The Point Networks - Tech Information

RAID stands for redundant array of independent (or inexpensive or intelligent, depending on who you ask) drives. It's a technology that allows you to create a large, single, high-performance, logical hard drive from an array of small hard drives.

As processors, RAM, and video cards get better, the bottleneck is increasingly your hard drive. Unlike the days when you might have had one word processing file open, people today work with large media files and access the hard drive much more. So RAID makes a lot of sense. It's one area where Leo thinks you're going to see some improvement in the overall speed of a system.

The main idea in RAID is something called striping. Imagine two HD drives, one on top of another. Now take one sector in the top drive and one in the bottom drive. When you stripe data, the information is written across both sectors, across the drives, as if they were a single unit. The data is interleaved between both drives in an overlapping fashion. As a sector is read from the top drive, the read continues onto the bottom on an alternate sector.

What are the advantages? It depends which RAID level configuration you use. RAID levels have nothing to do with quality. A level 5 RAID isn't better than a level 1 RAID. Each level has different benefits. A quick breakdown:

RAID Level 0: Performs pure data striping, offering the ultimate in performance and drive efficiency. An oxymoron of sorts as it doesn't offer any level of redundancy whatsoever. Requires at least two drives. Drive striping allows two or more drives to be used as one.

RAID Level 1: Mirroring. This one duplicates data from one drive to the others within the array. It offers the same amount of storage as a single drive in the array but with complete data redundancy. Disable one drive and the others can still supply the data. A level 1 RAID writes slow compared to a single drive but makes up for it with faster reads. Choose this level for a high fault-tolerance environment. Requires at least two drives.

RAID Level 2: Offers the use of Hamming ECC (error correction code). It gives each bit an ECC and stores it on an ECC drive. When reading data, it checks against the ECC drive(s) to verify data intergrity. Of the five RAID levels this is the least used. It's an inefficient use of drives, costs a lot, and the Hamming ECC is irrelevant since SCSI and most new HDs support built-in error correction. Requires at least three drives.

RAID Level 3: Offers data striping at the bit level. It stores data parity on a parity disk with data correction on reads. You'll need a hardware controller for best results though, and controller design is complex. It's best for environments using long sequential data, like digital video editing. Not good in a multiuser situation. Requires at least three drives.

RAID Level 4: Similar to Level 3 except writes data at block level, not bit level. Its read performance is similar to Level 0. Level 4 writes much slower because the parity data must be updated at each write, although sequential or large writes are much faster. It offers low cost per megabyte but makes for difficult data rebuild in event of disk failure. Little to no advantage over Level 5 with no support for multiple simultaneous writes. Requires at least three hard drives.

RAID Level 5: Similar to Level 4 but with parity data distributed across all disks. Great in multiuser environments like file or Web servers. Requires at least three hard drives, but better with five.

What are the disadvantages? Number one is cost. You need at least two drives, and while you can use software, a hardware RAID controller offers better performance. Second, above Level 0, RAID uses valuable storage space for data redundancy. Complicated setups can mean significant downtime. And it can be just plain overkill. If all you do is surf the Web and play the occasional 3D shooter there's really no need to spend the bucks for a RAID setup when you can send your kid to college instead.