There is no single universal backup strategy or product -- one size does not fit all. A backup strategy must be developed by each individual organization to accommodate its business needs, liabilities, technical expertise and budget. Even after a strategy is developed and put into place, it's only a work-in-progress that must be revisited and refined over time as the business changes.
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
Minimizing lost data
Lost data is absolutely unacceptable for Prairie Packaging, a foam food service product manufacturer in Bedford Park, Ill. With some 600 global users accessing 7-10 terabytes (TB) of time-critical data, the potential gap between data loss and recovery point was simply too large -- especially for its Oracle enterprise resource planning, scheduling and busy warehouse management systems. Loosing track of what was manufactured, warehoused or shipped between data loss and recovery proved intimidating. "Although we can try to recreate that [lost data], it becomes a very expensive and a very time-consuming project," says Manny Singh, director of IT at Prairie Packaging. "In our estimates, it took three times as long to figure out the missing pieces than it did to actually do the recovery." The drive to shorten recovery point objectivess eventually led Singh to continuous data protection (CDP) technology.
The adoption of CDP involved a serious consideration of data and its importance to the business. "If we loose 'this' amount of data, what does that do to the business from a cost driver perspective?" Singh says. This included an evaluation of fines, labor costs and other ramifications of missed shipping schedules. Those possible losses were then compared against the costs of CDP technology. According to Singh, the reworked CDP strategy took about 45 days to formulate and involved at least 10 people from across the company. "The group consisted of both business and IT people," he says, noting that the group considered accounting, distribution, supply chain and IT implications of the strategy. "We added up those numbers and then took those up to management."
For Singh, the single biggest problem with its new CDP strategy was finding a vendor to supply the actual products. "Although it took us 45 days to come up with the plan, it took us probably six-eight months to find a vendor that could fit that niche." Most of the products Singh evaluated were snapshot based, which typically required production to effectively stop during each snapshot -- an impossible expectation.
Today, Prairie Packaging is transferring critical data to Apple Computer Inc. X-RAID platforms, which is then passing the data to a Revivio Inc. CDP product. Prior to the introduction of CDP, data was passed from the X-RAIDs to tape using ARCserve software from CA Inc. Singh's team re-evaluates the backup strategy every quarter and has conducted three such re-evaluations since CDP implementation. "We've decided that we're going to add another Revivio to take on the remaining applications we have, and it will be hosted at a different location," Singh says. "It will really become part of our overall DR [disaster recovery] plan."
Eliminating the conventional backup window
Most organizations treat backups as discrete events that stop productivity and divert network resources to accomplish the backup task. This is the primary reason why most traditional backups are performed during periods of low network utilization -- evenings and weekends. With a growing number of companies embracing 24/7 operation, however, the notion of a "backup window" is quickly becoming outdated. The Weather Channel, headquartered in Atlanta, Ga., is one organization that has actively moved away from traditional backup windows by deploying a disk-to-disk-to-disk strategy.
"We don't have set backup windows," says Kevin Gungiah, director of technology solutions at The Weather Channel. "So [critical data] backups are triggered at different points of the day." Critical business applications are generally backed up to disk throughout the day using a hot/hot scheme, allowing one set of servers to be backed up while leaving the applications accessible from companion servers. Less critical backups are also made daily, weekly and monthly. Daily backups are kept for one week, weekly backups are kept for one month and monthly backups are retained for six months to a year. "Our recovery strategy determines what to backup and how we back it up, and the length of time we keep that backup," he says.
The Weather Channel had previously adopted a disk-to-disk-to-tape strategy but quickly discovered that approach did little to address recovery needs. "We did a course correction by realizing that we really needed to look at our recovery strategies first, before just backing up everything in the environment," Gungiah says, noting that 100% of backed up data is not necessary to recover critical business systems.
Today, The Weather Channel does not backup 100% of its 90 TB data storage capacity because much of that data is temporary. Instead, the focus is on backing up Office applications, Exchange data, sales and marketing information, and its operating environment. "When I look at the other systems that are really core to the business, we tend not to backup all of the data that supports those applications because it's so transient," Gungiah says.
According to Gungiah, The Weather Channel is constantly re-evaluating its backup strategy -- seeking to improve efficiency and utilize less space, archive infrequently accessed information and apply data classification concepts to move data more intelligently across tiers. "I back up half a terabyte in my Exchange environment today on a daily basis," Gungiah says. "My objective is to try and reduce that to 60 GB of active mail -- significantly reducing my backup volume."
Handling multiple networks and remote backups
A distributed enterprise faces some unique backup strategy challenges -- especially for smaller organizations that cannot afford to staff IT professionals at every location. Each site must often be backed up and restored independently, and this demands an unprecedented level of attention from the storage administrator who simply cannot be everywhere at once. "You have to be able to backup data across the board," says Al Zaccario, director of information technology at New Castle Hotels and Resorts LLC. "The site that has three nodes is just as important as the one with 70."
Zaccario is responsible for backing up about 60 GB each night from up to 25 managed hospitality properties. This full backup volume may seem small when compared to large enterprise needs, but it's still important business for him. "An incremental backup is harder to restore," he says. "We don't have the expertise in the field to be able to do that." Each property is configured independently with seven different backup jobs per week. Zaccario also uses LiveVault's InControl software to backup critical files on servers throughout the company to two Dell Inc. 250 GB PowerVault systems located at a central site across a WAN link.
Virtual private network access allows Zaccario to access the backup systems on each property to ensure that the job has been completed, though even routine completion checks are a logistical challenge. "There's 25 locations, and I probably can get to about half or three quarters of them per week," he says, noting that nontechnical personnel on each site can help to identify any instances of backup problems. Backing up mobile users is particularly problematic, since each user must manually execute a backup batch file. "That's where I get burned," Zaccario says.
As IT technology and WAN connectivity continue to improve, the biggest challenge is getting property managers to see the importance of backup tasks -- further increasing the need for careful backup planning. "Users are now so used to things going right that they are not prepared when things go wrong," Zaccario says. "The reliability of the systems is now so high that people are almost shocked when they don't work." Zaccario notes that backup planning is ongoing but concedes that the process is reactionary because of his limited resources. Backups typically continue until a problem arises that can then be addressed. "Strategic is not really one of my options."
Go to the next part of this article: Backup strategies: Future directions
Or skip to the section of interest: