Number and size of CSVs guidance

[Slingshotz]

I know that hardware specs will have a large influence on the answer to my question but I am trying to get a general idea of best practices. For reference, I am running three nodes on Dell servers each server having three RAID 0, 1.2TB SAS 15k drives, dual 10 gig links for sync and iSCSI channels and dual Xeon 2.2GHz CPUs.

My question is, which is generally better, have a bunch of smaller CSVs, like ten 600GB CSVs or two 3TB CSVs? My thinking is that if a node crashes and comes back, Starwind can synchronize smaller individual CSVs quicker than a few very large CSVs thus allowing the servers on those smaller CSVs to come up faster.

If this is true, then I assume that the priority on the HAImage is what determines which CSV to synchronize first? And can this be changed? I am not talking about the client request vs synchronization slider option, I am talking about the priority under the serial ID of each HAImage where it says first, second or third.

[anton (staff)]

Typically you want to have at least one per physical host. As we work within an IOPS "tunnel" and StarWind virtual LUN can do ±60K IOPS (sub-100K on a very fast machines and fast flash back end) you'll probably want to spawn more virtual LUNs and CSVs on top.