By Michele Hope
Babysitting arduous backup-and-restore operations is nobody’s idea of fun. One IT professional who knows about the effort required to monitor traditional backups is Bob Graham, senior vice president of information systems at Farmers and Merchants Bank, in Long Beach, CA. Someone in Graham’s group would spend 24 hours per week, on average, monitoring and correcting backup-related issues surrounding the daily tape backups of each of the bank’s 20 branch offices throughout southern California.
Despite providing each branch with color-coded tapes and a matching schedule, Graham admits the manual process was still fraught with problems. At each branch, non-technical staff had been assigned to monitor and swap out the tapes each night. Invariably, simple human error would lead to nightly backup failures at nearly half the branches.
“The process wasn’t working really well,” says Graham. “When we went to check, we’d find they hadn’t changed the tape correctly, or they had put in the wrong tape. The backup might have also failed for some other reason, such as a problem with the tape drive. Multiply that times 20 and you end up spending a lot of time asking for someone at the branch to help troubleshoot the problem.”
Graham decided a centralized disk-based backup architecture was the answer, and eventually implemented Symantec’s Veritas Backup Exec 10d to pre-schedule disk-based backups at each branch location. The local backups are then replicated nightly from each branch to a Dell NAS 220S storage array at the primary data center. These data sets are subsequently backed up to tape. Graham estimates this process has reduced his team’s backup monitoring time from 24 hours down to only one hour per week, while completely eliminating involvement from staff at each branch.
Reclaiming the time
“Disk-to-disk backup buys you a significant reduction in the amount of human effort needed to back up data,” says Phil Goodwin, president of Diogenes Analytical Laboratories, a consulting and research firm, in Erie, CO. Goodwin believes two-thirds of the time spent managing storage involves triaging daily backup-and-recovery processes. He claims disk-based backup architectures remove much of this effort and can cut an IT organization’s storage management time by as much as 30%.
Goodwin categorizes three types of disk-based backup technologies:
- Virtual tape libraries (VTLs) that emulate physical tape drives and libraries;
- Appliances that combine hardware and backup software; and
- Software-only solutions.
Goodwin notes that VTLs are gaining momentum due to their easier concept sale and purported ease of implementation into existing environments.
Earlier this year, Diogenes published Buyer’s Guides for enterprise and midrange environments that reviewed disk-based backup products in each category. Vendors included were ADIC, Alacritus (since acquired by NetApp), Data Domain, Diligent, EMC, Neartek, and Overland Storage. (For more information, visit www.diogeneslab.com.)
Based on his hands-on review experience, Goodwin cautions users not to trust vendors’ plug-and-play claims, regardless of the technology they choose. “There’s really no substitute for good system design. You still have to pay attention to your backup job schedules, and you also need to look closely at how you will recover and retain data,” he says.
Escaping from restores
Network administrator Ralph Miles decided to move to disk-based backup after seeing the time involved in restoring servers. “I grew impatient with my staff having to sit in front of those servers,” says Miles, who works for the Air Force Center for Environmental Excellence in San Antonio, which helps perform military base conversions and supports the construction of new military housing.
After a server failed, it would often take up to two days to bring another one back online, and the process involved re-installing all of the operating system software, applications, patches, and data.
Ron Roberts, an Atlanta-based reseller at BluPointe DRS LLC, suggested that Miles look at Asigra’s disk-based backup and bare-metal restore functionality. After seeing how it reduced server restores to just a few hours, Miles decided to use Asigra’s Televaulting software to back up Windows 2000 servers and desktops at both the local data center and regional offices in San Francisco, Atlanta, and Dallas.
The Televaulting software now backs up each computer’s application services, system state, and drive contents to a local server, while encrypting all sensitive data. The software also replicates remote backup sets to the main data center over a T1 line, where the data resides on a direct-attached Xiotech Magnitude disk array.
Miles now uses Asigra’s software to back up about 15 laptops in San Francisco as well. In the event a laptop gets stolen or damaged while an employee is traveling, he’s confident in his team’s ability to quickly solve the problem. “We can now rebuild the laptop from the data we backed up last night, then just FedEx the new laptop to them the next day,” he says.
Eventually, Miles plans to phase out his tape-based backup. For now, he still uses Qualstar tape silos and Legato NetWorker as another layer of insurance to back up the Asigra main vault. “I’m a person who wears both a belt and suspenders,” says Miles. “I’m still running my tape system as well as the Asigra solution. I’m going at disk backup full force, but I’m giving the technology time to season. You can think of it as holding onto my suspenders at the same time I tighten my belt.”
The mantra of “one full backup followed by six incrementals” has been drilled into most seasoned storage administrators. This can sometimes mean an uphill political battle to implement new disk-to-disk (D2D) solutions.
Bernard Shen, a principal consultant at Los Angeles-based Sofbel Inc., is familiar with convincing IT groups to part with such sacred cows in backup and restore. He recently helped a large, multinational defense sector company implement a STORServer disk-based appliance. The company planned to use the appliance to back up 27TB of data housed on 120 Windows and Sun servers.
The STORserver D2D appliance runs a version of Tivoli Storage Manager (TSM) that applies a different, object-based paradigm to backup. This allows TSM to perform only one full backup, followed by subsequent, ongoing recordings of only block-level changes made to each original source file. These block-level changes typically require backing up only 250GB of changed data per day across the site’s servers.
Because of this design, Shen notes the concept of full, weekly backups and even standard incrementals simply no longer applies in a D2D environment. Instead, the appliance’s disk-based caching allows the company to use smaller computers and a more affordable infrastructure for backup-while avoiding the need to invest in a larger pipe or faster disks.
Shen admits the paradigm shift is often a tough one to make for administrators used to the old ways: “Every administrator grew up doing full backups and incrementals. If you tell them you don’t have to do incrementals, it just doesn’t fly. Likewise, they believe, ‘If I don’t see a full backup, it must not be safe.’ ”
Brian Arnold, a senior systems administrator at a major Detroit area advertising agency, is also no stranger to slaying his organization’s own sacred backup cows. “We still do a full backup run every week,” he says, but his ultimate goal is to move his firm away from the use of traditional backup software altogether.
Arnold implemented FalconStor’s IPStor software with its own VTL service in an effort to consolidate separate islands of storage, tape silos, and separate backup solutions associated with his firm’s Macintosh, Solaris, and Windows environments. “Initially, we decided to use virtual tape to stage our backups on all platforms to a common tape solution and then send that to a Sun L180 physical tape library we have,” says Arnold. “Since the VTL then writes large data streams directly to physical tape, we’ve eliminated much of the tape maintenance costs that result from tape ‘shoeshining.’ ”
With that phase completed, Arnold has since turned to plans for implementing FalconStor’s TimeMark feature. This would allow the company to capture (and quickly remount anywhere on the network) granular snapshots of any data changes down to the minute, if needed. Once that happens, Arnold hopes his group can completely retire its use of tape-based backups using EMC/Dantz Retrospect and EMC/Legato NetWorker.
Michele Hope is a freelance writer and can be contacted at firstname.lastname@example.org.