What a difference a decade makes. Back in the "old" days, if one uttered the word "RAID," everyone in hearing range hushed in reverence. These days, nobody talks about RAID in terms other than commoditization.
"RAID is a commodity piece these days," says Kenneth Hill, Technical Manager of Sun StorEdge at Sun Microsystems. "It has been reduced to a component of a larger system."
Announcements about RAID 6 made at last month's Storage Networking World conference in Phoenix, Ariz. indicate change may be afoot. RAID 6 may be bringing some glamour to the mature and somewhat stodgy concept of RAID, and thus restoring the romance enterprises once had with it.
"The growing requirements of today's storage industry is driving mission-critical applications like database and Internet servers to implement methods of enhanced data protection, which is offered by RAID-6 technology," said Roger Cox, research vice president, Gartner Storage Research.
RAIDers of the Lost Disk
RAID means Redundant Array of Inexpensive (or Independent, depending on whom you ask) Disks. It is a means of pooling data storage using several hard drives to achieve fault tolerance, higher throughput, and other benefits. The term was coined in 1988 in a paper that outlined RAID levels 1 through 5.
The most widely used levels and their advantages and disadvantages are:
- RAID O, which provides data striping across multiple disks to increase performance but does not provide any safeguards against failure
- RAID 1, which provides 100-percent duplication of data for high reliability but effectively doubles storage requirements
- RAID 5, in which data is striped across three or more drives for performance, and parity bits are used for fault tolerance, thus cutting down the storage requirements while offering reliability
Since that initial specification, hybrids have been released combining the basic RAID flavors, such as RAID 10 (a combination of RAID 1 and RAID 0) and RAID 50 (a combination of RAID 5 and RAID 0). RAID 6 came out several years ago under various names Advanced Data Guarding (ADG), by Hewlett-Packard and RAID-5 dual parity, by IBM before RAID 6 became the accepted term.
RAID 6 is especially critical in Serial Attached SCSI (SAS) environments where systems often implement lower Mean Time Between Failure (MTBF) SATA spindles in combination with enterprise SAS hard disk drives.
"RAID 5 had a limit of somewhere between 10 and 14 drives in one system," says Richard Scruggs, HP's product manager for server storage. "RAID 6 enables you to use as many drives as you want, with a theoretical limit of 40,000 drives."
"Chose RAID 10 if you want high performance, but even RAID 5 will give you better performance than RAID 6 ... For high capacity and the best failover protection, opt for RAID 6." Richard Scruggs, HP product manager for server storage
RAID 5 is designed to accommodate one drive failure. When one drive fails, the system continues without any data loss. But if a second drive goes down, the data is lost. RAID 6's big appeal is that it continues running despite two disk failures. The chances of that second drive failing are accentuated by the fact that SATA disks have a failure rate six times greater than enterprise-class drives. In situations where you are rolling out SAS-based systems and using an abundance of less-expensive SATA drives, RAID 6 becomes very attractive. There is a performance hit, however.
"Chose RAID 10 if you want high performance, but even RAID 5 will give you better performance than RAID 6," says Scruggs. "But for high capacity and the best failover protection, opt for RAID 6."
HP is partnering with LSI Logic to bring RAID 6 products to market. An LSI Logic 4-port LSISAS1064 controller running on an HP SAS system enables IT users to recover data in the event of two drive failures. MegaRAID SAS 300-8E is an 8-port PCI Express SAS RAID adapter that connects to a 12-drive nStor SAS enclosure. It allows users to mix higher-performance disks with lower-cost Serial ATA (SATA) disks. In addition, LSI Logic has released a PCI Express to 8-port 3 Gb/s SAS RAID on Chip (ROC). Scruggs said this technology will be incorporated into the HP Smart Array 6402 and 6404 within the next month or so.
Meanwhile, other vendors are following suit. IBM will use Adaptec RAID-6 technology in an upcoming SAS RAID server. Ciprico has a 4 Gb/sec RAID-6 digital media system on the horizon, and Intel is working with vendors, such as Promise Technology, IBM, Ario Data Networks, and Areca to heavily promote the value of RAID 6. Intel has introduced a special RAID-6 I/O processor, which these firms are using in their various systems.
According to Intel's numbers, it would take 60 disks using RAID 5 only six months to experience a complete system collapse (i.e., two disks failing). With RAID 6, on the other hand, it would take 25 years for a system failure.
RAID-6 gear that is either available or about to hit the market, includes: the Promise SuperTrak RAID-6 system, Ario's SANArio RAID controller, and Areca's RAID 6 SATA controller.
"Our RAID-6 card will be out this summer," says Tom McKee, marketing communications manager at Promise Technology. "Our goal is to take high-level RAID-6 technology and make it available and affordable for the SMB market."
Rasilient Systems is also planning to release server-based SAS storage systems that harness RAID 6 by the end of the year. Currently, they are available in RAID 0, 1, 10, and 5. The company's business model is to keep costs low by using standard motherboards, chips, and other parts, along with Linux-based software.
"Some customers want SAN, others want NAS, iSCSI, SAS, SATA, or RAID 6," said Sean Chang, president of Rasilient Systems. "We offer a flexible product with high performance at low cost."