VEEAM over VMWare vCenter + hp.VSA

[CCastellanos]

Greetings VEEAM Community.

I would like to submit and share a platform design I have been working on for the past few weeks. Nonetheless, if anyone in the community had done similar work I would like to read and add my impressions rather than opening a new topic.
The concept is to create a separate VMWare vCenter deployment dedicated for VEEAM Infrastructure core elements/roles. I am using vCenter vSphere (ESS+ or up), 3 HP G7 Nodes and SDS based on hp StoreVirtual VSA.

Best regards
Carlos Castellanos

[dellock6]

Hi Carlos,
what is the use case for having a separated cluster?

[CCastellanos]

Hi Luca, thanks for the time and question and my apologies for the delayed response.

In my case there are various appealing reasons to evolve (perhaps) into this direction. Allow me list some of them:
- By virtualizing all VEEAM core infrastructure elements on a 3 Nodes vCenter cluster we add a level of resilience to our platform. We now gain HA on these VMs (e.g. VEEAM-DivA (primary for Division A), VEEAM-DivB (primary for Division B)
- We also add multiple factors of compute power we can distribute across these Veeam infrastructure VMs (scale UP)
- We can scale OUT both our Veeam infrastructure by adding new VMs (Proxies and Repos) and our underlying vCenter Cluster (licenses permitting)
- I am also adding a central SQL VM for both DivA and DivB and Enterprise Manager handling.
- We also gain a very important factor and that is isolation or dedicated resources. In other words, Veeam infrastructure elements like Virtual Labs can even reside on this very same dedicated platform without consuming production elements that could potentially affect business and/or require further planning (at production level).
- Now, by adding hp StoreVirtual VSA we also add much needed (by own experience) redundancy at the storage layer. We are actually adding a SDS layer with all its perks (and yes, its own management requirements).
I take that a similar approach can be accomplished with other solutions but I doubt at the same price point.
- hp VSA license allowing in our case we are also adding a storage acceleration layer with the implementation of Adaptive Optimization (AO).
- Allows for VSA scale OUT capabilities and even (slightly more elaborated) more than 2 tiers deployments (adding SSD (Tier 0)+SAS (Tier 1) on AO and then High Capacity SATA for a Tier 2/3)
- All other benefits your vCenter license allows
- Can actually expand to "fancier" solutions as you are now on a vCenter platform (e.g. PernixData or the likes)
- Pretty much as one would build a DEV platform, an IT platform (APM, NPM, etc), I believe Veeam with its almost software define backup and replication approach (or availability as the new trend) deserves and can benefit from its own pool of resources dedicated and isolated from the elements it serves. Not necessarily new as you can accomplish this with Windows 2012R2 scale out, SS, etc; but different in my view on various implementation stages as explained above.

[dellock6]

Thanks for the explanation, it sounds good indeed.
I was at first sight a bit worried of having repositories inside the VSA cluster, but it's not the production cluster where also VMs are deployed, so it's fine. As a heavy user of the VSA myself in the past as a customer I like this design. One thing to add: as Veeam has better chances to work at the application level when joined to the production AD, you may want also to have an additional domain controller into this cluster, so you don't use DNS and AD authentication if any major problem happens to the production AD.

I've solved sometimes this by keeping the Veeam components in a workgroup and using another dns, but the downside of this is the loss of AD options when doing restores and other things.

Luca

[CCastellanos]

Thanks Luca,

Good recommendation on DNS and AD within the cluster. This is indeed a pain point I have dealt with before as well during other DR events.

I am now working on v2 of this platform aiming at two sets of VSA Clusters, Same Management Group (MG). One set keeps AO but the other aim at larger SATA drives for ARCHIVE and COPY. This one will be on larger forma factor G8 servers.

I do have a few questions since you mention you are a big VSA user?
- Is it confirmed VSA even when licensed has throughput limitations inherent to the code?
- When using VSA AO on drives (SSD and SAS) connected to RAID Controller (e.g. P410, P420) I have seen mixed recommendations on whether RAID Controller Cache should be turn on or not for SSD Volume (in my case 2xSSD on RAID 10 per node), what is your take or experience?

Best regards
Carlos Castellanos

[dellock6]

The differences in licensing involves the maximum size you can have for a node (free is 1TB, than you can grow) and the AO that is not available in the free edition. I'm not aware of any difference in performances between editions.

Cache: I've never checked this deeply, but ar first I'd say to keep it disabled at the raid level, as this is already a sort of caching system enabled at the software layer, and just use the raid cache for the spinning disk group only.