The most significant and positive contrast in looking at full IT business planning today, compared to yesteryear, is how big a role the power and cooling facts and projections are playing in data center decisions. Those on the leading edge of this analysis curve are a step ahead of the rest, and will soon find data center managers quickly adding this criteria to their decision making factors. Power and cooling planning and solutions are a serious part of your spend and savings in your IT investments. There can be huge cost variances, over just a few years, between differing architectures under consideration and how the workloads are managed using advanced functions. The following articles will delve into the topic of power and cooling to help you get a better understanding and offer advice as you bring this important cost-saving measure into your decision making.

Enjoy this edition of TrendSetter and look for a Datatrend webinar covering this subject in the near future.

The Scramble to be "Seen to be Green"

Everywhere you turn these days, there's heated discussion about fossil fuel consumption and rising oil/gas prices — and about how renewable energy sources are at least part of the answer to both. Everyone wants to be "seen to be green", including President Bush who, in his promotion of higher ethanol to gas ratios and tax

Not surprisingly, most of the reasons come back to costs. But before we delve into any costs associated with "greening" (many of which are off-set by the resulting savings in other areas), let's take a look at how energy costs did their part to push us all towards the green "movement".

The eternal laws of supply and demand

It appears you can never get away from the fundamental demand rises, supply goes down, prices go up principle, whether the subject is houses, cars, crops or electricity. As the charts below illustrate, the demand for electricity over the past 10 years has increased by approximately 30% which, in turn, has driven the price per kilowatt hour up by about the same amount.

Given that energy costs consume approximately 20% of the average data center budget, power consumption and cooling issues are, if not taking center stage, at least no longer waiting in the wings.

Why data centers are the "SUVs" of corporate facilities

It is estimated that a typical 10,000 sq ft data center consumes enough electricity to turn on more than 8,000 60-Watt light bulbs. This amount is somewhere around ten times the amount of electricity needed to operate a typical office building during peak hours. In many ways, the data center design is inherently "un-green":

• Centers run in an always-on 24x7 state

• Powerful machines consume a lot of energy

• Energy consumed produces heat — heat requires cooling

• Much of the cooling equipment uses toxic chemicals which, in and of itself, have a negative impact on the environment.

Cooling as the "second order" effect

Replacing computer hardware with newer and more powerful machines cannot be the stand-alone answer to turning your data center a deeper shade of green. This is due to the fact that ironically, higher-performance machines can be part of the energy consumption problem. As machines become more powerful, they generate more heat. This falls under the "robbing Peter to pay Paul" heading: what you gain in efficiency per unit of computing, you lose in cooling power requirements. Stated another way, for every watt of power used by today's computer hardware, one or more is needed to remove heat waste (heat dissipation rises with increasing chip clock speeds). And, the more heat there is to remove, the bigger your data center's cooling bills will be.

A picture is worth a thousand words.. In the front is a Pentium Overdrive from 1993, complete with its cooler. The looming presence behind it is the standard Pentium 4 cooler from 2005. Will the future bring us even larger cooling devices?

While chip manufacturers are busily working on the heat dissipation issue, data centers around the world are discovering another key part of the answer: don't just replace servers, consolidate and/or virtualize them!  (See the article "Consolidate and Virtualize to Save Big" below for examples.)

Measuring Server Power Consumption

In practice, something as apparently simple as a vehicle's fuel efficiency isn't straightforward to measure and compare. While a motorcycle beats out an SUV in raw miles per gallon, an SUV can rate higher on overall efficiency because it can carry more passengers. Then again, SUVs generally don't carry passengers to capacity, so that has to be factored in. See how tricky this can get?

Now there's a contradiction!

According to a Ziff-Davis survey of 1,200 key players in the data center industry, 71% of data center owners identified power and cooling as the primary issues their data centers are facing, yet 62% of that same group reported those concerns did not affect their server purchases in the preceding 12 months. No doubt it's a contradiction, but an understandable one if there's no standard way to measure and report server energy consumption.

The latest metric

About a year ago, a group of EPA officials, researchers, and some of the computing industry's "heavy hitters" gathered at Sun Microsystems' headquarters to develop such a standard.

Despite some of the obvious obstacles to the process (companies inevitably opting for tests that make their own equipment look good, finding criteria to represent the performance of everything from processors to storage and networking), the conference attendees emerged successful.

To capacity or not to capacity?

