Ditch the Tape
The IT group at Pittsford Central Schools in New York was up against a wall. It had maxed out its physical tape library and was facing the prospect of having to rip and replace it to ensure data would be properly backed up. But the cost and manpower necessary to handle a bigger tape library was out of the question for the school district.
“Our tape library was so old we would have had no choice but to purchase a new one,” says systems administrator Charles Profitt. “The only reason you'd stick with physical tape is if you had plans to rotate out tapes, and we definitely did not.”
Instead, Profitt chose to go in a different direction. He deployed a virtual tape library (VTL), a disk-based technology that emulates pools of storage as tape-based drives. Virtual tape libraries can be software- or appliance-based and are available from manufacturers such as EMC, Fujitsu, HP, IBM, NetApp, Overland Storage, Spectra Logic, Sun Microsystems and Tandberg Data. VTLs enable IT teams to manage storage as a single entity and run concurrent backups to speed backup and recovery. Rather than mounting and searching through physical tapes, VTLs offer faster and more reliable access to disk-based data.
“VTLs appeal to organizations that are used to dealing with tape, because they offer tape-based functions without having to swap and rotate tapes,” says Deni Connor, principal analyst at Storage Strategies Now. VTLs are also budget savers, she adds, because they eliminate the data center footprint of physical libraries as well as the cost of tapes.
School districts like the flexibility VTLs offer. “You don't have to worry about how much data you can fit on a cartridge. You can size the volumes to whatever virtual tape you want them to be,” says Lauren Whitehouse, senior analyst at the Enterprise Strategy Group. In addition, IT avoids the hassle of drive failures and corrupt backups. They also can recover files faster because they don't have to recall a tape, load it, and search through to find the point where the data resides.
Profitt agrees. He deployed an HP StorageWorks 6218 Virtual Library System VTL, which he uses to back up his file servers, SQL servers and application data. The VTL is also interoperable with his HP StorageWorks Data Protector backup and recovery management software, so he can use the same backup policies he created for his physical tape library.
SOURCE: Enterprise Strategy Group
Many IT leaders have found that they don't need their physical tape libraries anymore because the VTL can hold enough data, according to Connor. That's certainly true for Pittsford, which has placed the 4.1 terabyte VTL offsite as a continuity of operations solution for its onsite servers. IT can also use a VTL as secondary storage until data is offloaded to tape, or roll out multiple VTLs for site-to-site replication.
Some VTL users have found that newer features, such as data deduplication, have helped them boost capacity. Data deduplication is a process by which only changed data is stored and is available in two varieties: in-line and post-processing. Connor says the benefit of in-line, which is offered by companies such as IBM, is that you don't need extra storage to hold the data while it's checked for duplication. However, in-line processing can affect performance as bits are checked at the source in real time. With post-processing, which is offered by HP and others, all data is sent over and then examined at the storage target.
Profitt says he did need to buy enough storage to be able to compare two full backups, but says the speed is worth the trade-off.
Here are some tips to help you avoid the potential pitfalls of virtual tape libraries:
• If you use replication for continuity of operations, test how long it will take to recover from your offsite VTL.
• Carefully plan out your storage strategy so that if a part of your VTL fails, it does not take down all your critical data.
• Determine whether you are going to destroy old or failed disks yourself or let the vendor take care of them. If you are leaving it to the vendor, make sure your data is either erased or unreadable.
• Don't count on the vendors' estimated volume sizing. Instead, study your own applications and storage needs as well as potential overhead. For instance, there could be more overhead if you're creating lots of smaller cartridges rather than several large ones or if you're using data de-duplication.