EMC has released VPlex as part of its private cloud vision. This technology takes a fresh look at data center management, access and mobility -- i.e., it enables organizations to move thousands of virtual machines (VMs) and associated storage over thousands of miles. This also facilitates the shifting of IT operations away from a potential disaster area to a backup site at short notice.
"This distributed storage federation solution eliminates the boundaries of physical storage and allows information resources to be transparently pooled and shared over distance for new levels of efficiency, control and choice," said Brian Gallagher, vice president of storage at EMC. "Federation of storage arrays is key to virtual storage."
Initially, two products are available: VPlex Local and VPlex Metro. The local version moves block-level data between EMC and non-EMC storage within one site. According to Gallagher, it will support any block devices on the back end. That includes NetApp storage that uses block-level storage, but not file-based storage. Over time, that may change.
The primary building block of this data center management technology is a VPlex Director that supports up to 8,000 virtualized storage volumes. VPlex Directors sit between x86 servers and storage arrays. They contain quad-core Intel Xeon processors, 32 GB of memory and 8 Gb/sec Fibre Channel (FC) host and array connections. Pricing starts at $77,000 for a single engine with two processor boards. Two quad-core processors are on each board. Alternatively, the software is available for a subscription price of $26,000.
Gallagher talks about VPlex being "array aware." What he means is that it can leverage the capabilities already built into the storage array, such as replication, complementary caching and automation (working in conjunction with EMC Fully Automated Storage Tiering or FAST).
For those that must move data rapidly to an external facility, VPlex Metro is the better choice. It provides the ability to link two separate VPlex clusters within a data center or up to 100 km apart and to federate some or all of the storage volumes across both clusters. At that distance, the round-trip response time will be about 5ms or less. System administrators can configure a volume for simultaneous access by applications in both locations. In addition, VMware VMotion and vSphere are integrated into this product to facilitate long-distance live migrations.
As VPlex Metro has double the possible engines of VPlex Local, it can deal with up to 16,000 virtualized data volumes. Organizations running Local will require separate licensing if they decide to move up to Metro status.
"VPlex Local is fine for those with a single server farm," said Gallagher. "VPlex Metro enables you to create virtual server farms."
In the next year or two, EMC plans to introduce VPlex Geo and Global. Once you pass the 100 km barrier, synchronous transfer suffers from too much latency. Hence, the company is developing VPlex Geo to enable asynchronous federation of VPlex clusters. Once available, it will be useful when it comes to data center migration, site consolidation and application relocation over large distances.
"The VPlex protocols minimize latency," said Gallagher. "But the application response time has too much latency beyond 100 km or so."
VPlex Global, of course, will eventually take things to another level. The idea behind it is to provide concurrent data access and workload relocations across multiple locations spread around the planet. It will use both synchronous and asynchronous technology to achieve this.
"One of the challenges about leveraging IT resources around the world is that you had to forklift the data from one geographic area to another," said Benjamin Woo, an analyst at IDC. "What VPlex enables you to do is access this information anywhere."
Although EMC has thrown a great deal of its annual $1.7 billion of R&D spend at VPlex, it isn't in a hurry to recoup its investment. Pat Gelsinger, president and COO of EMC's Information Infrastructure Products Division, noted that the company expects a slow roll out with revenue ramping up starting in 2011.
That said, the company was willing to demo its product at EMC World earlier this month. It set up a VM migration from its Hopkinton, Mass. headquarters to Boston, where the show was being held. During Gelsinger's half-hour presentation, he moved 500 VMs to the other site and had them online at the Boston location. Trying to do that by backup-and restore would take many hours, and in some cases, days.
EMC VPlex Line Up
|VPLEX Local||The local version moves block-level data between EMC and non-EMC storage within one site||Features a single cluster with support for up to 8,000 virtualized storage volumes||Starting at $77,000 for one engine|
|VPLEX Metro||Non-disruptive, transparent data mobility between EMC and non-EMC platforms over a distance up to 100 km||Features two clusters with up to 4 VPlex Engines each and support for up to 16,000 virtualized data volumes||Starting at $77,000 for one engine; minimum of two engines needed|
|VPLEX Geo||Non-disruptive, transparent data mobility between EMC and non-EMC platforms over continental distances||Will enable asynchronous federation of VPLEX clusters to support requirements, such as data center migrations/consolidation and application relocation over cross-continental distances||To be announced|
|VPLEX Global||Non-disruptive, transparent global mobility between EMC and non-EMC platforms||Will enable distributed concurrent data access and workload relocations across multiple global locations over both synchronous and asynchronous distances||To be announced|
Drew Robb is a freelance writer specializing in technology and engineering. Currently living in California, he is originally from Scotland, where he received a degree in geology and geography from the University of Strathclyde. He is the author of Server Disk Management in a Windows Environment (CRC Press).