208V @ 60A 3-Phase) power strip.

The majority of the implementations are for new data center builds for 240/415V power distribution. Some key points for 240/415V power distribution to the rack are:

Provides a means to increased rack Power densities.
Equipment will operate more efficiently using the highest supported voltage available.
Fewer branch circuits to the rack will reduce wiring, weight, & bulk.
Less wiring bulk increases airflow and cooling efficiency.

CramIT! A Datacenter Blog - Do large numbers scare you?

Marc Cram — Wed, 20 Feb 2013 23:16:00 GMT

The task of managing a large number of machines that are all expected to continuously run in perfect harmony can be a significant challenge if you let it. The complexity of various hardware configurations that are expected to support multiple hypervisor and virtual OS instantiations along with the countless thousands of potential client hardware, OS, and software stacks can be daunting. The combinations and permutations quickly run into the billions. Supporting all of this digital complexity is an analog technology that predates them all – power. The last thing you want to have fail when everyone is counting on you is power. Yet even power is succumbing to the digital tidal wave. Today your average intelligent power strip is capable of generating hundreds of data points per minute. When you multiply this by the hundreds or thousands of strips in a data center, or even across multiple data centers worldwide, managing the torrent of power data can be overwhelming.

The Sentry Power Manager (SPM) tool from Server Technology is one way of coping with the power data coming from your datacenters. With the built in reporting and alerting functions, there is no need to drown in a sea of power numbers. And when it comes time to report your PUE, having the compute load numbers at your finger tips can be a pleasant experience. Talk to your local Server Technology reseller today for a demonstration of the power of SPM.

The End of the Universe! - maybe? And a lot of data to boot.

Robert Faulkner — Wed, 20 Feb 2013 00:27:00 GMT

Some of you, my fellow nerds, were very excited when the Higgs Boson was evinced in July of last year. This past week, the LHC at CERN was shut-down for a 2-year maintenance and Mark Hoffman over at Science World Report has posted a nice article on the 75 petabytes of data recorded in its short three years of operation. That's an average of 25,000 1-terabyte drives per year - not bad!

That's interesting on its own, but now Clara Moscowitz over at Space.com has reported that the 126GeV Higgs boson may portend the end of the universe. Basically, the electro-weak vaccum may be metastable depending upon Higgs and Top quark masses. If that is the case, a true vacuum bubble could form and ripple through the universe, destroying it at the speed of light.

Oh well, it was good while it lasted...

Where will all these pixels go?

Erik Stabile — Wed, 13 Feb 2013 17:05:00 GMT

I love photography. I love cameras. I love the fact that an 8-megapixel camera is seamlessly integrated into my cell phone. The convenience of a camera phone, which takes decent photos, is something I would have never imagined possible five years ago. And the fact that Nokia has a “41-megapixel” camera is completely mind blowing. Of course with this camera you won’t get anything near the quality of a camera like Hasselblad H4D-40 ($15,999.00), a camera that can truly boast 40 seamless pixels of pure image perfection. That is because pixels count does not translate into image quality. There are many other factors at play, like the quality of the lens in front of the images sensor, the sensor itself, the ISO sensitivity, and most importantly—the shooter. But to a casual consumer, pixel count may determine everything. So it is likely that cell phone companies will continue to engineer cameras with high and high pixel rates.

So what does this have to do with data centers?

More than 43 percent of people used their cell phone as their primary camera.
There are roughly 300 million photos uploaded to Facebook every day.
There is nearly two days worth of video uploaded to You Tube each minute.
Instagram is nearing 100 million active users and 4 billion photos.
Flickr, which hosts hight-quality images, has around 75 million users.

Now imagine 50 or 60 percent of the population snapping photos with 15, 20 or even 40 megapixel cameras. The improvement of cell phone cameras will only help feed these social behemoths larger diets of pixel-rich imagery. It is a staggering amount of data and it only has one place to go.

So how will data centers accommodate all this new data?

Capasity planning. This is where we come in.

New PDU Power Pivot for Rackmount Data Centers

Julie Brown — Wed, 30 Jan 2013 17:05:00 GMT

You've never seen anything like this before. In fact, don't read this blog -just watch the attached video at www.servertech.com/pdupowerpivot. You'll see why we say that this is an innovation that will revolutionize the data center power distribution (PDU) market. The PDU Power Pivot is an incredibly flexible design with a 90 degree rotatable power cord. This makes it the ultimate PDU for multiple racks and multiple data center facilities.

