Sepaton Adds De-Dupe to VTL

Sepaton is joining the list of virtual tape library (VTL) vendors with data de-duplication, a feature that's rapidly becoming a must-have for any company that sells virtual tape solutions.

Sepaton will add DeltaStor de-duplication software as part of a product upgrade this week. Sepaton already offers DeltaStor as an add-on application for its existing VTLs or with new appliances.

Sepaton claims compression ratios of up to 50:1 for its new enterprise VTL and up to 25:1 for its SMB and legacy VTLs.

That said, we know customers don't always get the performance that vendors claim. (See Dealing With De-Dupe Doubts and Don't Get Duped.)

Still, Jeff Machols, systems integration manager at retirement and health benefits firm CitiStreet, says he got a 56:1 compression ratio while beta testing DeltaStore. Machols, a Sepaton customer since 2004, says he tested a combination of applications and Linux file systems and took backups ranging from 100 Gbytes to 120 Gbytes down to around 2 Gbytes. He says the mix made up about 80 percent of the types of data CitiStreet backs up."It's probably a little better than what I expected," Machols says. "The promises are all over the place from vendors on de-duplication. I would've been thrilled with 25:1, and we got twice that."

Machols plans to use DeltaStore on Sepaton VTLs in CitiStreet's Quincy, Mass., and Jacksonville, Fla., data centers. He won't upgrade his VTL hardware now -- his goal for using de-duplication is to stave off adding hardware by reducing capacity.

"Each library has about 50 Tbytes capacity now," he says. "As data grows and back up grows, well add incremental space, but de-dupe will eliminate that need for a long time."

By now, nearly all VTL vendors have data de-duplication or are working on adding it. (See De-Dupers Lining Up.) DataDomain, Diligent Technologies, and Quantum beat Sepaton to market. FalconStor also has a de-duplication application for its VTL software, although the de-dupe feature is in limited release.

Data de-duplication, which eliminates redundant files from the backup to save space and increase backup times, was a hot technology of 2006, and VTL is among its most popular applications. (See Insider: De-Dupe Demystified.)"If you're going to play in disk-based backup, you're going to need to have de-duplication," analyst Greg Schulz of the Storage IO Group says.

It's the larger storage vendors who still lack de-duplication in their VTLs. EMC, Hewlett-Packard IBM, and Sun use FalconStor's software for VTL but have not yet qualified the de-duplication software. Network Appliance does hardware-based compression for its VTL, but industry sources say NetApp is working on de-duplication. (See NTAP Disses VTL De-Dupe.)

The methods of performing de-duplication vary, however. Sepaton does its de-duplication outside the data path after data is written to the target disk. DeltaStor program manager Dan Chartier maintains doing it that way doesn't slow the backup down. "We don't do anything inline," he says. "De-duplication runs after the backup is complete."

There are several alternatives. Backup products such as Asigra's Televaulting, EMCs Avamar Server, and Symantec's PureDisk do the compression on the server before the backup. (See Symantec Dips Into De-Dupe and EMC Picks Up Avamar.) This method reduces the amount of data backed up, so less data has to be moved. However, the method takes up server CPU cycles.

Data Domain, Diligent Technologies, and Quantum DXi VTL do the compression on the fly while the data is being sent to the target. (See Data Domain Adds VTL Option and Quantum Leaps Into De-Duplication.) FalconStore Single Instance Repository (SIR) and Sepaton do it on the target after the data is backed up. (See FalconStor Extends VTL.)Schulz says doing the compression on the target after data arrives -- as Sepaton does -- avoids a bottleneck. However, it requires more processing power than the inline or on-the fly approaches. That's one reason Sepaton is also upgrading its hardware.

"You have to do another I/O step [to compress at the target]," Schulz says. "After data lands, you have to go back and reprocess it again. You need more processing power in your cluster to handle that operation."

Sepaton is adding 4-Gbit/s Fibre Channel connectivity and active failover to its enterprise VTL while extending the capacity from 42 Tbytes to 1 Pbyte. Sepaton also added an SMB appliance that scales to 7 Tbytes.

The enterprise drive is priced at $59,000 for 7 TBytes without DeltaStor,and $190,000 for 7 TBytes with DeltaStor. The SMB drive starts at $18,000 for 3.5 TBytes without de-duplication and $30,000 with de-duplication. DeltaStor licenses for existing VTL customers cost about $10,000 for 10 Tbytes of capacity.

— Dave Raffo, News Editor, Byte and Switch