[lug] duping minimal vm question

[karl horlen]

i didn't want to hijack the other recent thread but came across this nugget which i've been contemplating myself recently for a different purpose.

> > Maybe you could setup a virtual machine per client on
> the data fetch server
> > to provide more isolation. Depends on how many
> clients. With a known
> > transfer schedule, the VMs don't even have to run all
> the time.

i haven't decided which virtual manager to use yet but i'd like to roll out a server which isolates one vm per client.  each client would basically need a minimal lamp stack for a website but nothing else.  i want to keep it as simple as possible. questions:

- what's the best way to install a minimal centos install on a "master" client / guest vm?

- once i have that master guest, is it possible to simply use it to dup multiple future guests based on teh master?  if so, how easy is it?  

- i'm also wondering how many vm guests i can realistically expect to rollout on this server before performance suffers?  i know that's going to be fairly subjective based on the performance requirements of each guest website so guess i'm on my own here.  anybody have a formula / method to estimate usage to quantify this before setting up the vm guests?

fwiw, i imagine each virtualization implementation has it's own method[s] so i don't need specifics unless you want to share, especially since i haven't picked my vm implementation yet.  but if you feel one VM is generally better than another overall or at guest vm duping per scenario above, info appreciated.

thanks