WHEN IBM and StorageTek (now owned by Sun) first introduced virtual tape libraries (VTLs) in the late 1990s, the...
problem they were trying to solve was different from what VTLs are addressing today. VTLs were initially designed to improve tape cartridge utilization, says Dianne McAdam, senior analyst and partner at Data Mobility Group, Nashua, NH, but the technology has evolved into a way to improve backup times and speed restores.
To that end, VTL users are keeping more backup data online, rather than spooling it off to tape. Instead of a few days of backups, users routinely hold weeks, if not months, of backups on their VTLs, says Don Mead, VP of marketing at storage software vendor FalconStor Software. One customer has more than a petabyte of backup data stored on a VTL, he says. "When we first introduced a VTL, we never imagined the amount of data users would put behind it."
But no matter how cheap it gets, disk still isn't nearly as cheap as tape, and some users are bumping into the hard financial reality of keeping weeks' worth of backups online. "One of a VTL's shortcomings is that it uses a lot of disk space," says Brian Biles, co-founder and VP of marketing at Data Domain, which makes disk-based data-protection appliances and gateways.
One way a VTL can soften the financial blow of storing backups online is by compressing the data. Some VTLs like FalconStor's do software-based compression and, last month, Neartek announced that its recently released Virtual Storage Engine 3.0 can now be equipped with a PCI-based gzip hardware compression board for up to 2:1 compression without degrading performance.
Longer term, however, VTLs will probably adopt a more aggressive compression technique--data reduction. Data reduction keeps track of the blocks the VTL has seen and, rather than storing a block it has seen previously, it only stores a pointer to it. In the data-protection space, Data Domain and Avamar both use data reduction in their data-protection platforms; last summer, EMC VTL spin-off Diligent Technologies announced its ProtectTier platform with data de-duplication technology that can reportedly provide up to 25 times data reduction.
Last month, Data Domain announced that it would join the VTL fray and add a VTL/Fibre Channel interface to its platform, rather than the Ethernet-accessible file-system interface it has used thus far. "There are some cases where Fibre Channel is the default [in the data center] or where a tape library is the only thing that they understand," says the firm's Biles. "VTL is just an interface," he notes, so by adding a VTL interface, "we're just saying it's a non-issue."
Mead says FalconStor will also introduce data-reduction technology this year. That's significant because FalconStor's VTL software is at the core of numerous VTL offerings, including 3PAR's new Utility VTL, Copan Systems' Revolution family, EMC's Clariion Disk Library and Sun's Virtual Tape Library. It's also thought to be behind IBM's Virtualization Engine TS7510.
But Data Mobility Group's McAdam warns users not to let their expectations for data reduction get too high--at least not at first. One user McAdam knows reports getting 17:1 compression, but it took him six weeks to see it. "You don't get it right off the bat," says McAdam, adding that the data-reduction software has to start "seeing the repetitive patterns."