NEC has announced the details of its long-awaited grid storage
system, HydraStor, which should ruffle some feathers in the storage
industry. (See
NEC preps grid-based backup appliance.)
HydraStor is in beta and won't ship until the end of the summer,
but storage resellers and analysts said there's plenty to look
forward to. An entry-level system supports 7.5 terabytes (TB) of
usable capacity or about 146 TB of effective capacity when data
deduplication is turned on. The system can theoretically scale out
to thousands of petabytes (PB) of capacity in 2.5 TB increments,
according to NEC. A 140 TB system will cost approximately $100,000
with security, continuous data protection (CDP), replication and
migration features standard in the first release.
"It'll be the hallway discussion among all the storage players this
week," said Robert Gray, vice president of worldwide storage
systems research at IDC.
Most "grid" storage is actually clustering on steroids. It still
has a single control node for management, whereas with HydraStor
everything is distributed across all nodes so there is no
bottleneck, explained Karen Dutch, NEC's general manager for
advanced storage products. She said HydraStor, which is a file-only
system for now, will offer 100 megabytes per second (MBps)
throughput and supports an NEC patent pending process called
Distributed Resilient Data (DRD). Most clustering storage products
use a combination of RAID and
RAIN techniques for mirroring different
parts of the disk across two nodes so that data will survive
disk and node failures. However, Dutch points out that mirroring
would require a considerable amount of excess capacity. "It gets
extremely complex and ugly as you scale, and very expensive,"
she said.
HydraStor's DRD feature, meanwhile, has a default setting of
Parity-3 protection, protecting against three disk failures, but
users can effectively "dial up" as many parity chunks as they
require. Data is distributed across all nodes using DRD so an
application never performs a failover and there are no redundant
copies. More significantly, NEC said there is little to no impact
on performance during rebuild times, unlike RAID-6 rebuilds, which
regularly bring storage controllers to their knees. Nodes can have
different capacity and performance capabilities and can be added,
removed and upgraded nondisruptively, NEC said. In addition, manual
tasks, such as provisioning and reserving capacity, are not
required. As nodes are added, capacity is discovered and utilized
automatically, and existing stored data is automatically load
balanced across them.
Another feature likely to cause a stir in the industry is
HydraStor's deduplication technology, dubbed DataRedux, which
eliminates data duplication at the subfile level across and within
incoming data streams. NEC claims it reduces storage capacity by up
to 75%, without interrupting performance.
Jim Addlesberger, president and CEO of NavigateStorage LLC, a
reseller, said this is a crucial differentiator between HydraStor
and existing deduplication products. "While dedupe is a superb
idea, all the early implementations have had various shortcomings …
they couldn't examine enough chunks fast enough to scale, and the
manageability was horrible," he said. "As you scale, you buy
another and another, and soon you are managing a farm of them."
With HydraStor's grid architecture, controllers are added as
capacity is added and every node is aware of every other node,
easing performance and management issues.
Addlesberger said the only drawback to HydraStor is that it's a
new product, "and we'll have to see where the issues lie."
Similarly, Gray said it has yet to be proven, but he noted that
historically, NEC is like "the G.E. of Japan … very conservative
and with a long and proud tradition of engineering." Though like
IBM, he said that has meant a tough time being competitive
globally. "When you are so detailed and complete, being speedy is
difficult," he said.
Eventually Gray said he'd like to see NEC offer its APIs and
node logic into a standards body so that users can add nodes from
different suppliers. "Long term, we need to be able to build
heterogeneous grids, although that's a long way out," he said.
For now, HydraStor is only addressing the secondary storage
market, which Marc Staimer, president of Dragon Slayer Consulting,
describes as a "flanking maneuver by NEC … they are coming around
the side instead of hitting the enemy head on." In 2008, NEC
expects to change this up, rolling out security and classification
features aimed at the primary storage market, and by 2009,
HydraStor will support multisite implementations.