Observation of CPU Ready Time

[juststormY]

Hello together,

I think i have some problems regarding the system ready time of some VMs.

The problem: Some user of diffrent VMs reported performance problems on several clusters. I tried to find out where their are based on. I monitored the named VMs with perform (on Windows) and also with the VCI client. But it was not possible for me to find performance problems on the VMs (CPU, RAM & I/O) at any time.

So i had to go deeper into the systems and found out that the named VMs have sometimes peaks in the system ready time (esxtop). I think it may possible that the performance problems are based on this (about 5-30% per VM and CPU core).

Now i'd like to monitor the ready time of the VMs on the ESX system for a period of time (e.g. 1 week) and analyze later the data. I want to have it the same way like in esxtop, but in relation to the CPU cores.

Is there anybody out, who did something like this allready? Or can anybody help me to get this informations?

Thanks a lot!

Regards, Markus

[MrBiscuit]

While it is possible to write a script to poll the cpuready value of a vm regularly, I suspect what you are looking for is one of the utilisation monitoring and modelling tools that have been developed specifically for VMware ESX.

Some examples of such products, of which there are many, are Vizioncore's vFoglight, vKernel's capcity Analyser or Veeam Monitor. I believe Vizioncore and vKernel have downloadable trials while Veeam provides a free limited functionality version.

I am not a representative of these companies, nor am I saying any product is better than another, they are simply example tools.

[Wozzer]

Not done it myself, but right at the end of this article http://vmware.com/pdf/vi3_35/esx_3/r35u2/vi3_35_25_u2_resource_mgmt.pdf is a page on using esxtop in batch mode to capture stats at intervals for a period of time, and inputting in to a .csv file for later analysis. It might help towards your task.