VMworld 2014 – SDS Data Protection = Recoverpoint for VM

I’m super-pumped that Recoverpoint VM is going GA (formal GA date is in Q4 – right around the corner).   This is enterprise-class software-defined data protection and availability.

Put it this way… If you LIKE the idea of vSphere Replication – but:

• want more scale (more than the hundreds or so VMs you can do with vSphere Replication)
• better RPOs (better than the 15 minute or so async replication you can do with vSphere Replication)
• need enterprise-like behaviors like consistency groups (i.e. two or more VMs that need to be replicated together, and crash consistent to the exact same point in time – VERY common) and consistency group sets (which vSphere Replication doesn’t do)
• need more efficient WAN use (better than the minimal efficiency of vSphere replication)

… Then you’ll LOVE Recoverpoint for VM!

Look, I’m not saying that vSphere Replication isn’t good for some – but many need more than it can do.   Recoverpoint for VM can:

• Scale up!  The design target is for more than 10x the scale of vSphere Replication.  This isn’t as far as Recoverpoint physical does – which is thousands of consistency groups, gigabytes/second of replication traffic. Recoverpoint supports some MASSIVE customers, and I’ve seen clusters of vRPA configurations that do more than 300MBps (that’s 2400Mbps) of traffic.   That’s BIG.   I suspect that we’ll be able to take Recoverpoint for VMs further over time  and closer to the current maximum Recoverpoint scaling point.  In fact, making sure that Recoverpoint for VM can hit the scaling targets is the focus for the engineering team now.
• Can deliver async/sync – and the async behavior is tunable from minutes to seconds, and can also be expressed in other ways (rates of change)
• Can support huge numbers of consistency groups (VMs that get replicated with complete synchronicity together – regardless of the RPO – so they are in a crash-consistent state).   This is a key capability for many classic “Platform 2” applications that depend on infrastructure resilience.   Furthermore – there’s an ability to group consistency groups themselves into “group sets”.
• Recoverpoint is crazy efficient on WAN from a replication standpoint – it eliminates duplicate changes that occur between replication RPO periods, compresses and dedupes the traffic.  It does this much better than general-purpose WAN acceleration – and at moderate to large scale, the cost of the WAN itself can be a huge part of the TCO of a DR solution, and no one does this better than Recoverpoint.
• Recoverpoint doesn’t use a a snap/replicate replication engine (this is how vSphere Replication works, or most array-based approaches), but is a continuous replication engine.   The upside of continuous replication is that both locally and remotely – you effectively have a DVR, any point in time can be recovered to – and in general the richest set of replication capabilities.   The downside (and IMO, this is a small one) is that it requires a “journal” (a place to store the changes – and needs to be sized to keep up with the rate of change).   BTW – over time, we may make a Recoverpoint “mode of operation” that is a “downshift” to snap/replicate for people who don’t want continuous point-in-time replication.
• But – all these improved capabilities come in a super-simple package.   Look at the demo below for yourself to see how easy it is to use (and remember, we’re actually showing not the “easiest demo path”, but a more comprehensive set of use cases).   The install is also designed to be super-simple.

Recoverpoint VM is totally hardware agnostic – the splitter is embedded in the vSCSI portion of the IO stack, and the vRPA is a VM – so you can layer this on ANYTHING (including VSAN/EVO:RAIL, but frankly any storage solution who’s remote replication isn’t up to snuff)  - so interesting for almost everyone (including Service Providers looking at making a really, really good DR as a Service).  To understand better, hear a “story behind the story”, and see a cool demo – read on!

 

BTW - one of the “stories behind the story” on why this has taken some time to GA was for a while we focused in on vVol io filters as the way to do this – but with vVols pushing over the last few years, went back to the original prototype that was built about 3 years ago.   Of course, final product-ization (including all the vCenter integration) isn’t a simple task either.   But – it’s an example of how focused EMC and VMware engineering teams are on “what is the RIGHT way” to engineer something, and it’s really important to build these on top of APIs that won’t deprecate (thus breaking the solution).   This is a particularly tricky spot in the storage stack, where there are a narrow (but growing!) set of ways where 3rd parties can integrate.  BTW, when vVols GA – Recoverpoint VM will support the vVol IO filters.

Perhaps BEST OF ALL:

Recoverpoint will (in early Q4) join the growing list of EMC products (and I’m on the warpath to make this ALL products that are software-only) that is available with: non-fettered access; no necessary “talk to a sales person” step; and no necessary licenses/timebombs.  Go to town. 

Just remember – it’s the Fedora/RHEL model – if you use Fedora, you know that your only support vehicle is the community.  If you need support (and if I were deploying in production – it’s insane to deploy without the normal enterprise support vehicle) – license it!   The licensing model is simple and per VM.

The best thing to do is to see for yourself – check it out!   The demo starts with literally deploying the 3 tier app in that we’ll protect, so it’s “bottoms up”.

Thanks to the great EMC team that made this demo video: Brian Tobia, Brandon DaCosta, Jase McCarty, Bruce Hansen plus of course the awesome DPAD Recoverpoint team!

So – what do you think?  Are you a Recoverpoint customer?  Do you dig it?

Comments

[Jim Hathaway]

All this about RP for VM and only 1 sentence about VAIO?! I hope you're saving that for a future post...

[R]

Hi Chad, we've already invested in choosing SRM as our BCP/DR solution for virtualised workloads and were hoping that vRecoverpoint would give us the synchronous replication missing from vSphere rep.

However, based on what we've been told, there are no plans to integrate (maybe via an SRA of sorts) the SRM orchestration layer over the vRecoverpoint engine. This leaves us with a considerable dilemma as from day 1 it seems unlikely that the vRecoverpoint offering would have parity on the SRM feature-set from a orchestration/management perspective (nor would we expect it to).

Can you shed any light on whether there are any future plans to see these two products mesh a little better?

[Chad Sakac]

@R - thank you for being a customer!

There is a plan, but it requires the "next gen" SRM. SRM has at it's core only an externally available API for SRAs that deal with "data stores" as the unit of failover. vSphere replication hacks into there, and we're exploring it.

This is why we built some of the SRM functions (but not all as you point out) as the near term option.

Working on it!