Much as I always respected the engineering of NEC's products in the past, the one thing you wouldn't call them is exciting. The short form of the review of the S2500 disk array I wrote a couple of years ago in Network Computing would have read "Solid modular disk array has all the features you'd expect, but mothering unique. They sell a lot of them in the home market, so don't worry about being a guinea pig." Well, with HYDRAstor, it has something unique.

The eggheads at NEC's Princeton, N.J., lab (yes, the same town, not the same lab as Einstein) set out to create a new, general purpose storage platform. They put together a grid of 2u servers, each acting as either a front-end accelerator node or a 2.5-TB back-end storage node, and baked data de-duplication right into its DNA. With gigabit, or 10 Gb, Ethernet connecting the nodes together and a CIFS or NFS storage interface, customers can combine arbitrary combinations of nodes, adding more accelerator nodes for faster backups or more storage nodes as your archives grow. NEC's tested systems with up to 450 storage nodes with a capacity of well over a terabyte before de-duplication. It claims it could ingest and de-dupe data at a blistering 14 GBps.

As if that weren't enough, it threw out traditional RAID as too 20th-century and instead spread the de-duplicated data chunks across multiple disk drives, and storage nodes, with parity chunks to protect against drive failures. It even made how many parity chunks the system stored for each set of data chunks variable so you can make the disk-space-to-data-loss decision yourself. The default setting stores nine data chunks and three parity chunks for each data stripe and can survive three disk failures without data loss with only 25% overhead. It even distributes chunks across storage nodes, so on a HYDRAstor grid with more than 12 storage nodes, up to three storage nodes could fail without data loss.

NEC's marketing folks wisely choose to target HYDRAstor at the backup and archive secondary storage market where data de-duplication is gaining acceptance. Where the other three-letter vendors would tell you to buy a VTL for backup and a CAS system for the archive, with HYDRAstor you could just add more nodes. On the downside, you'll need to update your backup software to the new version that supports NAS for backups intelligently and re-schedule all your backup jobs, where a VTL would have been a drop in.

Setup looks EqualLogic simple, with NEC claiming a world speed record of 42 minutes from box to storage. All you need to do is connect the cables, assign IP addresses, give it a name, and create a file system with a data resiliency level.

The HYDRAstor is missing a few important features, such as asynchronous replication for disaster recovery and, more significantly, WORM functionality for archive applications. With courts not blindly accepting electronic records as authentic, I think WORM will spread from finance and pharma to become SOP for corporate archives.

At a cost of about $1 per gigabyte, the HYDRAstor's an impressive piece of kit.