I have a really nice server in my home lab with lots of RAM and lots of disks. It runs Hyper-V and has been a great lab machine for me. However, for a current testing project, I started working on it using my desktop and using VMware Workstation 8 as the hypervisor. While my Hyper-V server uses Intel processors, my desktop has an AMD processor. That adds some challenge in moving workloads between environments, so I don’t generally do so.
Back to sharing my story about my lack of foresight… my desktop is an AMD six core device with 8 GB of RAM and, right now due to a failure, a single SATA disk. From a processor and RAM perspective, I can run reasonable tests in Workstation but can’t scale too far due to the RAM amount. However, today, I remembered why it’s so important to keep disk performance in mind when creating a virtual environment. My test setup requires three virtual machines. With three virtual machines running on one SATA disk and all having relatively significant disk performance needs, my disk queue length was at 10 for a very long period of time while the third VM was installing.
Obviously, I was aware that I was pushing things a bit far so, to avoid having to move everything to my lab server, I simply plugged in a spare external eSATA disk and installed the third VM there and the disk queue length came back to earth and the speed of the overall system improved significantly.
Why do I share this? Because at first, I was just merrily working away and because this was just a test build which I plan to destroy tomorrow, I wasn’t concerned about performance but it became important when the system slowed to a crawl because of one bottleneck and is a reminder that all parts of the performance equation need to stay in balance in a properly architected virtual environment.