Phatsta wrote:
Yes, that would be true if the local storage were all hardware raid and running a raid 5 or faster, I guess. But one of the servers are only using single disks since it only supports software raid, and another server has slower SATA drives in a raid 5 array so it's not very fast either. The last server has a raid 5 SAS Array which probably would be faster than an IP SAN.
Oh, in that case, I'd spend my $5-6K budget on a new server with good local storage controllers and SAS drives. In RAID 10, not parity RAID configurations, assuming you have non-trivial write IOPS requirements. You'd pay for on-site support for a single machine instead of 3, which might allow you to budget for a higher service level (4 hour response or 6 hours to resolution).
You could keep around one of the servers and upgrade the local storage controller/drives to act as a replication target.
But it's not the only reason for getting one. Another reason is availability. If all VM's reside on an external storage, it's easy just to add it to any hardware in case of "emergency" (or maintenance). I'd rather depend on a high end SAN than on backups made nightly to a low end NAS
I don't think that actually increases your availability. If all your VMs reside on external storage and the external storage goes down, your entire environment is down. You still have one device can fail and bring down your environment. For 12 guests (or is it 12 guests times 3 hosts for 36 guests total?), your baseline config would be a single host with good backups and a good on-site service contract. Stepping up would mean a second host as a replication target. If you go to external storage, you need the kind of multi-chassis device (which would have two complete copies of your data) which is outside of a $6K budget.