Business was booming to the point that the company outgrew its enterprise resource planning (ERP) system. It was time to upgrade to new software, larger storage, and faster processors. The VP of manufacturing had no trouble getting approval for the upgrade and planned a one-month rollout process. No problem.
Actually, it proved to be a big problem. Hardware upgrades, software installation, and data conversion could be completed in a month, but that was just the start. Space limitations compelled the company to migrate to high-density servers and storage devices that required more power than the company’s power system could handle.
The power system overhaul would have been so costly that the company ultimately decided it was faster and cheaper to deploy the new ERP system in an entirely new building with new power systems. The end result: a cost overrun of more than 100 percent, and a wastefully underutilized data center.
Sadly, this scenario is typical. Even where there is ample budget and space to support IT growth, is there enough power? Often the answer is "no." This essential commodity, once taken for granted, now factors into IT planning decisions in a big way, for several reasons.
Your IT, Networking, and Telecom Systems Consume More Power Than Ever
A typical rack or wiring closet of equipment used to consume about 1.7 kW or 2 kW. A few years ago, that would have been high; it would be minimal today. For example, a rack full of servers today consumes around 12 kW. Fill that rack with new blade server technology and power consumption more than doubles.
How are you going to distribute clean, manageable, reliable power at that scale?
The capacity of your power system is only as great as its weakest link, only as high as the lowest rating along the chain. For example, suppose you need 6 kVA downstream, and you have 208 volt power coming in at 30 amps, but downstream breakers are only rated for 20 amps. Effectively you only have 20-amp, 208-volt service, derated 80 percent to 16 amps, per the National Electrical Code. If you sized every element of the power chain to avoid such bottlenecks, you would be wasting a lot of money. You would also sacrifice the flexibility to downsize, if necessary.
Power is proving to be a limiting factor in data center growth. For instance, while it is physically possible to put 42 1U servers in a single rack, that load would exceed all but the most ambitiously overbuilt power delivery systems – and the rack would be as hot as an oven. Hot spots of power consumption or cooling requirements create a big strain on facility systems.
Some managers address this concern by spreading high density equipment across several racks. Instead of attempting to pack 42 1U servers in a single rack, they might configure a rack with three or four 1U servers, interspersed with less demanding equipment, such as traditional 2U or 3U servers and networking equipment. This strategy levels the load on power and cooling systems, but it adds cost and complexity. Reliability may be compromised by the difficulty of maintaining widespread systems.
Your Power Requirements Will Inevitably Change
Business units within your organization are seeking to satisfy ever-changing business requirements, and they turn to their IT teams to provide the systems to make that work. They aren’t expected to consider all the behind-the-scenes logistics of making that happen, and they don’t.
Power isn’t even one of the top four worries on IT managers’ minds, according to the Network Computing survey, "This Old Data Center" (2005). However, power systems can make or break the business case, now that power costs are approaching 50 percent of a typical enterprise IT budget.
Suppose the organization needs new servers to support new applications, plus redundancy to meet Sarbanes-Oxley requirements, plus disaster recovery provisions for critical systems and data. You could be adding components today and scrambling to find the cables and power outlets to serve all those different plug standards.
Then, who knows? Next quarter the scenario could be reshaped by server consolidation, downsizing, corporate mergers, organizational upheaval, or technology upgrades. Flux is the reality of IT today.
Unfortunately, this flux brings unwanted complexity from a power perspective. How do you address this reality?
u Do you overbuild the capacity of the entire power system – from main switchgear to individual power drops – to be ready for future growth that may or may not happen?
u Do you maintain a large inventory of diverse power distribution units in anticipation of changing power drop requirements? Or do you plan on keeping an electrician on retainer, to call in whenever the configuration changes? These are costly and cumbersome propositions.
u Do you run power cables in spaghetti fashion everywhere you could potentially need a power drop to support new equipment? Even the availability of a simple outlet can become a major restriction to IT flexibility.
What is the Best Strategy?
The data center – one of many at the telecom service provider – was a Utopian paradox.
The gleaming, top-of-the-line facility had expansive floor space, twin power utility feeds, and the latest new equipment. Sitting in the middle of the vast data center, like lonely chess pieces, were six racks – about 4 percent of the data center’s potential utilization.
"Build it and they will come," had been the philosophy for establishing the luxuriously over-sized data center, as well as the company’s other data centers. But the people didn’t come. Ultimately, the service provider joined the annals of good intentions that failed, and their data centers were sold for pennies on the dollar.
Overbuilding is still the norm, even though this approach is costly and risky. Constant change brings a cascade of complexity. Since power costs are not going to decline, and power requirements are not going to stabilize, what is the best strategy? How do you deliver power to IT equipment in dynamic organizations?
The answer is to seek flexibility and adaptability in the power infrastructure, just as you seek in IT, networking, and telecom systems. However, the chameleon ability of the power chain must preserve power quality, availability, and reliability.
There are many theories about how to achieve flexibility and reliability; many organizations gain consistent results from the following strategies.
Standardize on a Set of Core Attributes
If you buy products that have consistent energy ratings, plug types, and management interfaces, you can greatly reduce the complexity of the power chain. Simplicity always saves money. For example, fewer plug types means less need for diversity in receptacles, power strips and power distribution units. Consistent energy ratings enable diverse equipment to be effectively served from a single UPS. Consistent management interfaces enable your system administrators to move easily from one system to another. Knowledge gained with one system can be applied to comparable systems across the enterprise:
Deploy Standardized, Pre-Engineered Power Solutions
There has been a lot of focus in recent years on pre-engineered, modular infrastructures to house communication and IT systems in data centers and wiring closets. This approach provides a pre-tested, ready-to-install infrastructure that is naturally more robust and flexible than piece-part solutions assembled on-site, often in an ad hoc fashion.
Why limit the benefits of structured solutions to servers, storage devices and communications gear? Pre-engineered infrastructure solutions can deliver equivalent benefits for your power system. Structured wiring enclosures are available to accommodate power quality, power monitoring and power distribution equipment as well.
Prefabricated wiring that have been engineered and tested at the factory, then shipped with all components in place within the enclosure have been proven beneficial. At your site, installation is easy, and your new wiring closet is a standardized part of the complete power chain. Structured wiring solutions address power needs in a cost-effective, energy efficient manner.
Benefits of a Structured Wiring Solution
For new or retrofit applications:
u Reduce space requirements by an average of 50 percent
u Reduce construction cycles by one or two weeks
u Mitigate job site risks with a pre-tested configuration
u Deliver clean and reliable power on demandu Configure modular components for specific requirements:
– Surge suppression
– Power quality and backup power
– Filtering and mitigating harmonics
– Emergency lighting
– Building management
– Rack-mounting IT equipment
Look at Both Sides of the Availability Equation
Mean time between failures (MTBF) is a good indicator of the robustness of power equipment, but the reality is that electro-mechanical devices, no matter how well made, will someday falter or fail, or at least require service. So, when evaluating equipment, it’s important to consider mean time to repair (MTTR) as well as MTBF. Both metrics should be assessed, measured, and optimized for the entire power system, not just for raised floor equipment.
In a perfect world, all the components of your IT environment would be classified in the top left corner of this chart. They would run a long time without a hitch, and be quick to fix if there ever was a problem. In reality, most electrical and mechanical infrastructure for IT falls into the top right of this chart – engineered solutions that run trouble-free for a long time, but are troublesome to repair if the time comes.
Achieving good ratings on both measures, MTBF, and MTTR has been somewhat of a balancing act of conflicting forces.
For instance, the industry moved toward modular components that could be easily swapped out for speedy repair. This approach often added complexity that made systems more prone to failure in the first place. Simpler designs with fewer components lead to greater reliability but can increase time to repair.
Today’s UPS products can optimize MTBF and MTTR through strategies such as these:
u Build self-diagnostics and predictive failure analysis into components and solutions.
u Reduce the number of components to reduce potential points of failure.
u Use standardized, off-the-shelf components wherever possible.
The result is a power back-up solution that is highly reliable, gives advance warning of potential issues, and can be quickly repaired with widely available replacement components.
Choose the Most Energy Efficient Components You Can Find
The more efficient the components of the power chain, the more real power available for your IT systems. With best practices and the right choice of equipment, data center managers can reduce energy consumption by nearly 50 percent.
If you manage a data center – or you care about the profitability of an organization that has one – flexibility and adaptability have become critical issues. But the IT infrastructure is only as flexible and adaptable as the power system that feeds it.
The good news is that technology solutions and best practices are available to mitigate the limitations of previous-generation power systems – and substantially raise the bar for new installations. Either way, you just have to ensure that power considerations are included in the planning horizon.
Ed Komoski is the vice president and general manager of power quality protection division for Eaton.
"Appeared in DRJ's Summer 2007 Issue"