If you want real virtual speed with your Jumpbox appliances use Xen not ESXi

As you will have seen on my blog recently we had a situation that meant I had to move one of our Jumpbox virtual appliances from a failed Xen server to a new ESXi server. Now these servers are exactly the same specification, HP DL380 G5 with 32GB RAM and 2 x Dual Core Intel Xeon CPU's , so not exactly supercomputers, but more than enougth to do what we want with them.

Well here is the thing, Jumpbox say that VMware are one of their main supported platforms and , Tier 1 partners I think they call it, and open source Xen is Tier 3. So how come then the virtual appliance runs significantly faster on Redhat Enterprise than it does on Vmware ESXi? When I say significantly, I'm not joking. The virtual appliance is the moinmoin wiki. On the ESXi platform trying to edit a post is painfully slow, 30 seconds to open the editor. On Xen it is 1-2 seconds at most, so I don't need a stop watch to see something is wrong.

I have gone through the settings for the Vmware server, something I don't have to do with Xen, and given it garanteed resources and CPU power, no difference. I could, if I fancied wasting a small part of my life, investigate the Advanced settings for the ESXi server, but again why would I when with Xen it just flys?

I don't know the answer thats for sure, and especially when you know that the Jumpbox appliance is running with the QEMU hardware emulator on Xen, as it's not para-virtualized! Emulators as I'm sure you know are not supposed to be that fast.

I think I'm going to chat to Jumpbox about this, as it really does seem odd. The only thing that may have affected the performacne is the fact that I had to migrate the appliance to ESX format from Server format, but surely once this process is completed the virtual image should be good to go.