February 14, 2003 By Merrill Douglas
Full and Incremental
The server backup software -- IBM's ADSTAR Distributed Storage Manager, now called Tivoli Storage Manager (TSM) -- brought major change to the city's backup procedure.
In the past, Wichita -- like many enterprises -- used a combination of full and incremental backups. Once a week, it backed up all data on its systems, taking a snapshot of the current information. On subsequent days, it backed up only the files that changed in the past 24 hours. On day seven it did another full backup. Without periodic full backups, Wichita had large volumes of tapes holding many generations of changed files; to restore a file, operators had to hunt through those tapes to find the right one.
TSM requires just one full backup, after which it makes incremental backups and automatically tracks storage locations. When someone needs to restore a file, TSM locates it in a pool of storage media. Using STORServer's software, a manager sets policies to indicate how long the system should store different kinds of files. After the expiration date, the system automatically writes over that file; this limits the number of tapes needed.
STORServer's system first backs up data to an array of hard drives, then transfers it to the tape library. An operator who needs to restore a file backed up the previous night can retrieve it faster from the hard drives than from tape, Norman said.
The software also changed the way Wichita provides for disaster recovery. In the past, Wichita stored its backups in an off-site vault; if a file needed restoration, someone retrieved the appropriate tape from that location. But TSM automatically makes two copies of each backup. The city keeps one set of tapes on site where they're handy for restoring on short notice while the other set is in the vault, safeguarding the data in case of a disaster.
A Half-Hour Job
The new system requires less time for daily backups, uses fewer tapes and requires much less operator participation. Instead of supervising the backup until 11:00 each night, an operator spends 30 minutes in the evening removing blank tapes and adding new ones as needed. During the day, an analyst checks the system to make sure it's performing correctly.
"My evening operator job is gone," Norman said, noting that loss saves the city the cost of an additional worker. "That position has been deployed to do other things."
Although the new backup system works faster, the process still runs all night. That's mainly because the city has added so many information systems and servers since 1996.
"We're still struggling to meet our backup window, and we always will," Norman said. "If we had it to the point where we could actually back up everything in two or three hours, we've overbought."
To meet the challenges of its ever-expanding information-services enterprise, Wichita has made two upgrades to the system since 1996. In December 1999, it replaced the tape library with an 80-slot system using AIT-2 technology, and early this year it replaced the entire backup system with a more powerful one, including a Compaq ProLiant ML530 server with 24 72.8 gigabyte SCSI hard drives and a 180-slot AIT-3 tape library.
Rather than buying one of STORServer's backup appliances, Wichita worked with Dataedge to specify the exact system it wanted, using faster equipment than STORServer offered at the time, Norman said.
Norman could not locate records indicating the total cost for all components of the original system. But he said the hardware cost $25,000, and the city has spent about $30,000 per year for maintenance. For the new system, the city is paying $42,000 for the server and all its disk drives and $90,000 for the tape library, including the drives and tapes. Those figures include a total of $15,000 for consulting services, installation and data migration.
Wichita has not tracked the return on its investment in the new backup technology. Payback wasn't even part of the justification, Norman said, explaining that the main goal was to build a backup system that could grow with the organization.
"We try to anticipate obsolescence and growth in demand for new systems," he said. "Otherwise, systems are replaced on an emergency basis, which can shorten the lifetime of the solution."
Bio: Contributing Writer Merrill Douglas is a freelance writer based in upstate New York. She specializes in applications of information technology. E-mail: email@example.com
You may use or reference this story with attribution and a link to