“This is a huge leap forward in data center PDU innovation,” said Brandon Ewing, president of Server Technology. “The PDU Power Pivot feature is a game changer for the data center industry. The flexibility that this innovation brings to the PDU market is remarkable – for installation and mounting options and for multiple racks and multiple facilities. We see data centers standardizing on the PDU Power Pivot feature going forward simply because of the flexibility and cost savings. This is a PDU that can adapt as a company grows or changes its data center,” said Ewing.

The PDU Power Pivot is the latest in an almost 30-year history of power distribution innovation for Server Technology, a company well known as the inventor of the intelligent PDU. The PDU Power Pivot is unique in that instead of having a static power cord; it has a 90-degree rotatable cord that can be positioned in a variety of ways. “Whether your data center has a raised floor or an overhead busway or any other way to bring power to the rack, the PDU Power Pivot can rotate for optimal configuration. Plus it can install easily to support racks from multiple vendors,” said Steve Hammond, product manager at Server Technology.

Secondary innovation in the PDU Power Pivot

The PDU Power Pivot also has an LED screen with an “auto-flip” display that actually inverts depending on the position of the PDU. “The auto-flip display works just like your Smartphone so the customer doesn’t have to tilt their head to read what’s on the PDU,” said Hammond. “When you combine the auto-flip display with the PDU Power Pivot, you really have a data center PDU that you can use anywhere,” added Hammond.

The PDU Power Pivot will be offered initially in four new products: three new 30-outlet PDUs in the basic, metered and smart PDU categories and one 24-outlet switched PDU, offering on, off and reboot capabilities.

For more information on the PDU Power Pivot and to see it in action, visit www.servertech.com/pdupowerpivot.

Data Center Systems Integration Puzzle Solved

Calvin Nicholson — Mon, 28 Jan 2013 19:12:00 GMT

It is clear that the rising cost of power, power availability, the great recession, a more competitive business environment, being “green”, political pressure and a number of other factors are leading organizations to focus on costs and being more efficient. This is particularly important for facilities that are high visibility, high cost and high users of power and water like the data center. In this blog we will examine the different systems that you might find within a data center like a BMS, DCIM and other tools that can help control costs and make your facilities more efficient. Hopefully we will also clear up some of your questions on how and where these systems may fit into your ecosystem.

BMS (Building Management System):

A BMS system is typically used to monitor and control systems like ventilation, lighting, power, fire systems, security systems and other factors within the data center. They are typically defined as monitoring the “back of the house” or are more facilities focused than IT focused within most data centers or organizations. Manufacturers include Honeywell, Siemens, Johnson Controls and a number of others.

We believed 7 or 8 years ago (before the concept of DCIM was widely accepted) that it would be the BMS system that would take over and monitor at least the bulk of the operations within a data center. This just made sense with a huge effort required in initial cost, installation, training and support to get these systems operational.

We have done both SNMP and API integrations (using our Cabinet PDUs & Sentry Power Manager software tool) with these systems but overall this has been more of the exception than the rule. Part of this might be due to foreign ideas and concepts that relate BMS systems much more to the industrial market place like Pulp and Paper mills, power generation, chemical plants and other industrial environments. These industries all use BMS systems and are more familiar with devices like the PLC (programmable logic controller) and communication protocols like C-bus, Profibus, and internet/open standards like DeviceNet, BACNet, LonWorks and Modbus all common protocols in “the back of the house”. In the end the total acceptance or lack thereof of the BMS for broad monitoring capabilities within the data center may have more to do with the divide between facilities and IT than anything else.

Data Center Management Software Systems:

I have given these systems their own category as these manufactures make multiple systems and some of them can be classified as a BMS or DCIM solution or both depending on what they are trying to sell you. Many of these different tools within the same suppliers offering will integrate together depending on the functionality you require.

Suppliers include APC, Emerson Liebert, Eaton, HP and IBM. Typically the one thing that is clear with these systems is that they are tied very closely with these manufactures products and the concept of an open architecture and implementing best of breed solutions are often lost on the “one stop shop” concept. As data centers become more efficient and designs are more cookie cutter, large organizations will look more to the one-stop-shop concept to reduce costs, speed implementation and deploy in multiple global locations. The question here is: long term, will companies still build their own data centers or will this be outsourced to the cloud or “experts” in the industry building new hybrid co-location facilities?

To be fair, in some cases these suppliers will integrate other products into their systems if they don’t manufacture that product, if the customer insists that the work be done or if the customer is willing to pay significant fees to get this work completed.

