VPLEX is one of those fascinating enabling technologies that doesn't neatly fit into any established category.
Some will call it "storage virtualization", and leave it at that. Yes, but not really.
Others will call it a "workload balancing solution". True, but there's much more.
Still others look at it as a much better answer to zero-RPO, zero-RTO business continuity requirements. Or perhaps a better way to do non-disruptive migrations at scale.
Yes, it does that, but ...
Since its introduction two years ago, the VPLEX platform has quietly enjoyed substantial customer adoption as well as a considerable amount of feature enhancement.
And, to this day, people ask me to compare it against competitive alternatives, and -- well -- I come up empty.
And not because I'm suffering from IT Vendoritis.
Sure, it's not for everyone, but if you're doing IT at reasonable scale, there's nothing quite like it today in the infrastructure world.
And it's certainly worth taking a moment to appreciate what it brings to the table.
What If You Could Spoof The Laws Of Physics?
The underlying premise of VPLEX takes a while to fully grasp: by using very clever caching technologies, it makes the exact same information appear to be in two places at the exact same time.
Once you get past your initial disbelief, all sorts of useful things are possible. Especially if you have more than one data center.
First, there's the availability angle. The farther your recovery site is from your production environment, the better.
Using VPLEX, we've been able to help our most demanding customers create zero-RPO / zero-RTO environments where return to production is measured in a small number of seconds vs. minutes, hours or worse.
But not everyone has that problem.
Next, there's the efficiency angle. We all appreciate the value of pooling compute resources with something like VMware -- what if we could pool computing and storage resources across data centers.
You know, move things back and forth easily and non-disruptively to balance utilization?
Not surprisingly, we're doing just that with select portions of our customers' environments. Today.
Finally, there's the user experience angle. Both applications and their data like to be close to the people who are using them -- performance is better, the user experience is better. And with VPLEX, it's getting easier to do just that as needed.
But there's the more pragmatic side as well. Storage migrations from old to new. Server migrations. Data center migrations. Creating better pools of resources in the data center -- both servers and storage.
You don't have to be on the bleeding edge to appreciate what VPLEX can do.
Indeed, since the pricing was adjusted downwards for the VPLEX Local version, there's been surprisingly strong adoption in more modest-sized IT settings. VPLEX isn't only for large shops these days.
Did I mention it's storage agnostic?
VPLEX is basically software (dubbed GeoSynchrony) that runs on hardened and redundant server appliances.
The software is offered in three versions: one for people who want to do this in a single data center environment (VPLEX Local), another for moderate distances (VPLEX Metro) and long-haul applications (VPLEX Geo).
"Metro" refers to a maximum of 5msec latency, or around 100 kilometers circuit distance.
As you browse the new features in GeoSynchrony 5.1, you can tell there's a lot in the new release.
If you're into performance and scale, there's good news here. First, thanks to additional multi-core optimizations, the team is claiming a 40% boost in performance from existing hardware, which isn't a bad deal. On the scale front, the initial 400 initiators limit is now 1600 -- again, good growth from existing hardware.
The team has been very focused on usability: the result is that the management console and its features have received a lot of attention, especially in the performance monitoring category.
In particular, the popular Unisphere management framework has now been brought to VPLEX. It's still a separate management console for VPLEX, but it has the look and feel of Unisphere.
For virtualized environments, in addition to the previous VSI plug in, there's now a robust VASA provider for VMware environments -- a native version for VPLEX that's separate from other VASA providers from EMC.
Perhaps the most interesting enhancements here are the first wave of RecoverPoint integrations. The RecoverPoint team and the VPLEX team are now integrated, and they've started to work to bring the products together in a very useful way, as we'll see here shortly.
The RecoverPoint Angle
If you're doing any sort of data protection, you really ought to understand the basics of how RecoverPoint works -- it's become our de-facto standard across different block-oriented storage platforms.
Continuous data protection is basically a Tivo for your apps -- you can rewind and recover to any point in time.
The technology is both very advanced and very mature, and has proven itself more than up to the task in many thousands of customer environments.
By the numbers: over 4000 customers worldwide (large and small) and a staggering 760 petabytes under replication using RecoverPoint.
The nucleus of RecoverPoint is the I/O splitter -- a small piece of software the duplicates writes and sends the copy to a different location (local or remote) for logging purposes. We've been working to put this component in more of our places: on the server, in the fabric, in the storage array -- and now within VPLEX itself.
RecoverPoint and VPLEX tend to get used together frequently. VPLEX makes an exact image of data appear in two places at the same time; RecoverPoint preserves recoverable state along the way.
So it makes sense that EMC invests in bringing the two technologies together into a more consistent and integrated environment.
And if you're thinking about zero-RPO / zero-RTO infrastructure with infinite granular recovery, you'll really appreciate these new features.
VPLEX and Oracle
There's a ton of bet-your-business applications running on Oracle (especially Oracle RAC), so there's been a great deal of work validating use cases around monster, high-transaction-rate databases that need to be in two places at once.
Most RAC shops have never invested in true active-active across two data centers at metro distances. The investment was just too daunting, but now there's a new solution worthy of your consideration.
Prior to VPLEX, high-end Oracle customers would use something like ASM in front of a rather byzantine everything-connected-to-everything-else configuration.
It's not hard to point out a half-dozen issues you'd have with something with this, especially at scale.
Not ideal, but -- hey -- people work with what's at hand.
After VPLEX, the picture is very different indeed.
The management model changes as well; the environment can be now run as a "single image" from the database administrator's perspective.
Not to mention the ability to easily add RecoverPoint for a rewind capability. It should be pointed out -- all of this is accomplished without the need for extra software on the server, database or application side.
It "just works".
We don't have any promoted public references yet, but during the briefing call, there were four very critical customer environments described where this was working quite well, including a massive 23TB production Oracle RAC database handling thousands of transactions per minute. One environment was an IT service provider who was very keen on building a premium enterprise service around VPLEX capabilities.
The final note is that this sort of solution also protects the context around the Oracle database itself: the support files and non-Oracle application elements that are inevitably part of any large production application.
At EMC World? You've Got To See This ...
VPLEX demos have become a staple at the tradeshows we go to. In particular, at EMC World we'll be showing live workloads being migrated back and forth from the EMC booth to the VCE booth being managed from an iPad. Or catch a YouTube video here.
These technologies will also be prominently featured in the hands-on vLab at EMC World if you'd like to give it a spin for yourself.
You Can See The Future From Here
I was asked a very interesting question the other day: after cloud, then what?
"Cloud" represents one recent swing of that pendulum: most clouds are implemented with massive centralization and consolidation of resources in a small number of strategically located and very efficient locations.
But consider the massive amounts of information we're all generating and consuming. There is no end in sight, is there?
Bandwidth increases, but latency isn't getting any better, is it?
Before long, there's going to be strong incentives to de-centralize once again: moving applications and information closer to where they're being used, and away from a small number of centralized locations.
If you need an acronym to help you visualize this model, try RAID: a redundant array of inexpensive data centers.
The ability to move information intelligently back-and-forth (and at scale) become paramount in this emerging world. It'll show up everywhere in the stack: the devices we carry, the applications we write -- and the infrastructure we deploy.
Given this scenario, VPLEX is a very interesting technology to watch indeed :)