In July of 2007, the EPA's Server Energy Measurement Protocol will go into effect and will provide a standard way of measuring power requirements for servers at different load levels. (Such metrics already exist for desktop and laptops computers.)

Yet one more reason why server consolidation is a key part of the greening solution for the corporate data center. The more "consolidated" your equipment, the higher the utilization rate, so the lower the overall power requirement will be. (Think of 20 machines running at 20% capacity with all the necessary power supplies and cooling equipment running at "full speed" vs. five machines running at 80% capacity with a fraction of the peripheral power and cooling equipment needed).

Energy Star "seals" the deal

And the debate is?

Intel and AMD are the chip vendors currently doing battle over this metric. Intel argues that performance per watt is not meaningful "enough" to consumers since it focuses only on the chip. AMD, on the other hand, maintains it's the most concrete, "apples-to-apples" measurement in the industry.

For its part, Intel would like to see the more nebulous "Satisfaction per watt" measure adopted — a measure which addresses not just raw power consumption but noise levels (with newly-designed fan struts) and better battery life (through a process of caching display images).

"What does this have to do with my data center?"

At this point you might be asking yourself, "If these measurements all pertain to new servers, why should I care about them?" Fair enough. There are two parts to the answer:

1. When comparing "Performance Per Watt" between a current server and, say a four year old one, the differences can be really big. Server architecture has come a long way in a short time, and if you put your older boxes up against the newest, you will quickly see how much there is to be gained (and saved) by modernizing your hardware.

2. When you are ready to make new server purchases (because at some point you will be), your decision will be an informed one from the power perspective. Consider the data center manager of the Ziff-Davis survey mentioned above who claims power and cooling are the biggest issues his/her center is facing, then buys new hardware without even taking them into account.

... on improving the efficiency of your data center.

Consolidate and Virtualize to Save (Perhaps Big)

Who wouldn't want to be "seen to be green" if there was an easy way to do it which could also save you money? As we've already alluded to, server consolidation and/or virtualization allows you to do your bit for the planet and do something really good for your organization at the same time.

Less (servers) is more

There are some alarming numbers floating around concerning when "legacy" data centers will run out of power — some percentage of them as soon as a year from now! In response, data center managers are turning to logical solutions like server consolidation and virtualization as key power-saving measures.

Example 1: Consolidate with BladeCenter Grid solution

Rack-based blade servers have become synonymous with the term server consolidation. Their rack-based design allows multiple physical boxes to be "consolidated" onto one server (with one or more CPU per rack), providing for:

• Less energy consumed for the same processing power

• Greatly reduced physical space requirements

• CPUs to be turned off while not in use, saving additional power

• More efficient peripherals, i.e. power supplies (see Score One for You, One for Green article below).

Example 2: Virtualize with VM Software on a BladeCenter

Virtualization takes server consolidation to the next level by allowing for the creation of multiple "virtual" machines with one CPU. Combining a blade solution with virtualization is a "double-whammy" in terms of power savings since you reduce the number of racks needed to get to the same number of servers (fewer racks means fewer physical CPUs running).

The decision "to virtualize or not" can be an involved one which might dictate a technology assessment, especially when applications requiring different CPU utilization are involved. There is also some discussion surrounding the "break-even" point for virtualization or the ratio for server reduction (industry average is currently 8:1).

... on server consolidation and virtualization in your data center.

Score One for You, One for Green

Taking the observer stance, we might even go so far as to say you should never do something just because it is a win for green. In the IT power usage reduction movement, there are many things you can do above and beyond server consolidation/virtualization which are clearly a win for you and a win for the environment. We'll start small and work our way up.

Power supply woes

The amount of power in the data center wasted due to inefficient power supplies is nothing short of astounding.

According to a joint study by Eco Consulting and EPRI Solutions, an inefficient, oversized power supply can waste up to three times as much net AC power as a properly-sized, efficient unit. The same is true for UPS systems, which often waste power when lightly loaded.

The same study found that the efficiency of server power supplies peaks at loads between 50% and 60%, then drops off dramatically at loads under 30%. Given that the average enterprise server operates at only 20% of capacity (see Measuring Power Consumption article above), you can see how quickly the waste adds up.

What you can do

So maybe you're simply not ready to switch power supplies right now (perhaps because of warranty concerns on your current ones). Just tuck this information in your back pocket when you go to purchase your next round of servers.

