It’s hard to understate the impact of NAND flash and SSDs. Over the last decade, these storage solutions have transformed the entire market. But long before we had massive solid-state storage, we had a different method of improving system performance and response time: RAID. The venerable storage system turned 30 yesterday and it’s still widely used in enterprise and server solutions.
The people who invented RAID, David Patterson, Garth Gibson, and Randy Katz, first defined the concept in 1987, followed by a formal paper in 1988. Specific concepts like mirroring had been defined a decade previously, but RAID didn’t just allow for a mirrored solution. It also provided a blueprint for dramatically accelerating disk performance by striping data across multiple drives at the same time.
Patterson, Gibson, and Katz defined a variety of RAID levels to suit various tasks and strategies. Today, RAID 2 and RAID 3 are rarely used, while motherboards typically offer software RAID support for RAID 0, RAID 1, and RAID 10. Some boards and controllers also support RAID 5. RAID 0 is data-striping only–you get the advantage of writing data to two disks at once (boosting performance), but you increase the chance of catastrophically losing data. If each drive has a 1 percent chance of failure (just as an example), then the chances of losing a drive and all of your data is 4 percent, since any failure will kill the array. RAID 1 is mirroring–all of the data on Drive 0 is simultaneously written to Drive 1. This protects data, but offers no performance improvements. RAID 10 (0+1) combines the two methods, as shown below:
RAID offered a way for small disks to match or exceed the standard mainframe disks of the same era. IBM’s 3380 offered a whopping 7.5GB of storage in 1987, but a price tag in the neighborhood of six figures could give you a spontaneous ulcer. The entire idea behind RAID was to develop a system storage array that could match IBM’s performance and reliability for vastly less money.
Why RAID Mattered
But RAID didn’t just offer performance improvements at the enterprise level, it had real benefits for consumers as well. As the graph below from StorageReview.com shows, a RAID 0 array built from Western Digital’s VelociRaptor 600 family could even edge out an SSD from the same time period.
RAID arrays didn’t always improve performance, and there were definite drawbacks. AMD users were often badly bitten by a bug in VIA’s 686b southbridge that could cause critical and irreversible data loss when the system also contained a Creative soundcard. But at the time, these drawbacks were often considered an acceptable risk. CPU clock speeds had been leaping ahead for decades, while storage performance grew at a much slower rate. When combine with other methods of boosting performance, like short-stroking the drive (using only the inner area), RAID arrays offered the best performance you could buy.
Hat tip to ITProportal for recognizing the anniversary.
Now read: How Do SSDs Work?