This is where monitoring and knowing how you configure the VM's comes in. The swap file ONLY get's used IF there isn't enough RAM available. Even if for some reason Windows swaps or Linux hits the swap file at times, this wouldn't be a problem if you size your VM's properly. Thus eliminating the need for swap as much as you can. You really can't completely eliminate swap in the guest. Also consider you move ALL your swaps to the local ESX server, now what? Where is your bottle neck? That's right ON the host, which is exactly where you DON'T want the bottle neck. It's better to just let this be, keep the swaps WITH the VM's ON the SAN. The SAN/NFS has much higher IO and distributed load than does the ESX server. ESX is your CPU Memory NOT disk. Keep the disk on devices where it should be, on NFS. You won't see (or shouldn't) any load that will impact the rest of your VM's, which is another reason to properly utilize SAN disks and not put too many VM's on the same RAID group. And you can remove the VM swap file (but not internal swap) by reserving memory on your VM's, so you can experiment by using reservations.. but that's another problem, you start reserving your memory you have to by vigilante in configuring the VM's so you don't run out of memory also.
↧