Perf monitor improvements / fixes

[marek w]

On a dual CPU server the performance monitor kind of brakes. It still works, but does not show the % utilisations correct (in comparison to what it says in task manager), and also makes it tricky to read / make sure of due the low % server utilisation.

Therefore would it be possible to: 1) Make the performance monitor show a smaller scale? Such as every .1%? 2) See here: http://prntscr.com/9owat5 and http://prntscr.com/9owbc3 to see the perf monitor not correctly reading the cpu value

P.s. I'm just putting these here as a record of future improvements / fixes that either yourself or someone else can complete when available.

[remygrandin NA]

The problem here come from 2 things :

1. the WMI service windows call to get a server CPU usage return it as an integer between 0 and the number of CPU core times 100. (Meaning, on a 2 core computer, a returned usage of 50 (/200) will be stored as a 25% usage overall). For exemple :

   • On a 2 logical CPU cores machine, you won't be able to get more than a 0.5 % precision
   • On a 4 logical CPU cores machine, you won't be able to get more than a 0.25 % precision
   • On a 8 logical CPU cores machine, you won't be able to get more than a 0.125 % precision

   and so on ...

2. Actually, SESM store the CPU value of each perf entry as an integer, so it will be needed to be changed in order to be able to store it as a floating point value. (I will probably change it on the next release)

For the graph display themself, the Y scale is ajusted automaticly according to the data to optimise the space. If you need to read a specific value, just hover your mouse on the point and a tooltip will show you the exact data