Data Center Infrastructure Management (DCIM Systems):

Right now DCIM is the rage with everybody discussing these systems and how they help to manage and monitor your data center. Each tool has its own advantages and disadvantages but many of these systems are capable and willing to monitor just about any product within the data center whether it is in the data center itself or in the back of the house.

Manufactures include FieldView, Nlyte, Rackwise, Cormant (Cable Solve) and a number of others including software systems provided by many of the hardware manufactures that monitor both their own devices, devices from their competitors and in many cases other devices as well.

The DCIM promise to cut costs, reduce downtime and automate network planning, implementation and operational tasks typically hinges on four key capabilities;

1) Infrastructure Management (means different things to different systems/users)

2) Environmental monitoring/optimization

3) Power consumption monitoring and control

4) Intelligent physical layer infrastructure management

Looking at the challenges and all of the things that a data center manager needs to do, it is hard to see how one system can do everything within a data center. The main requirements for a typical data center include but are not limited to multiple areas:

A) IT Asset Management, Change Management, ITIL CMDB (Configuration Management Data Base)

B) DC Capacity Management, Change Management, DCCMDB

C) Network and Systems Management, Power Monitoring Management

D) Power and Environmental Monitoring and Alerting

It is clear from these challenges that it will be some time before one system is capable of doing all these things within the data center - if ever. It is also clear that in the short term and possibly in the long term that these multiple systems within each data center facility are required and will need to communicate and share information to achieve the results customers require to efficiently run these facilities. Integration is the key to help bring together key information from multiple systems as required to maintain uptime, increase efficiency and provide the overall ability to manage and monitor today’s Data Center. Many systems today take advantage of the expertise and information available from other already existing systems within that Data Center using API (Application Programming Interface) to provide that “single pane of glass” view within a single system that customers require. These APIs are typically xml-based using industry standard tools like SOAP and REST. They allow the existing systems to provide greater value at the device level specific to those products or devices that they support.

Security - the cool new thing looks awfully familiar?

Robert Faulkner — Mon, 21 Jan 2013 23:27:00 GMT

The arms race that is security vs hacking may never end. Google is now talking about a hardware "key" to allow you to access your accounts. This is to provide 2-factor authentication. Robert McMillan over at Wired has a nice explanation of the proposal. I really like the 2-factor authentication Google has already implemented in which, whenever I try to login to my account from an unfamiliar location, I have to enter a code that is generated on my phone.

Now they are proposing that I carry around another piece of hardware that plugs into the USB drive of my computer. This made me have instant flashbacks of the old software protection dongle. I am already carrying my computer and my phone, please don't ask me to put something on my finger, key ring, tie clip, or shoelaces. Hardware is almost guaranteed to get lost or fail sometime. And, of course, hardware authentication is really challenged by the vast variety of platforms (e.g. laptops, tablets, televisions, etc.). I believe we are a long way from a standard working its way through all the means to access the internet.

It brings to mind the promise of biometrics for authentication. There you can potentially have truly unique identification with effectively unhackable security. But wait... Even that is hackable with false fingers, eyes, or whole faces.

When it comes down to it, I believe it is most likely that there is no such thing as truly unhackable security; however, there is such a thing as "good enough for me". What I mean by that is: Nobody would bother targeting me for my Google account, so the 2-factor software authentication with strong password should be more than enough. My bank account doesn't have much in it <sigh>, so I don't think I need an iris scanner to get in. For me, carrying around a tertiary piece of hardware just for accessing any account is more bothersome than it is comforting.

What do you think?

In space, no one can hear your data

Erik Stabile — Fri, 18 Jan 2013 00:02:00 GMT

If you find yourself walking though the Swedish countryside and you come upon one of Banhof’s modular data centers you might think you were walking onto the movie set of an E.T. remake. That’s because Banhof, Sweden’s oldest Internet provider, designed their modular data center to look like a space station. At the center of this modular design sits an inflatable dome, suitable for office work, with four container ports. These containers, made of the same steel used to armor tanks, house the servers. They too, are quite spaceship-esque. Even the doors open like those on the Starship Enterprise.

It’s an interesting design and it begs the question, could one of these, or something like it, actually work in space? And perhaps a more important question: is it necessary?

According to a piece published in Wired.com last year entitled “Why We Need a Supercomputer on the Moon,” Wired senior staff writer Robert McMillan describes the impending “deep-space network jam,” and how building a supercomputer on the moon might alleviate the problem.

