SuSE

hi,

currently we have SAP application running on a solaris machine that has 8 dual core CPUs @ 2.4 GHZ and the performance of the system is perfectly normal. We plan to migrate the app to a suse linux VM image on a vmware box now. So my questions are:

1) Should i size the linux image to also have the same number of cores or some number of CPUs ? Or should it be more/less considering the fact that it is suse and its a VM?

2) Since this linux is a VM image, it will get VCPUs allocated to it. So lets assume that there are 2 CPUs on the physical host and these CPUs are dual core, so does that means 4 cores are shared by all the VMs on this physical host ?

3) How about memory ? any memory requirement conversion matrix on solaris vs suse & any conversion matrix of a Suse on physical vs Suse on VM.

I know i am asking too many questions. Please answer as per your knowledge and experience. thanks.

Hiya....there are some very general statements there, and we probably need to delve a bit more deeply. Can you first post the exact spec of the solaris machine you have ?
What version of Solaris? Is it Sparc/intel ? 32/64 bit ? how much memory, how many physical CPU's, what model and speed CPU's, how much disk and of what speed, and whether there is hardware or software raid ?

Then I think you need to qualify the statement that the performance is "perfectly normal". SAP may be running normally on 1 CPU, or it might be using all of them. We need to get some proper measurements of the performance of the machine over an extended timespan. This article shows you how to measure such things, especially tools like "sar" that can collect data, historically over time.

Once you have that data you can then start making decisions about what new hardware you can run it on.

I hope this helps...

Also, you cannot allocate more CPUs than you have. In a VM, with VMware you are not allocating real CPUs, you are allocating access to the CPUs as time/considerations permit. Your application may run like a dog, or it may not notice at all.