I have a Red Hat Enterprise 4.6 virtual server built on 1 virtual disk running Oracle Applications on VMware ESX 3.5; the performance of the virtual server is not good because of IO bottlenecks. The ESX server is reporting minimal load, it's the virtual server which is struggling with disk IO because everything is on the one virtual disk.

I can create almost any number of virtual disks to spread the load across but I don't know what the best number and arrangement of disks would be or how to migrate the existing system off the original disk (which is too big and I plan to trash) and onto the new disk arrangement.

I have found out how to move the swap space onto a new disk using 'parted' and 'mkswap'. I tried it on a test system; I don't want to touch the live system until I am very clear about what I am going to do.

I was thinking about having a single new disk with three partitions for "/boot" (100MB), swap space (2GB) and a data backup area (xxMB). And then three (or more?) disks in stripe formation (using LVM) for performance (no fault tolerance required as the virtual disks would be on a RAID5 SAN) to put the main application on with another three-disk stripe for the Oracle database.

Unless I am forced into it, I don't want to create any additional servers to spread the load across. I just want to get better IO performance from the single server that I already have.

Any suggestions regarding the plan?
Remember I can create almost any disk arrangement I need (which is nice).
So what's going to give me the best performance?

And how do I actually migrate everything off the original disk?

Thanks for your suggestions and help.