Probably the most fundamental pieces of equipment in your IT asset arsenal are servers. But they are also probably the most disregarded when it comes to looking at efficiency. Many Data Center Infrastructure Management (DCIM) solutions concentrate on everything else in the facility but not on the humble server itself.
With virtualization and cloud computing now buzzwords in every enterprise, the idea that you could actually decommission servers and run your agency processes on a small server estate is attractive. A recent survey by 1E and the Alliance to Save Energy shows that 15 percent of your servers are not doing any useful work. The average cost of running a physical server for a year is $4,400, with a virtualized server at $1,000 (based on information in the Server Energy and Efficiency Report by 1E and the Alliance to Save Energy). But by introducing server power management and efficiency analysis, you could switch off as many as one in six of your servers, an attractive proposition.
Qualifying the problem
All too often, as agencies expand, so have their server farms. The problem is in not knowing if all servers still add business value. There may be servers employed for a project that has been shelved or for one that is delayed, but once put into service, server managers are loathe to switch them off, even temporarily. Although you may invest in the most efficient servers available today, it may be of little consequence if you do not run them efficiently. According to Pike Research, servers use about 60 percent of their maximum power while doing nothing at all and most only ever use a fraction of their available compute resources.
The public sector needs insight into the performance of physical and virtual servers and needs to identify the level of useful activity on them. If there is no useful activity, then your investment is giving you absolutely no return. In this tougher, tighter economy with soaring energy costs and upward inflation rate, federal agencies in particular, which are accountable to the public, are having to become leaner and more watchful of their budgets. This applies to all areas and IT is no exception. As computers and servers are great consumers of energy, IT is looking at how to reduce their energy draw without disrupting end users. Leveraging existing assets and their power saving technologies to the full extent will contribute a greater return on investment.
Server power management and efficiency analysis
If there are old servers (five or more years) in your estate, you will be likely to find their components were designed to run at maximum performance all of the time. Much like some new cars have technologies to save energy (such as engines which stop as the driver brings the car to a halt) you will find new servers are able to adapt their performance to save energy by running more slowly. Running software which is aware of when the server is no longer working on business critical tasks (analogous to stopped car) means these technologies can be used to great effect.
This awareness can be used to identify servers which are never used by your agency. Maintaining a dynamic list of such assets is valuable as it avoids unnecessary spend on new hardware. It often makes sense to keep a small amount of spare capacity to hand, akin to having one or two spare light bulbs. That said, aging servers are often of little use, even in an emergency, so age of server is an important factor when making repurposing or recycling decisions.
Virtualization has driven uptake of brand new servers and naturally flushed out old servers, yet has introduced its own set of inefficiencies. There is concern over increasing virtual sprawl which is partly due to a blithe approach to commissioning new servers. In a 2010 study conducted by 1E to measure federal agencies’ awareness of and ability to comply with energy management rules and regulations, 58 percent of respondents were aware of the issue of virtual sprawl, yet 29 percent did not monitor their virtual servers’ performance to determine if those assets were doing useful work, or the tasks for which they were bought and provisioned. This leads to an increase in the numbers of virtual servers which become an unmanaged liability and additional cost to the enterprise.
Measureable savings, improved ROI
It is pointless running servers uselessly; wouldn’t it be great to know where your spare capacity resides? If you gain visibility of server power costs, efficiency and CO2 emissions, you can then make informed decisions about which servers to switch off or recycle. This will help you avoid sprawl, and achieve improved payback through efficient use of technology instead. After all, you want all your assets to be working for you and running efficiently. You wouldn’t expect to have 15 percent of your workforce doing nothing for you every day, so make the same demands on your IT estate.
Robert Monine is director of federal programs with 1E.