hi all,

i am new to clustering, and as i understand it the simple way to go is with openmosix.

my friend is an admin of a windows network with a bunch of web and other services. we were wondering how rational it would be to run these boxes instead as virtual machines on top of a linux cluster.

i have googled around, and read some docs, but there is really so much out there i don't know where the best place is to start.

beyond that, here is what we have:

-a bunch of spare x86 pcs, and some running x86 servers
-everything is on windows 2003
-all kinds of services, too much for a migration to linux actually

what we hope to get:
-a system with good failure redundancy (ie. 2nd vm takes over where first virt server failed)
-high availability
-secure (obviously windows is responsible for the network security, no flames please ;-)
-easy to administrate, add and remove nodes with different hardware

what i wasn't so sure about was vmware server, and whether it ran in only 1 or 2 threads, or more. my understanding was a cluster could expedite jobs with lots of threads, or execute many compiles/whatever concurrently, but not spread the tasks associated with 1 thread over many machines.

what would be good is if eg. the web server, when swamped, could spread the work over the cluster. i don't know how trivial this is to implement.

can anyone direct me to a good starting point? anyone had similar experiences?

thnx

Clusters are really designed to split large, computationally intensive jobs across multiple systems. In order for something like VMWare to run on a cluster, it must be written to be cluster-aware. AFAIK, there's no virtualization software that's cluster aware in this fashion.

What you're really after, I think, is some combination of virtualization and failover. In this scenario, you'd have n machines running virtualization, with multiple Windows instances running on each machine. The Windows instances themselves would be set up to fail over to a Windows instance running on another machine in the event of a failure. For something like that, VMWare's ESX would work well:

http://www.vmware.com/products/vi/esx/esx_features.html

I've done this sort of thing with Linux--created mutliple VM's on multiple hardware platforms with applications failing over to VM's on other hardware in the event of a hardware failure. It allows you to build a pretty robust environment.

Chris