Truth or Dare… didn’t we all play that game in our younger years. Where you lived with the consequences of a dare if you got it wrong. In the Data Center there are a few things you don’t want to get wrong, and one of them is power (outages).
Vendors put a lot of effort into designing redundancy and high-availability into their products, protocols and architectures. Customers spend a lot of their valuable money building redundant and high-available infrastructure. All that means nothing if infrastructure goes down due to incorrect power sizing.
That is why Power-Calculators exist (HP calls it a Power Advisor, I’ll stick to the more generic term Power Calculator for this blog). They are a valuable tool to assist in ensuring power sizing is done in such a way that under worst case conditions, the PDU’s circuit breakers won’t trip. Of course some customers will choose to run an (for them) acceptable level of risk by assuming not all servers will require max power at the same time. However, even for those customers, to calculate that risk you still need truthful power sizing data.
Truth or Dare… Power-Calculators are either an accurate tool (Truth) or they are a marketing tool (Dare). Since customers have no easy way to test the power consumption of servers themselves, you rightfully expect and depend on vendors being truthful with you.
I was very surprised to learn that HP’s use of their Power Calculator tool is on the daring side. Cisco got curious about the power numbers HP was publishing and decided to do the testing that customers themselves normally would not do. The result of this first round of testing (Rack Servers) is rather shocking (pun intended).
Full report can be found here:
Server Power Calculator Analysis
The HP Power Calculator is under-sizing the power consumption by 14% (average) in the wrong direction. The Cisco Power Calculator seems a bit overly cautious with a 25% (average) margin on the safe side.
This blog isn’t about who makes the most power efficient server.
It is about is making customers aware that Power Calculators cannot be used to compare server power consumption for marketing reasons. It is also about keeping vendors honest. If HP tells customers that they publish conservative data, then customers should be able to depend on that.
Here is what HP has to say about their Power Advisor (some selected quotes as teasers)
The HP Power Advisor utility reduces the research and guesswork… is intended to be a conservative estimator of power… Proprietary software exercises the processors to the highest possible power level and operates all peripherals while taking voltage and current measurements…
The entire HP Power Advisor Utility Description document is of course available to you.
My Question: Why do the HP servers consume (significantly) more power then the HP Power Advisor say they do?
Do you want to play Truth or Dare…
Recent Comments