Do Not Rely on RAID

---

Realizing the shortcomings of backing up data, enterprises have turned to using Redundant Arrays of Independent Disks (RAID) on their servers. RAID has two main purposes: to speed up data flow by simultaneously reading from or writing to more than one disk at a time (disk striping) and to preserve data by replicating it on a number of disks so that if one goes down the others will still be accessible. RAID can be achieved by utilizing either a hardware or an operating system approach. Software RAID has the advantage of being less expensive. Hardware RAID, on the other hand, is more expensive but faster. Because a separate RAID controller is doing the data replication and distribution, hardware RAID does not tie up the operating system and impact performance. (RAID is covered in more detail in Chapter 1.)

Although RAID is a significant improvement over relying solely on backups, it has its drawbacks. To begin with, it is expensive. Cost precludes its use in low end servers. Second, while RAID does prevent data loss, it does not prevent service loss. When a disk goes down in a RAID array, the system may keep running, but the faulty disk must still be replaced. During the hours it takes to add a new one and transfer data to it, performance takes a huge hit. Although one's employees are not likely to walk off the job during this time (they will just flood the help desk with calls), where customers are concerned it is a different matter. When service is slow on a company's Web site, its customers may very well click over to the competition. Numerous studies highlight the dangers of having a slow Web site and what it means in terms of losing out to the competition. The average consumer will put up with a delay of no more than eight seconds before deciding to leave a site. Such an encounter often leads to buying from the competition, never to return.

---