HP VSA - HA failover / shared storage + multi site question

[dannyboy1121]

Having a look at the VSA solution and I'm pretty impressed.

I have a couple of qustions though which I'm hoping someone can help me with:

The config guide appears to show that the Storevirtual VSA is supported with any storage provider approved by the hypervisor vendor, so I can use it to virtualise 3rd party arrays. For me, there are some great use cases with this but the problem here appears to be cost. The VSA is not expensive in itself but if the only way I can have HA is to run network raid, then I'm immediately incuring a raid on raid tax.

I'm guessing that there is no shared storage model to avoid duplicating data? e.g. Node A fails and then the shared storage gets presented up via node B?

I have some DAS scenarios where network raid makes sense and the disk is cheap enough to run network raid and it still be cost effective. Frontending EMC / NetApp etc and having to duplicate up to half the data just kicks this in to touch for that scenario. Yes the array will have its own raid / HA protection but I have to take in to account ESXi host failures / ESXi patching and upgrades / VSA patching and code upgrades  etc.

The other question I have is about multi site.  Again, this is about the associated raid configurations and potential cost implications.

How does the raid config work in a two site solution? Is it raid 10 in that site A is raid 0 with site B being a mirrored copy? or is it a local raid 10 which is also mirrored to site B as a raid 10?

Thoughts appreciated.

[oikjn]

The VSA solution is great and gets you an availability level that is hard to beat at its price point (in $$ that is), but as you are seeing, the cost for the VSA solution really is in its raid on network raid configuration which means that this SAN provides a much lower ratio of usable:raw space than other SANs and this can become a significant cost when you are talking about 100's of TB+.  

Generally HP mitigates this by using hardware raid5 on the drives supporting the VSA and then network raid10.  

Network raid10 means there are TWO copies of your data, which means if you have two sites and you set the cluster to be site-aware, the SAN will store one copy of the data at each site.   NR10+1 and NR10+2 add a 3rd and 4th copy of your data if you are looking for site+node loss protection.

Bigger problem...  there is no UNMAP support so your thinn provisioned volumes will continue to grow until you migrate the data off them and delete the LUN.  This is a major issue in my book and HP is not saying when they will add this feature which has been "next release" since v9.5 and its not at v11.5.  I would press your sales guy on that issue since it sounds like you are concerned about space.