EMC Corp. today unveiled V-Max, a new modular version of its Symmetrix disk arrays with software-based performance improvements. The vendor also plans to add automated storage tiering software to the new system later this year.
The new Symmetrix V-Max follows the DMX-4 in the Symmetrix family, although the DMX-4 will remain on the market. V-Max has a virtual matrix or mesh architecture that connects multiple modular disk arrays, front-end and back-end director servers, and chunks of global memory to a high-speed interconnect similar to a switch. The interconnect protocol can be FICON, Fibre Channel (FC) or iSCSI. By contrast, the DMX-4 uses a direct-matrix architecture in which directors within the chassis are hard-wired directly to a backplane inside the frame.
As with the DMX-4, V-Max can flexibly move data around while hosts are connected using virtual logical unit numbers (LUNs). But the Symmetrix Enginuity OS sees storage resources differently in V-Max. It views and manages hardware assets, including ports and storage devices, in groups rather than as individual elements. V-Max includes new software wizards to make provisioning simpler with this release.
"This is the most significant redesign of [the Symmetrix] software OS ever," said Bob Wambaugh, EMC's senior director of product marketing. "When we did thin provisioning last year, we added a layer of abstraction that allows us to do many things in parallel underneath. It parallels what VMware did in the server world, where pools of physical resources are treated and acted upon concurrently."
With PowerPath integration into VMware, the system can also automatically load balance virtual server workloads within the array. "It kind of ends that whole I/O problem with virtual servers because PowerPath will automatically load balance across the system," said Steve Duplessie, founder and senior analyst at Milford, Mass.-based Enterprise Strategy Group (ESG).
'This is about software, not hardware'
With new multicore Intel Corp. processors, parallelized hardware and updated software algorithms, EMC claims up to three times the performance for V-Max over DMX-4. While DMX-4 can move virtual LUNs around individually without taking down the system, V-Max can theoretically perform thousands of these operations concurrently, EMC's Wambaugh said. However, he declined to disclose any specific benchmark performance numbers.
The use of x86 processors rather than specialized silicon is "brand new for the high-end space," wrote Benjamin Woo, vice president of enterprise storage systems research at Framingham, Mass.-based IDC, in an email to SearchStorage.com. "There could potentially be some Symmetrix purists who will find this not to their liking. Can't really help these guys. When it's all said and done, this is about software, not hardware."
One of those self-described purists is Tom Becchetti, an EMC and NetApp customer working with a Fortune 200 company that he requested not be identified because of corporate policy forbidding him to use its name publicly. "The performance upticks are interesting, but you might need them to manage all the extra code," he said. "The devil's in the details with these things—how much code overhead have they put onto the array for automation?" Becchetti said he generally believes that the place to automate IT processes is at the host level rather than within storage systems.
For now, there's a difference between what's available and theoretical in V-Max. While V-Max can theoretically scale up to 256 nodes, the first system available currently has a limit of eight nodes, or engines, as EMC calls them. A single engine can scale from 48 to 360 drives, and pricing starts at $250,000 with the minimum number of drives and minimal software licensing. Additional software licenses for virtual LUNs and other features must be purchased as capacity expands. At that price point, "it's not exactly going to put EVA or Clariion out of business," ESG's Duplessie said.
However, IDC's Woo said, "given the flexibility and modular nature in which the new Symm can be put together, it could [also] potentially cannibalize some very high-end Clariion sales."
V-Max can scale to 2 PB, with 128 host ports for mainframe and open systems, as well as 128 back-end connections for FC, SATA and Flash drives.
Automated tiered storage planned
Later this year, EMC will add automated tiered storage provisioning at the LUN level to the array software, the firm's Wambaugh said. When that happens, data can be moved automatically between tiers of storage, including Flash drives, according to frequency of access. EMC calls this FAST -- fully automated storage tiering.
If that sounds familiar, it's because there's a vendor with a similar capability already shipping—Compellent Technologies Inc. Atrato Inc. also announced similar capabilities this week and expects to make them available next month. NetApp and Xiotech Corp. have bulk-provisioning wizards for VMware hosts, while 3PAR offers the ability to cluster up to eight controllers and uses what it calls a Mesh-Active architecture.
Except for 3PAR, most of the other products delivering new features similar to those found in V-Max are focused on the midmarket, without the performance or scalability of a high-end disk array. "It depends on your requirements—if you plan to scale out with those features, you need to seriously consider a different architecture," said Tom Trainer, founder and president at Analytico Inc.
"Compellent's array eventually ends—this can do automated tiered storage across multiple arrays," ESG's Duplessie added. Tiered storage automation is also considered increasingly important to drive the adoption of solid-state drives (SSDs) throughout the market, according to industry experts SearchStorage.com recently interviewed.
Brian Garrett, technical director at the ESG Lab, said the real battle in high-end disk systems currently lies between the new Symmetrix and the Hitachi Data Systems USP-V, which introduced features like thin provisioning and storage virtualization behind the controller before DMX did.
"There's an arms race between these two in terms of performance and scalability," he said. "But it's hard to tell without [specific performance] data available who's winning on that front right now."