Here is the scenario:
We have two virtual machines running on the same machine using hyper-v, one is a database server and the other is a web server.
I am analysing performance information based on web requests. Each request to the web server also results in requests to services hosted on the same virtual machine as well as calls to the database server.
The information is taken from PsList which is running on both virtual machines and from JMeter which performs the requests. The information includes memory usage and cpu usage on both virtual machines over time as well as the time taken per request. I may be wrong but pslist
My first (noob) question is how to interpret the cpu usage of each virtual machine (given as a percentage). Is this the percentage of cpu that has been allocated to that virtual machine that is used, or is it the percentage of cpu usage on the actual machine on which the virtual machine is running? In other words, would you expect the total cpu usage of processes (including idle processes) on both virtual machines to total 100 or 200?
My second question is whether there is a better way to measure the performance of both virtual machines that could show the resources that are being used by each and by the host machine itself?
Many thanks, Nigel.
-
Measuring the performance from inside the virtual machine with regards to how 'much' cpu utilization (ie: a % of what the OS sees as available) will most likely give highly inaccurate results.
Here's an excellent write up on performance monitoring in Hyper-V.From Adam
0 comments:
Post a Comment