Until recently, one advantage that mainframe systems had over PC-based network solutions was a well-deserved reputation based on their ability to provide safe, secure, reliable and quick direct-access storage devices.
That technology has trickled down to the microcomputer world: The Redundant Arrays of Inexpensive Disks (RAID) drive subsystem architecture allows for data redundancy and protection using multiple drives.
At this time, there more than five levels of RAID implementation, with a few under discussion. The main five implementations are not aggregated levels, where each level builds upon the prior level; rather, they are five different methods of implementation. Recovering RAID 5 is often the easiest.
RAID Level 1 uses mirrored disks to provide complete data redundancy on a one-to-one basis. Each disk has a twin that contains the same data as the primary disk. This is the most common method used in corporate America today.
Novell Inc.’s NetWare provides for two different versions of RAID Level 1: In disk mirroring, two drives are attached to the same disk controller, while in disk duplexing each drive uses its own individual controller. These configurations provide 100 percent data redundancy at a 100 percent (or greater) increase in cost per megabyte.
RAID Level 2 provides for multiple parity drives, and has a slightly lower cost per megabyte because less than 50 percent of the total available storage is used to maintain data integrity.
Although it is prevalent in the supercomputer and mainframe worlds, however, no products that support Level 2 are currently available for the microcomputer market.
RAID Level 3, which provides for a minimum of two data drives plus a dedicated parity drive, is the most commonly available implementation of the technology.
The parity drive is used to maintain the error-correction code (ECC) information necessary to rebuild a failed data drive, and the information on the drives is striped across all the available data drives. Performance is improved over RAID Level 2, and the cost per megabyte drops even further because there is only one parity drive.
RAID Level 4 continues the use of dedicated ECC drives to provide data integrity. But performance increases because files are striped in blocks rather than in bits, as is done in Levels 2 and 3. Block-based striping improves performance because the necessity for synchronization is removed. Level 4 also provides the ability to perform multiple simultaneous reads, which also can significantly improve system throughput.
RAID Level 5 does away with the dedicated ECC drive by striping both data and ECC information across all available drives. Adding to Level 4’s ability to perform multiple simultaneous reads is the more important feature that allows multiple simultaneous writes. RAID Level 5’s greatest advantage is its ability to perform reads and writes in parallel. RAID Level 5 also can provide the lowest cost per megabyte in a redundant data-protection storage scheme.
Advanced features under discussion as part of RAID Level 6 include the ability to recover from multiple concurrent disk failures.
But that isn’t to say that RAID is completely flawless.
“To be frank, RAID arrays do fail quite often,” says Dave Masters, engineer at Hard Drive Recovery Associates in Irvine, a RAID recovery specialist, “In fact, too many people see the redundancy as a replacement for a good backup plan, which of course they should not.”
A non-RAID-specific characteristic becoming more prevalent and tied to the RAID phenomenon is the greater acceptance of fault-tolerant features, which include not only the previously discussed RAID implementations but also hot swapping.
Hot swapping is the ability to replace components without bringing down the disk subsystem. Currently, only hard drives and power supplies benefit from this capability. Failure of the hard-drive controller or SCSI (Small Computer System Interface) adapter still results in system failure.
Many manufacturers are jumping on the RAID bandwagon. Compaq Computer Corp., for instance, started the trend with the Intelligent Drive Array included with its Systempro computers. Dell Computer Corp. followed with its own RAID implementation, which showed greater performance on medium-sized networks. Zeos International Ltd. and Northgate Computer Systems Inc. have also announced RAID products.
Not to be left out, the hard-drive vendor community has also jumped aboard. RAID arrays are now available from manufacturers such as Legacy Storage Systems Inc., Micropolis Corp. and Core International which offer high-end RAID systems with hot-swappable drives and power supplies, with matching high-end price tags.
The enabling technology for RAID, in the form of software that allows operating systems to take advantage of the features provided by the various RAID implementations, is provided to OEMs by software vendors such as Chantal Systems Corp., which provides the software for products such as Micropolis’ Raidion drive array, and Integra Computing, whose OASIS product provides the software that IBM has been demonstrating to enabled RAID services on high-end PS/2 file-server systems.