The Effect of Removing Server Power Supplies

We had a great chat session a few weeks back that took us down a strange path. A customer on the chat informed us that they remove one of the redundant power supplies from their servers in order to reduce power consumption. His original question was how to remove the Red-X from the ITA console so they would stop getting error messages.

Then the discussion turned into a, "how much can you save by doing this ?" discussion. Since both Todd an I were in the lab at the time, and we had the server in question and power monitoring equipment, we decided to do a quick gut check of the difference in power consumption. This first very "unscientific" test consisted of Todd yelling out numbers from behind the rack and me typing stuff in the chat box.

The discussion came up again with Fred. Fred talks to customers at the Briefing Center on a daily basis. Fred told me that he's had other customers who are doing the same thing.

That discussion prompted me to spend some more time in the lab and do a more thorough investigation of the differences. The results are about what we expected. Older servers have more inefficient power supplies, so they see a greater difference than newer servers with more efficient supplies. Also, as the workload increases, the percentage difference decreases because the inefficiences represent a larger portion of the measurement at idle workloads.

Can someone tell me where this trend started ?

Scott - Back to SystemsEdge Blog