SANSymphony Virtual Appliance?

[grittyminder]

The SANSymphony / SANMelody platforms seem to be well suited for SMBs. I was browsing the DataCore website and found something mentioning that SANSymphony is now available as a virtual appliance: http://www.datacore.com/VMstoragefoundation/default.asp

So what exactly does this mean? Does this mean that if configured properly, the attached storage located in the physical machine that the DataCore virtual appliance is hosted on will be available via a virtual SAN to virtual machines located on the same, and other ESX Servers?

If this is correct, am I also correct in saying that such a setup would allow you to forgo the traditional SAN route (i.e. a completely separate storage device/appliance located on the network) and have all computing and storage needs taken care of from only a handful of physical servers while still being able to make use of VMotion, HA, and whatnot via the virtual SAN?

[Dave_Mishchenko]

You essentially have the right idea. This creates a VM that uses local storage to provide storage to other VMs / ESX host / other servers. You would configure the VM with the amount of storage tat you want to share out. It would allow to to forgo a SAN if you have sufficient local storage that is fast enough to meet your storage needs. LeftHand Networks has a similar appliance that supports features like clustering, snapshots, and remote copy .[http://www.lefthandnetworks.com/products/virtual_storage.php].

If you don't need the features of these appliances, you would also create your own Linux or Windows VM with a iSCSI target as well.

You could use these instead of a SAN, but you would have to give careful consideration as to whether or not these would meet your I/O requirements. For a place that wanted to put in a few hosts with just local storage, this would be an ideal way to be able to provide shared storage. But if you need to scale up the I/O then you might find that you need to go back to some sort of storage device that can accomodate the number of physical drives you need to meet your I/O load.

[Osm3um]

I recently heard of this option as well. I have some additional questions:

Doesn't ESX 3.5 have something similar where it can share local storage amongst different ESX hosts?

Sure sounds like an excellent option for VI3i! or am I missing something?

How does software such as this fit into the HCL? If the host hardware is on the HCL will that be sufficient?

As for the speed, I know there are a lot of factors, but given a number of factors being the same (i.e. SATA drives in Raid 1) can it really be much slower than say an EMC AX150i?

I am running about 10 VMs on an EMC AX150i with 65 users (Exchange 2003, Sharepoitn 2003, SQL, DC, etc.) with no problem althogh under heavy load it is a bit on the slow side....

Thanks for your help,

Bob

[grittyminder]

Thanks for the reply!

The SANMelody platform I think could potentially work for us as a short term solution. For example, right now I'm looking at a PowerEdge 2950 with 8 SAS disks (2xquadcore @ 3GHz, 16GB mem, RAID 10). If I get 2 of these PowerEdges and mirror them together I could conceivably 'balance' the reads across the two mirrors (i.e. by hardcoding paths to LUNs) to increase bandwidth. Just to make it crystal clear, for this particular configuration everythingthe virtual machines, the virtual SAN, etc would be located on the PowerEdges. Am I mad for even considering this?

Alternatively, I'm looking at foregoing the PowerEdge/SANMelody setup and going with 2x StoreVault S500s with 15 1k SATA II drives apiece. In this case, would the bottleneck be the CPU/controller (1x Xeon 2GHz)? I'm not sure how many simultaneous transactions the controller would be able to handle.

I think that the above two solutions could conceivably serve our needs for the short term. But as you say, we will definitely need to upgrade or add something in a year or two.

I'll look into the LeftHand offerings as well. Our budget is somewhat tight (we can go upwards of 40,000 USD); however, for that much money ideally we would like a solution that could accommodate 2 parallel storage groupings or tiers (i.e. fast storage, and slower storage for more static data) along with thin provisioning (our current storage utilization is not even 1/4 of a TB, but that should grow rather quickly in the somewhat near future).