Have you considered using NFS rather than FC?
It would allow you to have larger volumes (as you're not constrained by the same VM/datastore limit due to ESX locking on block based devices) and you'll therefore no longer need fractional reserve and will get better dedup savings?
If that's not a possibility, you'll have to consider whether 20 VM's per datastore is ideal; When I last used FC, best practice was less than that.
Darren
-----Original Message----- From: owner-toasters@mathworks.com [mailto:owner-toasters@mathworks.com] On Behalf Of steve klise Sent: 01 June 2009 21:39 To: toasters@mathworks.com Subject: ESX volume design question
What would you do? ESX 3.5 running all windows servers; FCP with all fiber drives We have an existing 6030 with 1 aggr, 1 large volume with ~5 LUNS 500GB a piece; running DOT 7.24. (upgrading soon, so no dedupe running yet). We have a new HA 6040 (PAM cards) =). The existing volume on the 6030 is still at 100 percent fractional reserve.
We have SMVI running in our test environment, but not in prod yet. We wanted to snapmirror from the 6030 to the 6040 and visa versa. The 6040 is running 7.3.1.1 so I can only go from the 6030 to the 6040 for now. No vmware on the 6040 yet. Question: I have 3 shelves each for each head (6 300gb total for the 6040) and 1 big aggr on the 6030, 11.5TB agr 4.97used in the volume on the 6030.
I was thinking of having the volumes for the 6040 @ 750GB and 1 LUN. I am averaging ~20 windows boxes. I figured when SMVI runs, it is only 1 snapmirror update vs running the job with allt he LUNS in 1 big aggr and multiple snapmirrors happening. I will make the fractional reserve set to 50% on the 6040 with thin provisioning at the LUN level, and dedupe.
Pros: I think its more granular. Having to find a location for 6TB is a bit difficult
Con: Not as much dedupe by having all VM's in 1 volume.
Thoughts, experiences, gotchas.?