As the proud parent of a teenaged son still in high school and a daughter in college, I’m amazed at how much our conversations relate back to simple, basic principles of energy efficiency in a data center. Oh sure, you’re saying – those aren’t the kinds of conversations we have, Jack. Okay, so only a few, but they are worth examining.

Data center efficiency has many faces, and one can always learn from their kids. So with that intro let’s move into CUPS (Compute Units Per Second) as described in an Emerson white paper on energy efficiency and metrics, and kids.

So the other day my son and I are chatting about my recent evening of playing Halo 3 when I suggested that my game is improving significantly. My son quickly says “Hey Dad, glad to hear your scores are improving and your rank is going up,” to which I reply “Well, not really. But man I’m having fun, met some neat people, not dying quite so quickly, and I’m getting a better feel for the new maps.”

Then my son, who needs to learn when to leave a wanna-be gamer in his moment, says “But Dad, if your score isn’t getting better how can you say your Halo game is improving?” I won’t bother you with my reply about no more allowance, forget the iPhone, and who do you think bought you that car anyway?  But we should consider the implications to data centers.

I can hear my son asking a data center operator – so how do you keep score in this thing? His point on “scoring” is well made. How do we score/rate our data centers?

Data center efficiency can be “scored,” as my son would say, with relative ease if we understand two simple aspects – the power that goes in and the relative useful work that it produces. Notice I said relative and not actual. Why?

That’s because only you, through your business policies, IT practices, and HR polices, can define “useful work.” So for now we have to make an assumption that if the data center is alive, well, and executing applications – be it sophisticated data regression analyses, Monte Carlo models, video streaming, online gaming, online sales, MRP/ERP, supply chain management, or simple e-mail and web browsing – then the business has determined that these applications need to be served and are important to the success of the business.

So we start with the basics – power in. Yes, it is that simple – just measure the power coming into your data center. What? You are in a mixed-use facility, you say, with one giant meter at the campus entrance and a shared chilled-water plant. Or perhaps you’re in a high-rise with numerous tenants, apportioned utilities, and shared chilled-water. Well, nobody said it would be easy.

That’s why the first recommendation is, if your power-use scenario is not 100 percent clean – meaning you know the source of all power and where it goes in the data center – then it is time to bring in the professionals. Either contact your local utility provider or your local Liebert Services professional, who can provide you with an energy assessment that will help you identify the flow of power in your data center from building entrance right down to your IT assets. With that in hand, you can now begin to develop a framework for “scoring” your data center.

But you could use an even easier baseline if you are in a hurry to “do something,” as is often the case when management comes screaming down the hallways looking for answers NOW. Start with the output of your UPS. Ideally, you will have this monitored in real-time and reported to some form of management program.  However, you could use sneaker-net and a clipboard with daily readings say at 10 a.m. during what should be peak use. At the very least this gives you a baseline number to reference for improvement.

Because let’s face it – the IT assets connected to the UPS are driving your data center. The lights wouldn’t be on, and the service technicians and application writers would all be at home if not for the data center. So a simple UPS reading gets you started. Note that UPS isn’t perfect, but if you do not focus your improvements on the “useful work” side of the load equation you probably are not getting the best bang for your buck.

Now comes the tricky part – somehow quantifying the output or useful work. Let me point you to a simple tool based upon CUPS (Compute Units Per Second) that will relate performance capability and percent utilization to total power. You’ll receive a score that you can use over and over again to gauge the improvements you have made in your data center. And guess what – you can create a report to management that reflects the CUPS score for your data center over time, thereby showing them just how much more work per Watt your data center is producing today than it was last year or the year before.

And a really neat aspect of taking this simple step to create a CUPS/Watt model for your data center is that unlike online interactive games you can actually predict your future performance based upon changes to your IT asset mix. This will make going to management and asking for the latest next generation multi-core, multi-threading based wonder box a lot easier to justify than going to the wife and asking for a new large flat-screen TV to boost your Halo scores. (It didn’t work for me, by the way.)

So, in summary, here’s how to keep score of your data center’s relative performance:

1.  Start with total power in
     a.  Sneaker-net works! (But will not relate two different data centers)
2.  If not available, seek professional help and an assessment
3.  Identify your current IT asset performance
     a.  CUPS tool
     b.  May require an up-to-date inventory (you need one anyway)
     c.  Percent utilization is required (again – you really need to know this
4.  Divide IT MCUPS from tool by Power from audit / clip board
5.  Determine CUPS/Watt score for your data center
6.  Report to management
     a.  Track monthly
     b.  Watch closely with large IT and system changes
     c.  Trend with workload, season, etc.
     d.  Use to justify new equipment and vacation
7.  Go online, post your score, and brag to your buddies (just kidding)