CommVault Adds De-Dupe to Simpana

11:30 AM -- I chided Dell and CommVault last year for pitching Dell's DL200 backup server (driven by CommVault's Simpana backup, formerly known as Galaxy) as bringing data de-duplication to the masses when it really did file-level, single-instance storage. Now, CommVault's new Simpana 8 suite leapfrogs the major players by integrating block-level data de-duplication into the core data management agent behind the backup and archiving functions of the Simpana suite. As if that weren't enough, they've also extended the de-dupe functionality to tape.

Before sending data to a media server, the Simpana 8 backup agent also performs the blocking and hash calculations and sends the hash values along with the data. The media server then identifies duplicate blocks and stores the data on any disk resource available to Simpana -- DAS, SAN, or NAS.

This approach should use somewhat less host CPU cycles than Avamar or PureDisk, which conduct a more complex conversation with the data store server(s) to identify unique blocks before sending them but will send more data over the net. Simpana will globally de-dupe to minimize network traffic between media servers so remote offices with local backup pools will use less network bandwidth than those without.

When a job spools datasets off to tape, it copies the blocks that contain data from any of the files or other objects in the dataset and creates a new hash catalogue that it writes to the tape. The tape can then be read by any Simpana 8 media server, but restores will require some cache disk space.

Simpana is content aware, seeing all the file system metadata as backups or archive jobs run and taking that into account when dividing the data up into blocks to de-dupe.While EMC customers with Networker and Avamar have a common server agent and can use Networker to schedule and manage Avamar backups, they still need to set up a separate Avamar data store for de-duped backups. Similarly, Symantec's started integrating PureDisk with the core NetBackup product, but still requires servers running PureDisk to serve as the de-duped data store.

These integration issues point out how Symantec and EMC built their software portfolios through acquisition and then created a Frankenstein's "monstore" that backs up, de-dupes, archives, and replicates. But you can see where the pieces were stitched together. CommVault developed the Simpana components in house, so you can use a common de-duped data store for archives and backups without jumping through hoops.

CommVault is charging $3,000/TB for disk-only deduping and $5,000/TB if you want to de-dupe data as you spool it off to tape as well as when it's on disk. This makes de-duping competitive with de-duping appliances. A Quantum DXi7500 with 9 TB of useable disk has a street price of about $50,000. An EMC AX4, Nexsan SATABoy or Snap Sever 620 with 9 TB usable will set you back around $15,000; add in $27,000 for de-dupe option licenses from CommVault and the total cost is $42,000, or 15% less. Given that the price you actually pay will vary depending on volume, your negotiating skills and how closely you buy to the end of the quarter, thats the same ballpark.

Since Simpana does single-instance storage to disk as a standard feature, Simpana users won't see the 20:1 or 50:1 data reduction rates that the use of the term de-dupe may bring to greedy little IT managers' minds since single instance already eliminated much of the de-duplicate data. Beta tester Jason Hall of home furnishing retailer Rooms To Go reported disk space savings of 30 to 50 percent.

CommVault gets points from me for moving de-dupe into the core backup engine and being the first to extend block de-duplication to tape. Combined with its stellar integration between modules, Simpana 8 looks like a winner.I am a bit concerned about the price however. Spending $3,000/TB to get even 3:1 data reduction on storage that costs only $1,800/TB may not be worth the investment in today's market. Data center space and power savings are worth something, but I can get a Data Domain box for the same money…

Of course, compared to the Backup Exec and Arcserve these companies are probably using now, Simpana will use a lot less disk to store the same data, or allow longer retention with the same disk.

— Howard Marks is chief scientist at Networks Are Our Lives Inc., a Hoboken, N.J.-based consultancy where he's been beating storage network systems into submission and writing about it in computer magazines since 1987. He currently writes for InformationWeek, which is published by the same company as Byte and Switch.