Stretched vSphere Clusters? DR with SRM? Why not BOTH?

A VERY BIG release today with VPLEX 5.1 and Recoverpoint 3.5.    Why?  Here it is in a nutshell:

• EMC can help customers have true active/active storage target that has very rich local replication and remote replication in addition to the active/active property they love!    With VPLEX having an integrated Recoverpoint splitter creating point in time copies, Tivo-like continuous data protection, and also remote replicas with best-of-breed capabilities and efficiency…  Well, they are all simple, easy, and integrated.
• VPLEX has led the way on the vSphere Metro Stretched Cluster support model with VMware for many moons (welcome to the party HP Lefthand!), but the model of a stretched cluster isn’t always right.   Often people want to have a vSphere stretched cluster that is on the VMware vSphere Metro Stretched Cluster matrix – and ALSO have a simple SRM DR solution also on the support matrix.   While customers recognize that there’s no single tool for every use case – when they have a blended use case, they want the tools to be simple, integrated, and not mutually exclusive. They want an AND, not an OR.

For perspective – there are a TON of customers using VPLEX and RP.   Both are very feature rich – but have a shockingly low cost entry model – but can each scale to equally shocking levels – which means they range from small enterprises and tiny public entities, all the way up to the biggest customers in the world.

A mind-blowing fact when you consider that VPLEX is “relatively new” – there is currently 75+ PB of storage that VPLEX has made “active/active”, and in many cases, with non-EMC storage behind it.  That’s HUGE.  Some other neat VPLEX facts:

But those to new sets of use cases (active/active with rich services, and making stretched vSphere clusters/VMware SRM and AND not an OR) will create the next wave of customers.   Read on past the break for some of the nuts and bolts in these huge releases.

VPLEX 5.1 and Recoverpoint 3.5 have the usual bigger/faster/stronger stuff (and it’s pretty material).  

• VPLEX:  in pure software, VPLEX performance went up 40% (both in latency and maximum number of IOps – which is astronomical even with a single storage engine – think ~500,000+ IOPs), and scale in terms of maximum number of devices went up by 4x (to more than 1000 devices per storage engine).    VPLEX now also supports VAAI h/w assisted locking – important for scaling, as it’s the actual storage target that to the host terminates the storage traffic.
• Recoverpoint: In Recoverpoint 3.5 (which works with any customers using Gen 4 RPAs – for those on older RPAs, they will need to work with us to upgrade their RPAs), the bandwidth per RPA has more than doubled -  which is key.   You can now reasonable expect 150MBps – 300MBps burst replication from a single Gen 4 RPA running 3.5, on top of all the Recoverpoint goodness (continuous protection, compression/deduplication of data on the wire, vCenter integration and more).

Remember – while these seem “big” (they are impressive) – these are the low-cost, ENTRY config specs – and don’t come with a high price tag.

But – while “faster, better, stronger” is always key – I’m most fascinated by major changes in function or ease-of.   Here are key things in the new releases:

1.  Common, simple UI.   Like VMAX, VNX, VNXe, VPLEX now has the common Unisphere look/feel in the management model.  This is a topic EMC (correctly) has been lambasted by customers who have lots of EMC technologies: “make management easier – integrate the portfolio!”.   Unisphere first showed up on the VNX/VNXe – but as you can see, it was always intended to “unify” more than just EMC’s block/NAS management models, it is designed for common look, feel, function across pretty well everything we do, and make it based on modern UI models, and make it easy for new technologies to integrate with (and present increasingly via REST APIs too).   There are lots of other examples (AppSync being another new one).  Here’s what it looks like:

2.  Better Performance Reporting of VPLEX virtual volumes.   This was a big feature request – you asked, we answered!

3) Oracle RAC active/active stretched support.    One huge thing that we also finished was certification by Oracle of stretched Oracle RAC use cases – which is huge!   This is the Oracle “peer” of vSphere Metro Stretched Clusters, and can be a killer active/active HA model that is simple, transparent – and can be used in conjunction with Dataguard and other approaches.   It looks something like this:

…  And through a TON of work – we’ve created a GREAT joint whitepaper with a ton of partners.  This shows a real full stack example: a virtualized SAP landscape (on vSphere 5) with a stretched vSphere cluster, running on top of a stretched Oracle RAC environment – built for total active/active protection and operation in an HA model.

