> RE: power monitoring every second. Of course you'll see fluctuations; you > are influencing that by making the MC work hard to poll the sensor, queue the > data, send it though the communication interface, etc... Long ago I tested > systems like that and soon realized that the sampling caused a 40 - 60 watt > increase in power usage (or error) versus sampling over a longer period of > time. I compared the results with an external watt hour meter. There wasn't > much of a difference between the longer sample period average and the real > watt hour meter. I dug a little deeper into this and I gotta wonder what type of system you saw such levels in. Here are a couple examples of collectl output, measuring several sensors every 10, then 1 and finally .1 seconds (in the last case I chose to include msecs in the times) and I didn't see any real impact on power:
# VFan Temp1 Temp2 Temp3 Temp4 Temp5 Temp6 Temp7 Temp8 Temp9 Power 09:46:58 37.240 46 33 30 30 32 30 30 57 22 206 09:47:08 37.240 46 33 30 30 32 30 30 57 22 204 09:47:18 37.240 46 33 30 30 32 30 30 57 22 206 09:47:28 37.240 46 33 30 30 32 30 30 57 22 206 # VFan Temp1 Temp2 Temp3 Temp4 Temp5 Temp6 Temp7 Temp8 Temp9 Power 09:49:33 37.240 46 33 30 30 32 30 30 57 22 204 09:49:34 37.240 46 33 30 30 32 30 30 57 22 204 09:49:35 37.240 46 33 30 30 32 30 30 57 22 204 09:49:36 37.240 46 33 30 30 32 30 30 57 22 206 09:49:37 37.240 46 33 30 30 32 30 30 57 22 206 # VFan Temp1 Temp2 Temp3 Temp4 Temp5 Temp6 Temp7 Temp8 Temp9 Power 09:50:15.426 37.240 46 33 30 30 32 30 30 57 22 204 09:50:15.502 37.240 46 33 30 30 32 30 30 57 22 204 09:50:15.629 37.240 46 33 30 30 32 30 30 57 22 204 09:50:15.702 37.240 46 33 30 30 32 30 30 57 22 204 09:50:15.831 37.240 46 33 30 30 32 30 30 57 22 204 My main interest is all this is to try and correlate power to applications by having a finer grained view of the impact of what an application is doing. I too doubt 1 second sampling is really necessary but I do think 10-30 second samples could prove useful. -mark ------------------------------------------------------------------------------ This SF.net email is sponsored by: SourcForge Community SourceForge wants to tell your story. http://p.sf.net/sfu/sf-spreadtheword _______________________________________________ Ipmitool-devel mailing list Ipmitool-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/ipmitool-devel
