E-marketing firm deploys multifaceted grid storage system

Robert Reeder knows what failure is. Earlier this year, a drive on the company's EMC Clariion RAID 5-based system failed, corrupting all its data. When the IT staff turned to its tape backup system to restore the data, that procedure failed as well, resulting in a week's worth of lost data.

This is not an experience that Reeder, CIO at W.A. Wilde, an e-marketing firm in Holliston, Mass., can afford to repeat. The company puts 500 million pieces of direct mail into the postal system each year. "We're all about data now -- it's what drives the documents and what you do with them," Reeder says.

Disaster lurks

That's why, when W.A. Wilde began beta-testing a storage system whose design philosophy seemed to assume that yes, hardware fails, and yes, you have to plan for it, Reeder took an immediate liking to the system. "With RAID 5, the system is called 'fault tolerant,' but it's not," he says. "If one bit is out of place, there's the potential for not being able to restore your data."

The system that Reeder has been testing is Advanstor from ExaGrid Systems in Westborough, Mass. Advanstor is a "self-protecting" NAS system that offers several levels of data protection in one integrated system, including primary RAID 5-based storage, local and remote disk-based backup, off-site disaster recovery, data archiving, data restore and automatic migration across two-tiered storage.

As Reeder sees it, ExaGrid's philosophy seems to be, "Let's build a strategy where we always have a copy of data, no matter what fails." And because the copies are saved on cheap disk, "you don't have to count on sophisticated technology never failing," he says.

NAS and beyond

W.A. Wilde didn't start out with the notion of acquiring such a multifaceted system. Over the past decade, it stored data in many different servers throughout the company, including two in Holliston, and one in Brockton, Mass. The company used EMC Clariion servers for its databases and e-mail, and a variety of other storage servers for individual customer projects.

However, as prices declined on NAS-based systems, the company began looking into consolidating its storage to better control its capacity and storage utilization. "When we had growth areas of the business, we were sometimes using storage that wasn't meant for that purpose, but it's where we had room," Reeder says. "We wanted to do NAS so that everything was housed in one or two locations."

Reeder considered a NAS system from Dell, but that opened up another can of worms, he says. "It only solved the NAS part. We thought, 'How does this change our backup needs? Do we have enough tape capacity?' It was a whole other problem."

In this way, W.A. Wilde was like a lot of other medium-size companies that use a manual process to protect their data. Most start with one NAS box, and when they need more capacity, they simply add more NAS boxes that are backed up to a tape subsystem, says Mark Kaufman, CEO at ExaGrid. Those tapes need to be shipped offsite, which adds a lot of human intervention.

As companies begin to think about off-site disaster recovery for further protection, this just adds to the cost and headaches, Kaufman argues. "People are spending a lot of money, dealing with a lot of complexity and leaving a lot of room for human error," he says.

Advanstor is intended to target these people by integrating the functions of backup, disaster recovery and data migration in one NAS system. Moreover, by having the solution integrated, "the sum of its parts is greater than the whole," Reeder points out.

Three systems in one

Here's how it works: The front end of Advanstor is composed of a Dell RAID-5 multiprocessor server running Windows Storage Server 2003, as well as ExaGrid's data protection and migration software. This is connected through a Gigabit Ethernet backplane to a grid of SATA disks that function as a virtualized repository.

When it's time to back up data, the system takes a snapshot of what's on the RAID system, and for new files, it stores at least one full copy on one or more of the grid disks. For existing files that have been changed, it saves the latest image and then creates a "delta" file to preserve a record of older copies. This way, there is always a master copy of data, at least one full backup image of the latest version, and delta copies for reference or deletion at will.

In addition, data copies are also sent through a VPN connection to another fully configured Advanstor system at a remote location. Each site is capable of replicating to the other. "By having the system handle disaster recovery in this integrated fashion, it's trivial for the system to fail over to the other site," Reeder says.

The back-end grid disks are critical to the self-healing nature of the system, Kaufman explains. If one grid disk fails, another grid disk can take over, not just to store the data that was on the failed disk but also to fulfill backup and protection policies. "The grid disks are not managed as individual grid disks but as a single repository," Kaufman says.

Reeder also takes advantage of the third component of the system -- data migration. When the front-end NAS filer begins to fill up, migration software detects files that haven't been recently accessed. It moves these files to the back-end grid repository, leaving a pointer on the NAS filer.

According to Kaufman, "To the users, it's still on the filer because if they reference it, the system brings the data back -- and quickly, since it's over a Gigabit Ethernet connection." This accomplishes two things: you aren't unexpectedly running out of space on the main storage system, and the most expensive storage platform is only storing active data.

Notes Reeder, "The whole concept is that I can use the most cost-effective storage, and I don't care if it fails. The NAS takes care of hierarchical storage management, backup and disaster recovery, and it's all available via one Web interface, so the IS department can manage what used to be three different functions."

Reeder can also customize protection policies to meet the requirements of each client. For instance, some projects might need a noon backup, while others should be backed up at night, since data is continuing to be updated during the day.

Big changes for Wilde

The total cost for a 1TB Advanstor system is $62,000, which includes local and remote site configurations, as well as primary NAS filer storage. In terms of capacity, each front-end NAS filer is capable of storing 600GB of active data. The back-end grid disks can scale to 10 terabytes of protected data.

To Reeder, this is a bargain. "If we got a high-speed tape jukebox that could automatically back up a terabyte or two of data every night, we'd be looking at an investment that's comparable to what ExaGrid is offering, and we still wouldn't have their storage management or disaster recovery capabilities," he says.

For W.A. Wilde, Advanstor doesn't spell the end of the tape backup system -- at least not yet. For instance, a big job may come in that will take a week to print, but once it's finished, there's no need to hang on to the data. In these cases, tape is used for nightly backups of the work in progress.

"It's a short-term need, and we don't want to fill up the filer," Reeder explains. "We look at ExaGrid for data that has persistence and for which we need to track versions as it changes."

Still, Advanstor has permanently changed how W.A. Wilde approaches storage. "Part of centralizing all our storage is we're also cleaning it up and figuring out what we need to put on Exagrid," Reeder says. "We want to really organize our storage, which hasn't been done in 10 years."

Mary Brandel is a freelance writer in Grand Rapids, Mich. She can be reached at mary.brandel@comcast.net .

Copyright © 2005 IDG Communications, Inc.

7 inconvenient truths about the hybrid work trend
Shop Tech Products at Amazon