4) VPLEX customers now have an option for rich data services on active-active volumes.   Customers love the VPLEX architecture’s core design (a scale-out model that is truly active/active) – but historically would need to have point-in-time replicas driven at the underlying extents on the storage behind VPLEX.   This would make recovery, or test and dev automation with SQL Server/Sharepoint/Oracle/Exchange/SAP/whaever more complex.   Now – all those things (snaps, clones, CDP – what we call “data services”) are simple to do with VPLEX and Recoverpoint integration.

5) vSphere Metro Stretched Clusters + VMware SRM = awesome.    the biggest part of this release (at least to me) is that now there is a simple solution for customers wanting best-in-class, solid-as-a-rock, scaleable active-active stretched storage for vSphere Metro Stretched Cluster use cases, but recognize (correctly) that this is NOT “disaster recovery” (rather sync-distance HA), and want to ALSO have simple, easy, scalable disaster recovery which is integrated with VMware SRM.   This is the general availability of the idea that Lee Dilworth and I alluded to at VMworld 2011 here.   Beautiful thing?   It uses the same Storage Replication Adapter (SRA) that is widely used today for Recoverpoint customers – well tested and proven at customers.

 

Are VMware and EMC working together for even more (like active/active async models) – you bet.   Like I consistently point out – what you see us do/say at VMworld and EMC World every year around futures is invariably not the “way out there stuff” (we don’t talk about that), but rather stuff we are actively working on – and you see appear in shorter timeframes – like this simple integration of vSphere Metro Stretched Clusters + VMware SRM using VPLEX and Recoverpoint.

Summing this new VPLEX and Recoverpoint goodness in an elegant simple visual:

VPLEX and Recoverpoint customers – what do you think?   Right direction?   For those that have been holding off (or using something else?) – would LOVE to hear what you think too!!!

Comments

[Dominic Rivera]

So I've looked at this solution pretty extensively, and the problem that I see with it is that it's extremely difficult to manage the pathing policies on the ESX hosts, there's not a real easy way to script out and determine which LUNS belong to each side of the VPLEX. Dedicating HBAs to each side of the VPLEX makes it easier to determine which paths should be active, but you still end up in a situation where if you reboot the host you're going to lose those pathing policies ( and now youve also doubled your HBA costs ). Mabye I'm missing something? What do other customers do to get around this. I even use PP/VE and there's no value-add in that PP/VE isn't any smarter about automatically selecting the right path/HBA to make it easy to work with.

[Joshcoen]

The RecoverPoint VPLEX splitter is an awesome feature and will definitely make things easier when it comes to, among other things, volume recovery in a VPLEX environment.

You mention that RecoverPoint and VPLEX have a "shockingly low cost entry model", but with this new integration, will I now have to license RecoverPoint per TB and then turn around and, essentially, re-license that same space in order to present it to VPLEX and take advantage of this new integration? If so, I'm not sure I would consider that "low cost", but I would call it shocking.

[PJ Spagnolatti]

All cool and all great Chad... but what about the fact that Recoverpoint 3.5 STILL DOESN'T support (Clariion/VNX, not to mention SanTap splitters) VAAI?

And by "support" I mean really support it, not just reject the commands. I really can't believe VAAI is getting so little attention from EMC. What's the point of having a VAAI-enabled array if one can't use these features?

Very disappointing.

[Justin]

Stretched clusters in VM ONLY work with L2 network... EMC fails to mention this in all their slide decks... even at EMC World :)

[Craig]

You guys are on the right path, but you don't have a truly integrated "3 Data Center" solution here, yet. In your Tokyo/Kawasaki/Sapporo example above, only Kawasaki has a relationship with the async site, Sappporo. I think most customers would want either Tokyo or Kawasaki to have the same data protection in Sapporo, without having double the copies of data there (the current "workaround" to this issue). Vplex and RP need to be one product, not two and they need to embrace a true 3-site model. We need a VMware stretch cluster across two sites that can share a common RP/async relationship with the third site.