But don’t send Banhof’s module to the moon quite yet. It costs around $10,000 per pound in shipping to get materials into space, and $40,000 more to get it to the moon, and by the looks of it those Banhof server containers are heavy. So it’s unlikely that Banhof designed their new modular data center for space use—but I'd like to hope they, and other manufaturerers, are testing the water.

New Applications are Increasing Rack Power Demand

Bob Parente — Mon, 14 Jan 2013 18:42:00 GMT

Do the most with what you have.

New application deployment in the data center continues to drive increased rack power demand. Contributing to the increasing power demand is the progressive adoption of applications in the cloud and the equipment required supporting them. Internet Companies, Social Networking, Gaming, Corporate Control application use (such as CRM applications), and wireless technologies are examples that continue to adopt virtualization technologies (Hardware, Software & Application Virtualization) and application redundancy that require powerful servers, high speed networking and storage equipment. This typically correlates to increasing rack densities.

As data centers find it necessary to place more & more equipment within each rack (increased equipment density), more power is required (increased power density). With more equipment and the need for more power, generated heat increases. With higher density racks, maintaining equipment uptime remains a key issue; where any failure can potentially influence other equipment if best practices to control and monitor the power delivery are not utilized.

Rack configuration changes necessitate capacity planning. Implementation of a solution that can monitor energy use per rack and/or per IT device will provide insight to energy usage and prevent power under-provisioning failures and over-provisioning waste. More importantly, the best solutions would provide trending &/or predictive analysis to predetermine power limitations and aid with changes that will require provisioning for more power.

As the rack equipment and power density increases, it becomes vitally important to choose a solution with the tools to maximize energy efficiency. Today more than ever, it is important to do the most you can with what you have. As new applications and technologies are deployed, understanding each racks power demand continues to validate the need and benefits of power monitoring tools that provide power capacity planning, trending and predictive analysis.

Photo: Some rights reserved by kamshots

Procrastinating in the Data Center? Then Try the Gravity Simulator!

Brandon Siri — Fri, 11 Jan 2013 09:59:00 GMT

Happy Friday Data Center Experts! I hope your data center is up and operating well. No tripped circuits? Available power maximized? All your power is safely monitored? C14 plugs are plugged into C13 receptacles? Quality PDUs mean you are good to go!

Never mind installing that new HP Gen8 Server today….that’s Monday work. Today is Friday, and it’s time to geek out for a bit. If you are an aficionada of terrestrial objects perched in an interstellar medium, void of stellar winds and supernovae, then read further.

Computer power consumption has never been better utilized than it is hosting this interstellar flash game where you can send digital ‘solid’ particles hurling towards infinity in a 2d space. The catch, each particle has a specific gravity based on its mass (Tiny to OMFG). It’s a fun toy. I can say that I have spent almost an hour just testing different scenarios – flinging different sized matter in all sorts of different trajectories and seeing what fallout would occur. Brilliant! This is a fail safe way to achieve a fulfilling Friday afternoon. Make sure to add this little game to your data center checklist for the day. You have to try this now:

http://www.nowykurier.com/toys/gravity/gravity.html

CramIT! - A Datacenter Blog: Your New Year's Resolution

Marc Cram — Mon, 07 Jan 2013 19:08:00 GMT

Image: Some rights reserved by Chris Dag

Don’t read this! It’s time to turn your 2013 New Year’s resolutions for improving yourself and your datacenter uptime into reality.

Capacity Measurement and Capacity Planning – An ounce of prevention is worth a pound of cure. Set time aside each month to review how much power capacity your data center has remaining, and to take a forward look at what you will need in the coming year. Look at ways to consolidate loads, turn unused servers off, and update to more power efficient hardware.
Vendor management – Put an hour into your weekly schedule where you will allow vendors to come in and meet with you. If the vendors cannot accommodate your schedule, so be it. But use this time wisely to educate yourself and to cultivate potential strategic partners for your future efforts. The knowledge you gain and the relationships you can build from this could be instrumental to your success going forward.
Educate yourself and your organization – Still haven’t networked your intelligent PDUS because the IT security team doesn’t want them on the network? Take time to learn how SSH, Radius, and TACACS+ can be used to provide a high level of secure access to the power infrastructure of your data center. Then have an informed discussion with the security team about the benefits versus the risks of making the PDU infrastructure accessible on the network.
Automate – While trite, the phrase “work smarter, not harder” applies to the datacenter in spades. Find ways to leverage the tools you already have to better manage the hardware and software you have in place.