Hitachi Struts Mr. Universal

NEW YORK Hitachi Data Systems (HDS) ended the speculation around its next-generation storage system today by launching the Universal Storage Platform (USP) -- presumably named because it offers something for anybody who needs the highest-end enterprise storage (see Hitachi Preps Lightning Strike).

For those that want performance, Hitachi claims its system’s architecture delivers 68 GBytes per second of cached bandwidth, 2 million IOPs (input/output operations per second), and 256 concurrent memory operations; and supports 332 TBytes of capacity. For those that want functionality, Hitachi announced new virtualization, partitioning, and replication software.

HDS offered specifics on those features for the first time at a news conference today. While the performance numbers clearly give it an advantage over competitive systems from EMC Corp. (NYSE: EMC) and IBM Corp. (NYSE: IBM), HDS’s doubters wonder whether the new features have broad enough appeal to win substantial market share.

The system -- dubbed Hitachi TagmaStore Universal Storage Platform -- is not officially part of its previous high-end Lightning family. Hitachi will continue to sell Lightning and midrange Thunder systems, and they can work alongside the new controllers.

The new systems come in three models:

All models support Ficon, Escon, and NAS connectivity. The systems are available immediately, and pricing begins at around $700,000.

HDS executives say the software enhancements are the key parts of the new offering, although they will be the last in place. Virtualization and replication functions won’t be available before December, and they involve connecting to storage arrays from EMC, IBM, and Hitachi’s Lighting and Thunder platforms.

HDS will extend the virtualization capabilities it already has across Lightning and Thunder platforms, allowing customers to pool storage that resides on hardware from competitors. IBM claims it will have the same capability with its SAN Volume Controller by the end of the year, but the two go about it differently (see IBM Revs Virtualization Engine and IBM Sets Virtualization Ship Date).

IBM uses an appliance running software as its virtualization engine. EMC is planning yet another type of virtualization, driven by placing the software on intelligent switches (see EMC & IBM in Virtual Skirmish, EMC & McData Get Smart, EMC OEMs Brocade's Rhapsody Switch, and EMC, Cisco Do the Deed). HDS performs virtualization by mapping LUNs into cache and accessing it from there. In essence, HDS initiates virtualization inside the controller.

“Competitors are trying to put virtualization in the wrong place,” HDS CTO Hubert Yoshida says. “By putting it into the network, they’re introducing a new level of complexity between the network and storage. We’re putting it behind the network and storage. The heavy lifting should be done at the edge of the network and not in the core.”Yoshida says IBM’s in-band virtualization has limits because it uses network resources and doesn’t scale. As for the switch approach, he says that adds latency because it involves extra read and write operations. Yoshida says Hitachi has no plans to embrace intelligent switches from Brocade Communications Systems Inc. (Nasdaq: BRCD), Cisco Systems Inc. (Nasdaq: CSCO), and McData Corp. (Nasdaq: MCDTA), but he says their multiprotocol routers would be used with the Universal Storage Platform.

Hitachi also uses a different approach for long-distance replication. Instead of placing upgraded data in cache, its Universal Replicator places updates into something it calls a disk journal that contains the data and metadata. For partitioning, HDS software allocates internal and external physical storage resources -- including ports, cache, and disk -- into independently managed Private Virtual Storage Machines. Users can modify any partition according to the applications it holds.

Financial analyst Shebly Seyrafi of Merrill Lynch & Co. Inc.

says hardware improvements will be key to gaining market share although software improvements are an important part of the new platform. He says the partitioning and replication capabilities could hurt EMC, “because EMC gets a lot of money from its replication products.” As for virtualization, Seyrafi says Hitachi needs to validate its solution in the marketplace.

“They’re going to have a tough time succeeding with virtualization, because I don’t think customers are ready for it,” Seyrafi says. “Virtualization has been out a while, and we haven’t seen a vendor succeed with it. I don’t know if customers want to see Hitachi become the focal point of virtualization. And customers don’t want to mix and match EMC and Hitachi arrays.”

If the software is so important, why did HDS significantly upgrade its hardware performance? To keep up with the software, says Claus Mikkelsen, HDS senior director of storage applications. “We really need that bandwidth,” he says.By putting the emphasis on the software, Hitachi is trying to change the way it is perceived. HDS has a reputation for being long on performance and short on service and software (see HDS's Hard-Wired Vision). Now, HDS president Dave Roberson says Hitachi has already hired hundreds of people in the services and software departments, with plans to hire hundreds more. Hitachi brass muttered Information Lifecycle Management (ILM) -- a phrase EMC acts as if it invented -- a few times during today’s news conference and talked about “solutions selling.” That translates into services, of course.

“We see the new platform as a real driver of software and services growth,” Roberson says. “This is something that needs to be worked on with our customers, rather than say, ‘Here it is, figure it out.’ ”

— Dave Raffo, Senior Editor, Byte and Switch