Installing a brand new vSphere. Server has 8x300GB local storage, grouped in 2 sets of RAID 5, and 32GB RAM.
Not knowing how many VMs will eventually be hosted on it, my questions are:
Why 2 sets? You can do more by increasing your spindles, and splitting the RAID on 4 disks instead of a big RAID on 8 disks will essentially mean you are losing 600GB of space, and less IO as a result.
If you setup a RAID 5, that's 2.4TB. Losing 1 disk for parity. That's 2.1TB, you could allocate say 200GB to ESX install, and leave the rest for datastore, and you don't need to split the RAID. Then keep the performance gain of 8 disks.
but in particular I want to know the size of /swap partition. I've read somewhere to allocated 8GB, while other places it's recommended to do 1600MB max.
Swap is for the ESX CONSOLE, not the server. If you run ESXi you don't need swap AT ALL. So this should be default, increasing it to 1600 is a waste, you will probably never run anything IN the console anyway, so no reason for swap.
This server is primarily (and initially) will be used in a DR site. Is there a specific configuration that will optimize the data protection, as opposed to performance? And if later on it's used a production server, will I be able to do the configuration switch easily?
Once you configure it, that's it. To reconfigure means moving VM's and starting over.
We have ESX 3.x & 2.x host in-house and we have a 100MB pipe between our facility and DR site. Is there an easy way to migrate/replicate those with VMs on them to the DR site?
there is but that is a very slow pipe. WE are talking MAX 10Mb/s speed transfer, and for a 30GB file you are looking at around 2-3 hours to transfer. That's if you only transfer one file at a time... so yes there are ways to migrate / replicate but there are much better 3rd party tools to do this, like VizionCore or Veeam.
