Power Utilization Effectiveness: Where’s the Beef?
Mark Thiele, Executive Vice-President, Data Center Tech at Switch, has responsibilities across several functions, including industry thought leadership activities and working with the team on new data center technologies, partnership development and cloud.MARK THIELE
My Power Usage Effectiveness (PUE) is 1.10, so my data center is better than your data center running at a PUE of 1.20? Could be, but not necessarily. As most people in the data center field know today, PUE was created to help provide a common, standardized metric for determining the relative “effectiveness” of your use of power across the data center and associated IT systems. I’m personally a fan, and have used PUE to help me build and manage data centers against a pre-defined goal of cost-of-ownership. However, the use of PUE does come with significant caveats, like the fact that higher server power use makes your number look better, or even more importantly, how much did you pay to get to your PUE?
Cost of Ownership vs. PUE numbers
Many of us in the industry are happy to shout about how great our PUE number is. At a minimum we should have a simple way of certifying and validating whether the “achieved” PUE is real, but we should also be able to determine what the likely “annualized” PUE would be.
However, what I’m proposing is that for the PUE number to really be meaningful, it has to be combined with the other newer metrics of water usage effectiveness (WUE) and carbon usage effectiveness (CUE) from The Green Grid, and most importantly, it also needs to be measured against cost-of-ownership.
That’s right, cost-of-ownership is important in the effort to achieve any type of efficiency. After all, what is the point of getting 40 miles per gallon, if I pay 2X as much for basically the same vehicle that gets 25 miles per gallon? A little simple math looks like the following:
Measures Vehicle 1 Vehicle 2 Delta Cost 25,000 12,500 12,500 MPG 40 25 15 Average yearly miles 12,480 12,480 0 Cost of gas per year (based on $4.00 per gallon) 1,248 1,996 748 Cost based on 7 year life 8,736 13,972 5,236 Investment value based on 2% per annum return (1,750) (1,750) Total Cost 33,736 24,722 9,014
According to the simple chart above you can see that while vehicle 1 was almost 60% more efficient from an MPH perspective, it would cost you $9,014 extra over a 7-year period.
All of the sudden your gas mileage benefit seems more like a boat anchor.
What am I suggesting?
I’m not suggesting that you look for cars with poor gas mileage, but I am suggesting that you should be careful about your buying decision and weigh your decision factors very carefully. In the case of PUE and your data center, the issue is exactly the same as the car example, only it involves a lot more money over twice the timeline (15 years vs. 7 years).
Those of us who are making a living building data centers for our companies need to step up and justify how we’ve reached our PUE number and, more importantly, what it really cost us to get to that number. When buying a car the gas mileage isn’t your only decision factor, and neither is the price of the car. However, once you’ve decided on what you actually want/need in a vehicle, there is the potential of comparing one against the other on price and value, with mileage being a factor. In data centers, the cost drivers are more complex and involve a much greater sum of cash, but they can be compared.
What should we do?
The first suggestion is think long and hard about whether building a data center for your company is in the best interests of your company and its resource investments.
1. How does a data center differentiate you from your competitor?
2. When you consider the cost/time involved in planning for and then building a data center, can you justify it against other business opportunities?
3. If the decision to build still makes sense, make sure you have a plan to successfully own the facility. Here’s some tips:
- Consider using the Data Center Pulse Data Center Stack (PDF) to help you manage the entire data center as a system.
- Think long and hard about whether or not your company is willing to set aside the resource and funding to manage the data center effectively over a 15-year lifecycle
- Keep in mind that the “shiny new toy” won’t be new in 2 to 3 years and your team and/or your leadership might have changed. Can you be relatively sure that the same focus and energy will be applied to the data centers maintenance and support?
Generally speaking, your best bet will be to develop a clear understanding of what your company needs from a growth and geo-distribution perspective. Then, find a partner(s) who can take on the risks and overhead of building and managing data centers, allowing you to perform function for your business that differentiates it from its competitors.
Industry Perspectives is a content channel at Data Center Knowledge highlighting thought leadership in the data center arena. See our guidelines and submission process for information on participating. View previously published Industry Perspectives in our Knowledge Library.
[...] fellow Datacenter Pulse (DCP) college Mark Thiele wrote a good article on the use of PUE in the datacenter industry. He basicly argues that you should look at the TCO of [...]
Jan WiersmaPosted November 22nd, 2011
My 2cents on the subject: