Aubrey Li wrote: > On 7/23/07, Cyril Plisko <cyril.plisko at mountall.com> wrote: >> On 7/23/07, Aubrey Li <aubreylee at gmail.com> wrote: >>> Hi list, >>> >>> I got a chance to measure the power consumption on linux and solaris, >>> and of course on the same machine >>> The following is the result from the power meter. >>> >>> 1) Ubuntu 7.04, linux kernel 2.6.20-15, Zero workload. >>> No. Watt PF Vol Cur >>> 0 375 .871 PF, 237V, 1.816A, >>> >>> 2) Solaris community express, solaris kernel onnv_b67, Zero workload. >>> No. Watt PF Vol Cur >>> 0 408 .892 PF, 237V, 1.93A, >>> >>> Any thoughts? >>> > > Firstly, I would correct the data I posted on the last mail is not credible. >>From several times experiments, I found there are 2 factor affecting > power consumption results so far. > 1) Environment temperature. Different environment temperature results > in different result. > Today the air-condition is the lab is fixed, and the watt from the > power meter lows down to the 380W. > > 2) CPU temperature. The first time testing and the second time testing > has different result. > That means longer the testing runs, higher the power consumption is. > >> Aubrey, >> >> it may be interesting to rerun your tests with build 70, >> which is closing today. The SXCE usually follows in >> two-three weeks after that. It is this build, that includes >> support for Intel Speedstep technology [0]. I guess it >> may make a difference. >> >> >> [0] http://opensolaris.org/os/community/on/flag-days/pages/2007071601/ >> >> > I checked just now, between Build67 and hg tip revision 4700. There > are no big difference.
Forgive my ignorance. What is the hg tip revision 4700? > As the resolution of the power meter is 1W, I got almost the same > result from those two build rev. They are jumping to and fro in the > range (380W to 381W). > > There could be other factors impacting the power consumption value > that I didn't image. > So, how can we get the credible data to driver our power optimization work? I don't have a good answer. But it seems to me that the data that we are most interested in is comparison data between different Solaris instances? That is, if we want to test an expected improvement, then we'll need to do a baseline run first. We can't have a set baseline that we always compare against. Mark
