| Clustered storage systems, traditionally used by high-performance computing (HPC) shops that need performance and capacity far beyond typical enterprise requirements, have been gaining traction among enterprises over the last two years thanks to data growth. The advent of Web 2.0 apps and the rise of online service providers over the last year has further blurred the line between HPC compute farms and enterprise IT, and pushed demand for massive, easily scalable storage systems.
Until now, startups such as Isilon Systems have been riding this wave, but bigger storage vendors are getting ready to weigh in. Analysts say new products from Hewlett-Packard (HP), IBM and EMC, all slated for 2008 release, will be the catalyst for a major move toward clustered storage in the enterprise over the next year.
HP and IBM are both dipping into the clustering market through acquisitions. HP is planning to couple clustered file-system (CFS) software it acquired with PolyServe last February with its own server hardware to create clustered nodes that combine processing power and storage. In January, IBM acquired Israel-based XIV, a startup founded by former EMC chief engineer Moshe Yanai. Unlike the majority of clustered storage products, XIV's Nextra system is block based rather than file based. IBM is positioning the product as a complement to its monolithic disk arrays at the high end for Web 2.0 apps.
Perhaps the most anticipated product has yet to be previewed in detail: EMC's Web 2.0 clustered storage system, first revealed by president and CEO Joe Tucci at a press and analyst event last November. The release of the product--hardware codenamed Hulk and software codenamed Maui--will be a tipping point for adoption of clustered storage, according to Robin Harris, senior analyst at Data Mobility Group, Nashua, NH.
"I compare it to IBM's announcement of a PC in the early '80s," says Harris. "IBM wasn't the first to offer a PC, but its blessing made the PC respectable for business." When the PC arrived, the workstation market was dominated by makers of minicomputers, the most famous being Digital. Minicomputers were proprietary, expensive and vertically integrated with apps by vendors, much like today's storage subsystems, says Harris.
Just as the PC introduced a low-cost, industry-standard workstation and the concept of a standardized OS, Harris predicts clustered NAS products built on lower cost, industry-standard components will bring about a similar paradigm shift in enterprise storage.
"It won't happen overnight, just like the PC didn't take over right away," he says. "It's a fundamentally difficult thing for businesses to move from 60% [profit] margins--which is where the proprietary storage subsystem vendors are today--to around 30%, which is where clustered systems are trying to bring them."
Another challenge is that developing clustered products has already proven easier said than done for some established vendors, including Network Appliance (NetApp). Just over four years after its acquisition of clustered NAS intellectual property from Spinnaker Networks for $300 million, NetApp is still working to integrate enterprise data management features like SnapMirror with its clustered Data Ontap GX product.
So far, GX has garnered the most interest from users with specialized environments, such as Vancouver, BC-based special effects studio Image Engine Design. According to COO Peter Muzzers, the ability to scale the system quickly, rather than full software feature support, was his company's primary concern when it deployed the system.
"The one feature we're still waiting for is better support for SNMP for reporting and monitoring," says Muzzers. "There seems to be some provisions for it in the code--they're obviously trying."
"GX doesn't have all the data replication capabilities we'd like it to have yet," admits Jay Kidd, NetApp's chief marketing officer. "But it's a matter of how to prioritize our development efforts." Kidd says NetApp's R&D focus has been on developing features for its existing monolithic systems to broaden their appeal, but it does plan to offer a single OS based on the best of both worlds.
For now, that's still a ways off. "We've found that the clustered storage market is still typically limited to single homogeneous workloads," says Kidd. "It's an important market, but it's not a mainstream market yet."