Wayne Tolliver, systems manager for university communications at OSU, backs up and archives approximately 60 TB of data on a Spectra T380 LTO-4 library to help control storage that doubles every two years on his disk array. The OSU communications office stores rich media data such as high-definition video and digital photography for university publications and Web sites.
"We just came to the realization that we cannot store everything on disk," Tolliver said. "It's not environmentally responsible, and it's not responsible from a cost perspective or from an infrastructure perspective to keep all these things spinning on disk. You can't offsite disk. EMC will be happy to sell you another disk array that's identical to the one you bought, and then sell you TimeFinder and Symmetrix Remote Data Facility [replication software], and then you have to call your networking company and have a very fast network link connecting them. That gets expensive. Not to mention, it's not green. So we've made our bet with tape."
Tolliver's shop runs an Apple-based storage area network (SAN), consisting of an XServe and two QLogic Corp. SANbox 5802V 4 Gbps Fibre Channel switches. He also uses Spectra's virtual tape library (VTL) option, which includes portable disk cartridges. But he considers many disk storage products overpriced and overhyped, especially for archiving.
"I think the disk vendors today do a tremendous disservice to the storage industry in general by trying to convince everyone that tape is dead," Tolliver said. "Tape is making resurgence for a whole host of reasons -- including cost and compliance, reducing the carbon footprint and continuity to keep things up and running. If your building blows up, you're losing your disk with it, I hate to tell you.
"There are archiving solutions based on disk, but then you're spinning more disk, you're buying more disk. And you have to duplicate everything in case your data center blows up. Our department's not big enough to afford all that infrastructure. We can afford a tape library, we can afford tapes, and we can just have multiple copies and offsite them."
Tolliver said his department installed Spectra Logic in 2006 when he was the only full-time IT staffer, partly because its systems were easy to manage for nontechnical people if he wasn't around. He started with a Spectra T50e but upgraded to a larger T380 last year after the department was charged with protecting data from the university television station.
He's been using Spectra's BlueScale management software since 2006, and upgraded to version 11 last year. Tolliver said he likes BlueScale's media lifecycle management capability -- it determines if a tape cartridge is going bad and retires it before it fails and includes an automatic drive cleaning feature that saves him from cleaning the drives too often (and wearing them down) or not enough (which impacts performance).
He also uses the BlueScale Vision Camera, which inserts inside the library and lets him see what is going on through a Web browser. "Besides the 'gee whiz' factor, you can see if something happens to the library," he said. "Our library is in our data center and I'm on a different part of campus, so I use it to see what the library is doing."
Tolliver added Atempo TimeNavigator Navigator Enterprise Server in 2007, scrapping EMC Retrospect because EMC failed to upgrade the Mac-based software application it acquired when it bought Dantz. He also uses Atempo Digital Archive (ADA) archiving software.
Tolliver said Atempo's Mac support and the integration between TimeNavigator and ADA are valuable for his department.
"Eighty percent of our user base is Mac," he said. "When I got here in 2002, there were all Windows servers on the back end. Any IT shop that runs things well will align their servers with a majority of their clients. It's better for the user base and the administrator. We have Mac servers now, we're running Snow Leopard. So we had to research and find the right vendors."
Tolliver said only three of the 40 hosts he protects are directly connected to tape, but those three servers have about 75% of his data attached to them. "The rest of it is all connected to the LAN," he said. "Those are the remote servers and servers not on the SAN, or laptops or desktops. That goes to disk first, then gets moved off to tape."
Data archiving tips: Use a big library, backup-friendly software
Tolliver said ADA "cooperates with our backup system" rather than requiring a separate catalog server or any dedicated hardware. "ADA shares the library, shares the slots, shares the resources, so we did not have to set up a redundant infrastructure for agents," he said. "It uses the Time Navigator setup. Some data archiving applications require you to run a separate partition in your library, and you have to pay for a separate partition license. Much to Spectra's chagrin, we didn't have to come back to them and ask for a license for partitioning because ADA uses the same partition as Time Navigator uses."
He said when a file is archived, Atempo sends it off to tape and inserts a stub. "If [his users] try to open or touch stale data that's been moved or archived to tape, there's an automatic process where they just have to wait about 90 to 120 seconds and their data is back," he said.
"If you want to do archiving, you should have an adequately sized tape library," he said. "We have capacity we can grow into for archiving. So you want to size your library appropriately. The other thing with archiving is, if at all possible, your archiving and backup software vendor should be one and the same. That way you have less redundancy in infrastructure, you have simplified support, simplified service, and you realize all the synergies of the products. I've heard of people having [Quantum] StoreNext for archiving and Symantec for backup; I'm like, 'What are you thinking?' I wouldn't want to deal with that."