7 Ways To Kill Data Center Efficiency
(Image: Dario Ruglioni via Pixabay)
Considering the constant pressure on IT budgets and the "green IT" push that's been going on over the past decade, you would think that IT would have its act together when it comes to identifying and resolving data center inefficiencies. But according to a report released June 3, this is far from the truth
The report -- by Jonathan Koomey, a research fellow at the Steyer-Taylor Center for Energy Policy and Finance at Sanford University, and Jon Taylor, a partner in resource consultancy firm Anthesis Group -- talks about what it calls "comatose" servers, which use electricity but have delivered no information or compute service for six months or more.
The authors cite a 2008 McKinsey study that found that server utilization in enterprise-class data centers "rarely exceeds 6%." Additionally, the authors said, upward of 30% of servers were identified as being powered on but sitting idle, or "comatose," citing research by McKinsey as well as the Uptime Institute and TSO Logic.
Koomey and Taylor estimate that there are 3.6 million comatose servers in use worldwide today. It's these types of numbers that get CIOs and IT managers in hot water. Yet, for many of us in IT, the numbers won't come as a surprise.
I've witnessed inefficiencies in the vast majority of data centers I've been in. It's not limited to long-forgotten or over-provisioned servers. The problem extends to other areas of the data center, including powered but unused or unnecessary network equipment, security tools, and even over-cooling of data centers. These factors conspire to create massive inefficiencies, since power is the No. 1 operating expense at any facility.
Koomey and Taylor said in their report: "In the 21st century, every company is an IT company, but too many enterprises settle for vast inefficiencies in their IT infrastructure. The existence of so many comatose servers is a clear indication that the ways IT resources in enterprises are designed, built, provisioned, and operated need to change. The needed changes are not primarily technical, but revolve instead around management practices, information flows, and incentives."
Indeed, as we look at why these inefficiencies exist, you'll see factors that have little to do with technical issues. Instead, poor oversight and management by those operating the data center, or managing equipment within it, are causing most of these problems.
While it's impossible to approach system utilizations of near 100% or make sure that every device that is powered on is actually serving a purpose, there's plenty of room for improvement. For example, some of the servers identified in the report as comatose are likely to be servers that are set up as hot-standby or disaster recovery systems in case there is a failure. If that's the case, does the system truly need to be in a hot-standby mode or would a cold/warm standby mode be more cost-efficient and provide nearly the same amount of redundancy?
Here are seven ways you can identify -- and resolve -- data center inefficiencies. If you have your own theories on why inefficiencies occur or how to resolve or prevent them, tell us in the comments section below.
Planning For The Unknown
IT administrators tend to overbuy data center hardware to account for unknown projects or pivots in business strategy. And while it's acceptable in some cases, this strategy can quickly get out of hand. Instead of keeping idle equipment boxed up and stored, it's typically installed and kept powered on and ready to go in the unlikely event the equipment ever becomes needed.
No Decommissioning Strategy
As applications are replaced with modern systems, the old systems (and data) linger because nobody knows what to do with them. There's a common fear I've seen among data center administrators that as soon as the application or data goes away, someone will ask for it.
No Idea What It Is, But It's Been Here Forever
In enterprise organizations with high turnover rates among IT admin employees, you'll find data center components hanging around because nobody knows what they are, and they're too afraid to ask. Instead of taking the risk and turning down an unused system, new hires let the components live on forever.
Shift To Shadow IT
Users who are unhappy with an application managed internally by the IT department are taking to shadow IT to get what they want. They're able to implement a Software-as-a-Service app, for example, to meet business needs without involving IT. As users migrate onto the shadow IT app without telling IT, the internally managed app and related hardware sit idle.
Over-Provisioning In Anticipation For Growth
When it comes to right-sizing hardware for application requirements, most admins don't perform proper due-diligence. They typcially figure out the minimum requirements for memory, processing, and storage, and then go far beyond this recommendation when building the system. The rationale for this approach is to give the application plenty of breathing room in case the need for it grows exponentially. More often than not, those lofty application demand predictions never happen. Considering how easy it is to modify these settings on the fly using virtualization software tools, over-provisioning servers should already be a thing of the past.
Poor Record Keeping, Lack Of App Audits
Performing regular audits and maintaining an up-to-date hardware/software/storage inventory is one of the most time-consuming, boring tasks for a systems administrator. Within enterprise-class data centers, losing track of applications and servers can happen relatively easily if records are not kept up to date.
No Drivers To "Go Green"
In many cases, there is no incentive for data centers to spend time tracking down inefficiencies because there is no obvious payoff for the IT department. In situations where colocation space is rented and power is already paid for as a flat, "per-whip" monthly cost, the incentive to power down unused systems is even lower. Metered power billing is becoming increasingly popular, though, and could change the dynamic here.
Conclusion
IT's dirty little secret regarding data center inefficiencies is quickly leaking out. Once upper management gets wind, they're going to be all over this. The best thing for IT do is start figuring out how to identify and quickly eliminate all the low-hanging inefficiencies within your environment. In many ways, once this becomes a company-wide priority, the shift in philosophy should make your job much easier. It's better to start planning now so you can be applauded for being proactive instead of criticized for being reactive.