Brian Gallagher, president of EMC 's Symmetrix and Virtualisation Product Group, sat down with Computerworld US at EMC's annual user conference, EMC World, held in Boston earlier this month, to talk about the company's new VPLEX synchronous data replication product . Gallagher explained what differentiates it from rival products and EMC's existing offerings, such as Symmetric Remote Data Facility [SRDF] replication technology and Invista storage virtualisation software.
Excerpts from that interview follow:
So what's the difference between VPLEX and your existing replication and storage virtualisation products such as SRDF and Invista? VPLEX and its ability to do distributed cache coherence over distance literally creates a new category. It is completely differentiated, as we look at active data over distance. No other technology can do that, whether it's Invista, SRDF or competitive technology. If you look at datacentres today, they're traditionally active-passive until something happens. Then the remote datacentre can become active, allowing you to access data. What VPLEX does is it creates active-active data over distance and it manages the coherency of the access to that information.
What do you mean by active-active? Is that just another way of saying synchronous replication? In this case, yes. I am talking about synchronous for the VPLEX Metro [two or more VPLEX appliances clustered over distances of up to 60 miles]. There are still things the architecture can do, such as active-active multi-sync. However, when you look at these other-use cases, there are other things you have to worry about, like when you have multi-writes over distance. How do you keep [those writes] from tripping on each other? Our architecture supports that. As we move forward, we'll be rolling out the VPLEX Geo and Global versions of it where we'll target specific use cases for this.
Invista was designed to do zero-latency storage virtualisation to third-party storage providing for high-speed data mobility within a data centre. Why we brought [VPLEX] to market was really to do virtual storage over distance. And that required an architecture to do cache coherence.
What we will be doing with this access-anywhere technology in VPLEX is just like what we did with our FAST [Fully Automated Storage Tiering]. We had a concept and then we proliferated that across our storage products.
Does that mean you'll be replacing SRDF and Invista with VPLEX? No. I think what you'll see is we'll be better leveraged with technologies like SRDF. For example, today we can create a VPLEX environment that does active-active [volume] synchronously and a volume being replicated off the back of a Symmetrix with SRDF asynchronous replication. Again, VPLEX is not just about Symmetrix, this is about anyone's storage products.
You said previously that you don't yet support other products. You said you were in talks. What we don't have in place is an agreement on management – so we're driving SMI-S's standards and leveraging management capability. EMC's been a leader in terms of SMI-S standards. We'll continue to push that and that's how we intend to leverage third-party management of devices. With VPLEX, our statement and direction is that we will drive management through SMI-S and leverage [the data] copy services of the array itself.
What are some of the management capabilities you'd be able to leverage through SMI-S on competitors' arrays? I'm talking about things like local replicas, whether mirrors, snapshots or clones, and remote replicas, depending on the customer's topologies. We can do [EMC] Recover Point integrations for continuous remote replication and we can leverage other vendors' synchronous and asynchronous replication.
Is VPLEX how you'll also consolidate management of Symmetrix and Clariion arrays? Today, Rich Napolitano [president of the Unified Storage Division] is talking about Unisphere. That is about our steps toward management convergence. Symmetrix has SMC [ Symmetrix Management Console ], which leverages the same user interface across all of our products. These also link into EMC Control Centre [management] software. You'll see more and more convergence, but the first step of that is being announced today. VPLEX will also leverage common management. Today it is a separate user interface, but tomorrow you can envision that as being integrated.
You refer to distributed caching a lot. What do you mean by that? Typically, caching has been localised. Most cache algorithms operate on a physical memory or set of memories that are localised. We've built our systems out of cache architectures. Symmetrix is a cache-based architecture. Clariion is a cache-based architecture. And, our competitors have built some products on cache-based architectures. What VPLEX does is distribute the caching function.
You can have a cluster linked to a cluster linked across synchronous distances. That is what we call VPLEX Metro. But, in reality, it is just one cluster. It is sharing cache among those members within a cluster. If they share volumes, all members of that cluster can share the same information. If one member does a write to a segment of data in a volume, it claims ownership of that segment. Any of the other members sharing that volume will be notified of that write through a lightweight directory process.
How does it notify them? It notifies through simple messaging within the cluster or across the synchronous distance. It says, 'Hey, you guys are sharing the same volume. That segment of the volume, I've got it now. If you need it, let me know and I'll give it to you.' It knows who has access to the volumes and notifies them when updates are made to them.
Some users on the exhibition floor said the VPLEX seems really expensive. I mean, $77,000 for a base configuration of the VPLEX Local. Will EMC be considering a lower-end model? Right now, it's packaged in an appliance, but over time it will be consumed differently, whether it's in a lower-end appliance or a virtual appliance. But I will say we priced this aggressively for the VPLEX Local. We believe we'll be able to compete on price very aggressively out there in the market. We also have a subscription model. So you can get into it thorough a subscription model [starting at a very low price].