We all know that data centers pull a lot of electricity off of the global grid.
Just how much? Worldwide, digital warehouses now consume about 30 billion watts of electricity, roughly equivalent to the output of 30 nuclear power plants, according to estimates compiled by industry experts for The New York Times. With server utilization rates often in the single-digits, the reality is that much of that energy is simply wasted.
Needless to say, energy management at data centers is now more critical than ever—not only for reducing environmental impact, but also for protecting a company’s bottom line.
Of course, we can’t just turn servers off. From business to government, everybody relies on immediate access to data in order to keep revenues up and everything running smoothly. And we all know people who melt down if they don’t have Facebook for a day.
So the challenge we all face is this: what can we do to save power, cut costs, and go green when service level agreements and end-user experience are always on the line?
The fundamentals of saving energy
When it comes to saving power at data centers, the standard menu of options is focused on capital-intensive infrastructure changes: from facility retrofits, to server virtualization, to all-new energy-efficient hardware. On top of those, data center infrastructure management (DCIM) solutions allow operators to coordinate everything more effectively.
All of these solutions should certainly be included in plans for new builds and retrofit projects. But before you can even get to that point, there are a number of fundamentals that need to be addressed:
Good data is indispensable. Do you know how many transactions you can process per kWh? How does that number change over time? Collect everything you can so you can answer these kinds of questions—before, after and during your changes.
In order to understand where your power dollars are going, it’s not enough to look at big-scale aggregate stats. You need to go down to the level of individual applications and servers. And if you don’t think that you’re paying for data center power, look again—you’re paying for it somewhere, and it is all negotiable.
Based on my experience, most companies will only embrace sustainability if it has a clear value proposition. Everybody from facility operations up to the C-levels needs to understand how energy waste affects the bottom line. You should be able to show exactly how costs trend alongside revenue, and you should be able to assign business value on a per-application basis.
It used to be that buying more servers was the solution to everything. With the right data, you can replace the expensive “brute force” approach with informed capacity planning that accomplishes much more with much less.
Don’t get caught scrambling to deal with surprises. Every change to your applications and workload affects power consumption, which can all be planned for in advance when you’re properly equipped.
Get started with your current infrastructure
The industry has been choosing from the same menu of infrastructure-focused solutions for quite a few years now, but there are newer options that don’t call for major infrastructure changes. That’s because most data centers already have a huge opportunity built right in.
One example: online shopping in the US is at its peak around Black Friday, when there are millions upon millions of transactions processed each day. Then, in January, the online traffic subsides. What happens to all those servers? The short answer is that they keep spinning and wasting massive amounts of power, even though the revenue is no longer there to justify it.
We can take advantage of these peaks and valleys in demand with a new, software-based approach called application-aware power management.
When you stop typing on your laptop, it’s smart enough to tell that you’re taking a break and put itself into a low-power state. Servers can do the same, but they can’t do it safely without fully understanding the incoming workload.
Application-aware power management software takes a detailed, application-level look at your incoming workload, and figures out how many servers or virtuals you need to support that demand perfectly. Then, based on algorithms as well as admin settings, the software automatically controls the power state of servers, returning everything to full power as soon as there’s an uptick in workload.
It’s not just a matter a shaving off a few kWh here and there. Typical data centers that we work with have the potential to save as much as half of their server energy costs with this approach. Needless to say, that’s a big slice of the pie all with one solution.
One last thought: equip people for change
Power reduction sometimes runs against the traditional “always-on” mentality at data centers. For good reason, IT administrators tend to worry whenever they can’t hear the hum of a server or the roar of a chiller.
The other day I was on a flight. One of the crew got on the intercom to tell us that they save fuel by actively adjusting their jet engines, and that we might hear a difference in sound as this happens. This is a great example of proactive communication that helps people anticipate and adapt to change.
By combining that kind of communication with detailed, deep data and total visibility into power consumption and performance, and you’ll have everything you need to start saving without compromises.
Before founding TSO Logic, Aaron spent 15 years building and managing large-scale transactional solutions for online retailers, with both hands-on and C-level responsibility in data centers around the world. He has previously held positions with PNI Digital Media as president and COO, Fuji Film as their vice-president of web development and operations, and with Microsoft’s digital imaging team. TSO Logic provides software for monitoring and reducing energy consumption at in-house server farms and large data centers. Their Application-Aware Power Management™ software allows data centers to automatically control the power state of servers based on application demand, without impacting availability or performance and without changes to infrastructure. www.tsologic.com
