What you will learn from this tip: Want to avoid
Here's a question for you: If hard disk drives get cheap enough, will storage system vendors start to bypass complex, parity-based data protection schemes like RAID-5?
It's an idea whose time may have come. Consider ExaGrid Systems, a network-attached storage (NAS) startup in Westborough, Mass., which last month announced the general availability of Advanstor, a storage system that incorporates NAS, backup, automated data migration -- a.k.a. hierarchical storage management (HSM) -- as well as remote replication and disaster recovery, but forgoes RAID altogether.
"It's not that there's anything wrong with RAID -- there isn't," says Mark Kaufman, ExaGrid president and CEO. "But for us, it just added cost and complexity without adding much value."
Instead of RAID within the array, Advanstor protects the data on its primary storage tier by mirroring it multiple times to separate, inexpensive storage nodes, "GRIDdisks" in ExaGrid-speak. GRIDdisks are 1U servers equipped with four 250 GB SATA drives in a JBOD configuration, plus ExaGrid's proprietary software. By default, the system provides "2x2" protection -- one copy on primary storage, one on local GRIDdisk storage and two copies at two distinct GRIDdisks at a remote location.
ExaGrid's Advanstor is not the first storage system to use this architecture, usually referred to as Redundant Arrays of Inexpensive Nodes (RAIN), says Arun Taneja, consulting analyst and founder of the Taneja Group, Hopkinton, Mass. EMC's Centera does as well. But at $2/GB for raw Centera storage, its mirroring architecture brings the price up to $4/GB -- a lot for archival-class storage. These days, Centera also supports EMC's proprietary RAID-S, which brings the cost of protected Centera storage down to approximately $3.5/GB, Taneja says.
The economics of ExaGrid's Advanstor are different, though. An Advanstor bundle that includes 1 TB of high-performance primary NAS, secondary GRIDdisks and all the software is listed at $62,000. Additional 1 TB GRIDdisks cost approximately $5,000.
From a raw capacity standpoint, Advanstor's pricing may be a bit high, says Taneja. At the same time, the system is a good fit for small and medium-sized businesses (SMBs) that "want to buy something as close to self-contained as possible," he adds.
Foregoing RAID has another advantage -- simplicity, says Dave Therrien, ExaGrid CTO and founder. Because there's no software or hardware RAID to track, ExaGrid can easily change hardware suppliers for its GRIDdisks. "With every tweak you make to the hardware layer, you have to track another layer of firmware," he says. "It's better to keep it simple and make up the functionality in our software."
Advanstor's 2x2 protection equals four copies of a file. That may seem like overkill to some, but not to Robert Reeder, CIO at W.A. Wilde, a direct marketer and fulfillment company in Holliston, Mass. The company is beta-testing Advanstor as a replacement for a more traditional storage setup, i.e., file servers protected with tape backup and no remote replication.
"I'm not really bothered by it," Reeder says of the 2x2 protection. "I have seen so many backup and recovery strategies that try to be foolproof and fail." By spreading the data across four independent disk locations, "I'm not relying on a single server not failing, but on the laws of probability that state that I can't possibly lose all four copies," he explains.
Aaron Sawchuk, CTO at ColoSpace Inc., a hosting service provider in Rockland, MA, and another Advanstor beta-tester, agrees. "This gives us the ability to take a significant hit at one location and still have protected data," he says. ColoSpace is using Advanstor as a backup target for its various application servers, and replicates between its Boston and Watertown, Mass., offices.
But while companies like ExaGrid are examining ways to forego parity RAID schemes, others are looking to enhance them. InoStor, a spin-off of Tandberg Data, recently announced the InteliNAS line of NAS arrays that feature the company's patented RAIDn technology, where "n" can be set to the number of simultaneous disk drive failures you want to be able to withstand, explains David Licosati, vice president of business development.
Other vendors offering RAID levels that protect against multiple disk drive failures are thus far limited to NetApp with RAID-DP and Hewlett-Packard with its Advanced Data Guarding (ADG) option on select storage controllers. IBM has also stated that it's working on dual-parity RAID protection.
In the meantime, plain old RAID-5 will probably be around for a long time to come. "There's a lot of industry inertia around RAID-0, 1, 4 and 5," says Jeff Hornung, vice president and general manager for NetApp's Gateway Business Unit. "It's very well adopted, and IT shops are very comfortable with it." Mirroring, in his mind, is on the outs. "Deployments that do require mirrored architectures are moving toward more mission-critical environments. In local environments, I think you'll see less and less of it in favor of dual-parity systems."
Checklist: Seven RAID configuration essentials
About the author: Alex Barrett is "Storage" magazine's trends editor.
This was first published in December 2004