When NEC first briefed me on its Hydrastor product last year I loved the idea of a deduping backup target that used the RAIN (redundant array of independent nodes) architecture based on standard Xeon servers. Now NEC's releasing new storage and accelerator nodes that boost capacity to 12 TB (raw) on each storage node and the data ingestion rate to 300 MBs for each acceleration node.A Hydrastor grid's accelerator nodes manage the cluster's file system, allowing backup and archiving applications access via a CIFS or NFS / NAS like interface. The new accelerator nodes are available with 2 10-GB Ethernet ports eliminating the 1-GB bottleneck that's held NAS backup appliances behind VTLs in backup performance. The accelerator nodes chunk and hash incoming data and pass each chunk to the storage node that's responsible for that portion of the hash universe to compress and store.

Because any chunk of data is stored just once in a deduplicated data store, it's especially important for a deduplicating storage system to provide sufficient fault tolerance to prevent hardware failures from causing data loss. Rather than use a RAID controller in each storage node, Hydrastor storage nodes break each data chunk into data and ECC slices and then distribute the slices across disk drives and storage nodes. Users can set the data protection level to survive anywhere from 1 to 6 disk drives. The default level of 3 has about the same overhead as RAID-5 but can survive 3 drive failures or the failure of up to a quarter of the storage nodes in the cluster without data loss.

These new nodes demonstrate the true strength of a grid system like Hydrastor. Any of NEC's 50 U.S. customers for the original Hydrastor can add new storage nodes that hold four times the data in the same 2U space and just join them to the existing cluster. They also can eject the old nodes and, when the system is finished migrating the data to the other nodes in the grid, return them to the leasing company without disrupting their backup process.

The grid architecture also means that the Hydrastor can scale from one accelerator node and two storage nodes to speeds and capacities big enough for most any organization. Even better, users can grow their Hydrastor grids in capacity and performance independently, adding accelerator nodes for performance and storage nodes for capacity. NEC's list prices are $70,000 for a storage node and $50,000 for an accelerator node.

With the new nodes, NEC also announced a set of preconfigured systems that include not only the Hydrastor nodes but also the racks, Ethernet switches, and interconnect cables to make ordering and installing a Hydrastor easier, ranging from a 1-accelerator node - 2-storage node system at $180,000 to a 110-TB 16.5-GB behemoth that is like Commodore Vanderbilt's yacht -- "if you have to ask how much it costs you can't afford it."

That behemoth also sets a record speed for inline data deduplication. In fact, it's faster than Sepaton and EMC's biggest VTLs, even when they're not deduping.