VPLEX: Redefining the Boundaries of the Data Center

May 10, 2010

In March, Pat Gelsinger introduced EMC’s vision for Virtual Storage (my blog post is here). Today, EMC announced the VPLEX family of products—the first new products that deliver on the vision.  VPLEX is a Distributed Federation solution, which means that we can enable a single copy of data to be shared, accessed and relocated over distance.  When new technologies are introduced into the marketplace, people are most comfortable in making comparisons to things that they already know.  In this post, I’ll compare traditional replication solutions with VPLEX Distributed Federation.

Federation Instead of Separation

In a traditional replication solution, there is a primary site and a secondary site for failover or recovery of data.  In contrast, with VPLEX, the data is active at both sites at the same time.  Rather than putting together complex failover scenarios, having the flexibility to have active data between sites will allow for maintenance without downtime and workload balancing across multiple data sites.  Instead of thinking of “site A” and “site B” separated by policies and distance, we can start to think of our pool of federated resources that spans multiple physical data centers.  VPLEX can change how you think about deploying data centers; rather than a disaster recovery site, deploy data centers to achieve additional efficiencies (such as taking advantage of low cost energy) and performance (optimizing locations to be close to clients).

Near Today, Far Tomorrow

The network between the locations is critical for any solution to work.  Similar to replication solutions, we can consider whether a solution is synchronous or asynchronous and how the network fits into a customer’s existing environment.  Enterprise customers will typically have network connectivity between locations that are close to each other.  When these existing links have the appropriate bandwidth and other network characteristics–such as those required for replication solutions–the same links can also be used for VPLEX.  Synchronous solutions of 100km are available with VPLEX Metro today.  Longer distance (asynchronous) solutions require special links. Asynchronous solutions will be available with VPLEX Geo in 2011.

As I hinted at in my previous post, the VPLEX solution ties into other development work from partners.  VPLEX will expand the solutions of VMware VMotion (see this great VPLEX/VMotion whitepaper) and Microsoft Live Migration, plus work with the latest networking technologies from Cisco (including leveraging OTV) and Brocade.  Virtualization of your servers, network and storage redefines the boundaries of the Data Center by creating a single pool or resources separated by distance.


Click on the graphic above to go to the VPLEX launch site which includes customer videos and whitepapers.  You can also hear more about VPLEX live on May 10th with streaming internet TV from EMC World: Pat Gelsinger at will be on at 3pm Eastern and Brian Gallagher at 4pm Eastern.

Stuart Miniman



Virtual Storage, not just another V-word

March 11, 2010

In the realm of storage, I have spent most of my time outside of the storage box, especially how storage networking, servers and operating systems (OS) impact storage.  I began working with Cisco and VMware long before the VCE initiative.  Some of the broad industry trends that we have seen across the OS and network stacks are also applicable to the storage layer.  Pat Gelsinger, EMC’s President and Chief Operating Officer of Information Infrastructure Products, spoke today with a group of analysts to lay out a vision for Virtual Storage.  The audio and slides of Pat’s presentation can be seen here (may request your email address to view).

This strategy relates to Private Cloud. If you look at the diagram below, virtual storage lives in the virtual information infrastructure layer.

To understand the trends, let’s look at the OS and network layers of this stack.

Server Virtualization is much more than hypervisors

The first thing that most people think about when they look at server virtualization is how it can help consolidate resources, which leads to a reduction in hardware and therefore to savings in power and space.  When I think back to my first interactions with VMware in 2003 (when I was in EMC’s E-Lab), it was actually the abstraction/isolation of an environment that was so useful.  Old operating systems (Windows NT) or “odd” environments (FreeBSD) could be put in a VM and EMC’s storage could then support the configuration.  In addition to extending the life of an existing configuration, VMware is tremendously valuable for test and development environments.  Around the time that we were first working with VMware, they came out with a product that everyone thought was magic when they first saw it – VMotion:  Live Migration of a virtual machine.  The liberation gained with this feature turned a useful/interesting product into a game changer.  Individual server resources could be turned into a pool of resources.  With the addition of VMware Distributed Resource Scheduling (DRS), VMotion becomes automated.  As a final note here, take a look at some of the plans around extending VMotion over distance that were discussed at VMworld last year.

Convergence and Cloud Internetworking

As with server virtualization, the first step towards network virtualization is consolidation of resources.  With the adoption of 10Gb Ethernet and storage protocols such as Fibre Channel over Ethernet (FCoE), iSCSI and NAS, we reach a single network for LAN and SAN traffic: a building block of the private cloud.  In particular, FCoE allows existing infrastructures built on FC to move towards a unified environment.  Another enabler of cloud internetworking is Cisco’s Overlay Transport Virtualization (OTV) which they unveiled last month.  OTV allows for mobility of resources across datacenters or with a compatible service provider.  OTV leads networking down the same path of automation and mobility that we have taken with server virtualization.

Virtual Storage, not just another V-word

As my mainframe friends have told me, virtualization is not a new idea, and storage virtualization has been available from EMC and competitors for many years.  Similar to what is happening in the OS and networking stack, Virtual Storage goes beyond the basic functionality prevalent in the market today  by adding flexibility and automation.  The goal is to transform individual storage resources to a pool using Federation.  As defined on the Wikibon site, Federated Storage enables a “loosely coupled set of storage resource nodes to act unilaterally and still be managed centrally, organizations can create networks of virtually limitless capacities and eliminate disruptive migrations”.  While many vendors may be working on federation, Pat Gelsinger discussed EMC’s vision, in which an enabling Distributed Cache Coherence technology allows for Global Federation.   In a recent Forbes interview, Pat discussed the value of leveraging federation over distance:

“I want to flexibly move those across shared infrastructure, which can drive up the utilization dramatically. I also can get load balancing across data centers, do mainframe-like high-availability and failover. I also can build a cloud-like architecture that’s private and then federate from my private cloud to my external environment, where I can have spillover.”

Similar to what we’ve seen in the OS and networking stack, virtual storage will allow the use of both old and new resources, regardless of location – providing an onramp to adoption of a Private Cloud model.  As Chuck Hollis has said: “And maybe we can start thinking entirely differently about that age-old catch phrase of “IT delivered as a service”.”

Comments are always welcome.  You can be sure that we’ll be talking a lot more about this at EMC World, May 10-13 in Boston.  I will be presenting Converged Data Center: FCoE, iSCSI, and the Future of Storage Networking.  Any bloggers that are attending should sign up for the Blogger’s Lounge.  Hope to see you there.

Stuart Miniman