DR/BC design relating to replication

[ZiPPy]

I've been working on a DR/BC + SAN performance design between our data center and SWITCH Supernap.  In short, we wanted some serious storage performance in our data center so we've chosen to go with EMC XtremIO.  At the other end over at SWITCH, we'll be placing a VNX5200 array.  We'll be using EMC RecoverPoint for continuous replication.

I was reading the Storage Evolution chapter in the Data Center Virtualization Fundamentals book, and I came across a statement that caught my eye. It stated, "Commonly, these replication technologies do not permit write I/Os on the replicated data from the secondary array either".  Prior to that statement, SRDF and HDS was mentioned.  I'm thinking they were specifically referring to those technologies, but it just got me thinking this could be a limitation using RecoverPoint.  I'm still researching but I haven't found anything.  I haven't checked directly with EMC yet, because they are promising me the world right now.  I haven't signed for the gear just yet, so I think the answer will be "it all works" just fine.

The reason for this concern is, if in fact a disaster occurs and our data center is gone the end goal would be all users connect from another location and/or home and continue working out of SWITCH.  All data will be mirrored between sites up to the minute, respectively.

Any thoughts on this matter?  I really want to stick with XtremIO, so the VNX was merely an add-on bonus to close the deal.  I don't see the point in wasting resources and putting an XIO at both ends.

Cheers,

[deanwebb]

Hmmm... could you show the whole paragraph? That could help interpret the statement.

[Otanx]

Not a storage guy, but I can play one on the internet. I would think that it does not work as an active/active type system, but if your primary goes down that the secondary array would take over as the primary, and start to accept write I/O. Of course there are a ton of things with storage that make me scratch my head so I wouldn't be surprised if I was wrong.

Side note: If you come out for the Switch deployment let me know, and we can grab a beer. It is just down the street from me.

-Otanx

[ZiPPy]

Yea, I'm still thinking about this one.  I'm not super worried, because let's be honest here what's the chances of our shit completely dying and everybody has to login to SWITCH.  Watch I just said that, and they will be down by Friday.  But seriously, even if they never used it I want to make sure the design is correct.  Simple because I designed it and will build it, so it has my name on it.  I'm starting to think out in left field now, maybe it's the beer I'm drinking right now.  But what if I had different VSANs, one purely for off-site DR where it would meet HIPAA, PHI ect type requirements and then had another VSAN for 'continued workload' and I would have some kind of dump from one VSAN to another, or in that aspect ...somehow.  I'm just spit balling right now.

I actually already deployed the DR site at SWITCH about 10 months ago.  I just have a few NAS devices and servers, over a couple 100Mbps links connected into the SWITCH mesh network.  This design is will make it how I always envisioned it, but I needed to at least get my foot in the door at SWITCH.  I was actually out there about 3 weeks ago picking up a server to build out one of the clusters and just do some maintenance.  I'll be out there again very soon, so I'll take you up on that beer.

Cheers,

[ZiPPy]

I had a long meeting this morning with a Systems Engineer over at EMC, who got into the nitty gritty of this design.  The answer is yes, we can continue working from the VNX array after a failure.  Once the XIO array is back online, the VNX array will replicate back over and you could switch production back to the XIO.  This is all using EMC RecoverPoint. 

The arrays should be on-site in about a week, so I'll definitely be testing this pretty thoroughly.

[deanwebb]

That sounds about right.