For quite some time, RAID was basically limited to high-end SCSI workstations and servers, far out of the reach of most power users. However, in the past couple of years we've seen several IDE RAID solutions that aim to bring RAID's benefits to IDE systems. These solutions, available on PCI cards and motherboards, are popping up everywhere these days. Given its ready availability and theoretical benefits, it's only natural that we give IDE RAID a real-world workout to see how it really performs.
We've run the Highpoint 370 IDE controller (on an Abit KT7-RAID motherboard) through a gauntlet of benchmarks that reflect and measure real-world performance to see if RAID is really worth the cost of an extra hard drive. In true TR fashion, the answers are presented with large amounts of gratuitous graphing. And the results might surprise you.
Before we get into the results, let's take a moment to discuss what RAID actually is. RAID stands for Redundant Array of Inexpensive Disks (some insist the I stands for Independent), and comes in several different flavours. For the purposes of this review, we'll just be discussing RAID 0, RAID 1, and RAID 10 (also known as RAID 0+1), since they're the only RAID variants supported by most IDE RAID chipsets.
- RAID 0 (Striping) - RAID 0 combines two or more physical hard drives into a single logical drive. Data is striped (distributed in blocks during writing) across all drives, giving you a single logical drive with a capacity equal to the sum of the capacities of all drives used. RAID 0 increases performance by spreading data onto multiple drives, though it does so at the expense of reliability. Because data is distributed across all disks, if one disk in the array fails, the result is the same as if all had failed: total data loss.
- RAID 1 (Mirroring) - RAID 1 involves one or more physical drives being exactly duplicated on other physical drive(s). Any data written to one drive or set of drives is written to its "mirror" drive or set, providing an exact duplicate of the original drive(s) on the mirror drive(s). RAID 1 provides increased reliability through data redundancy; in the event of a drive failure, a copy of all the data from that drive exists intact on that drive's mirror. Of course, this peace of mind comes at the expense of storage space. Since all data is stored twice, total storage capacity is equal to only half the total capacity of the drives in the mirror.
- RAID 10 (0+1) - As you may have guessed, RAID 10 is essentially a combination of RAID 1 and RAID 0. You can think of it either as a striped array of mirrored drives or a mirrored array of striped drives. In any case, it uses at least four physical drives and gives the equivalent storage of half the drives in the array. RAID 10 is designed to give you both the performance benefits of RAID 0 and the data redundancy of RAID 1.
Before we get into benchmarks that will show the performance aspects of RAID, let's talk about RAID's other benefit: minimizing the potential for data loss.