When you do, keep in mind that blade servers tend to have more efficient power supplies and use less power overall than non "rack-based" servers. The latest generation of blade servers can also control the output of power supplies and even turn some off while maintaining redundancy. One day in the not-so-distant future you'll be reaping the benefits of lower energy bills when those efficient power supplies are humming along!

What you can do

To quote the experts at Berkeley Lab, data centers can obtain between 10% and 20% energy savings by switching from AC to DC power.

With the exception of monitors and other non-DC peripherals, DC can power the entire data center and since only one AC to DC converter is required at the location of input, there isn't as much power lost. By plugging some numbers into this DC power savings calculator, you can quickly see how much this change could save your company.

Of course you will want to proceed with caution, but if and when you're ready, all the major IT vendors now offer:

• DC-based servers

• Redundant DC connections and circuitry

• UPS systems which run on DC power.

Relocate or colocate?

The statistics on the sheer volume of data center square footage coming online within the next several years are simply staggering: 12 million sq ft by 2009! Just to give you a frame of reference, the United States' largest shopping mall in Minnesota, the Mall of America, covers only 2.5 million sq ft.

A colocation center (aka "colo" or carrier hotel) is, quite simply, a hosted data center — both the logical and physical components. Not so simply, it's defined as a data center where multiple network service providers locate connections adjacent to each others' networks and where users of these services place their network, server, and storage gear for the purpose of reducing cost and complexity.

What you can do

With a decision of this magnitude, a "relocate or colocate" assessment would be the first order of business. If you opt for building your own center, consider:

1. Geographical placement for maximum cost savings

2. Ways to build a green data center, e.g. the U.S. Green Building Council's Certification for Leadership in Energy and Environmental Design (LEED). The criteria include sustainable site development, water savings, energy efficiency, materials selection, and indoor air quality.

If the colocation option appears more attractive to you, there is also much assessment work to be done. What you gain in economies of scale and lower costs, you might lose in convenience and accessibility.

... on improving the efficiency of your data center.

TechTip:
Rack ‘em Stack ‘em… just remember to keep your AIRFLOW in check…

By Debi Riedel, Lead Architect, Datatrend Technologies

Today, most server and rack level cooling issues are caused due to either the inadvertent mixing of hot and cold air or insufficient airflow.

Airflow efficiencies are typically addressed during the new build or major data center design, but can be overlooked as racks and server components shift and change, as new equipment is added, and as data center personnel become absorbed in daily production demands.

The mixing of hot and cold air, a more subtle issue, can dramatically reduce the efficiency of a cooling system and also has the potential to negatively impact airflow.

To minimize these issues, here is a list of things to consider:

• Understand airflow requirements for specific equipment.

  • There are four basic airflow scenarios: front-to-back, side-to-side, bottom-to-top, top-to-bottom.

  • Understanding the requirements for specific equipment will enable an efficient rack-level design and cooling strategy.

• Map the technologies under consideration as part of an IT Design and Performance Planning activity, against the thermal characteristics specified by the specific vendor.

• On racks designed for high-density environments, standardize power and thermal policies to maximize rack effectiveness.

  • Avoid shallow racks to eliminate airflow obstructions caused by in-rack cabling.

  • Plan for adequate fans, cooling units, or localized supplemental cooling for individual racks which can accommodate future high-density systems without compromising room-wide efficiency.

• Arrange racks in rows to establish hot and cold aisles.

  • Align racks front-to-front along cold aisles, and back-to-back along hot aisles. Within each row, tightly abut the racks. Cold air needs to be delivered to cold aisles and hot air needs to be extracted from hot aisles, for this strategy to be effective.

• Use blanking panels.

  • Blanking panels improve airflow through the rack, minimize air loss and help prevent exhaust air recirculation.

• Ensure adequate airflow to individual racks and systems.

  • Clearly define power and cooling requirements for room, row, and racks.

  • Ensure sufficient airflow to racks based on system-level inlet air temperature and airflow requirements, and use thermal and aerodynamic analysis tools to model and design your cooling solutions.

Contact us | Visit Datatrend website

All trademarks, registered trademarks and service marks are the property of their respective owners.
IBM, the IBM logo and other referenced IBM products and services are trademarks
or registered trademarks of the International Business Machines Corporation in the United States,
other countries, or both. All rights reserved.

Brought to you by Datatrend Technologies Inc.

121 Cheshire Lane, Suite 700 Minnetonka, MN 55305