We have some web, database and application servers that process requests filled in on a website.
A few times each year this is open for requests, we get a high ammount of users connecting to the website filling in these requests.
We notice on all servers a big increase on CPU when a lot of people do this. CPU will go to 100% and this causes problems. On all these servers CPU is really high.
How can we optimize our CPU commitments to these important machines?
We have 3 hosts with 6 servers that need to perform optimally on CPU.
The hosts have 2 sockets 12 cores and 48 logical processors.
The servers currently have 1 virtual cpu and 4 cores.
What would be a optimal setup for more performance, keeping cpu readytimes, overcomitting etc in mind.
How many virtual cpu's and cores should we assign to one server.
1. Use ESXTOP and See the virtual machine performance. Share the Matrix here if you can. Also you can use vRops if it is configured in your environment.
2. Specifically look for %idle %ready and %costop
3. Do not increase vCPU in case %idle is high on that VM.
4. Enable CPU and Memory reservation.
5. Set Power Managment setting to "High Performance" at ESXi as well as On Hardware BIOS Setting.