Storage

Updated on August 25, 2007

Data is typically stored on electromechanical hard drives. While they are much faster and more reliable than the first versions from the 1970s, they are still mechanical devices and as such, they are the slowest and least reliable component of a server.

SCSI vs ATA

The two popular interfaces between PC motherboards and drives are ATA and SCSI. ATA (previously called IDE) has been the standard PC drive interface for over 20 years, but SCSI has been favored for servers. At first, this was because SCSI had DMA and ATA did not. DMA allows a drive controller to do read and write operations independently of the CPU, freeing it up to perform other tasks. Since modern ATA drives (such as SATA and Ultra-DMA) now have their own DMA controllers, this isn't an issue anymore. While manufacturers still build top-of-the-line hard drives with SCSI interfaces aimed at the server market, most servers will perform just fine with cheaper ATA drives.

SMART

SMART stands for Self-Monitoring, Analysis, and Reporting Technology. SMART capable drives, drive controllers and software can analyze a drive and report potential problems. SMART is not foolproof, however. Drives can fail at any time without warning, so you should always be prepared for drive failures.

RAIDs

One approach to overcoming the limitations of drives is to combine several to work as a single drive. The theory is that if a single drive fails, then the array of drives can keep working and you can replace the failed component at your leisure. This configuration is called a RAID (for Redundant Array of Independent Drives). A RAID can also improve I/O performance by striping read and write operations over several drives.

There are many possible RAID configurations. The simplest is RAID 1 and it comprises 2 drives that mirror each other. If one drive fails, then the other continues to provide data and the server continues operating. Some RAID setups will even allow you to replace the failed drive without shutting down the server (hot swapping).

Backups

No matter what storage solution you decide to use, it is extremely important that you backup your data on a regular basis. It is also critical that you backup your servers with a variety of methods. You don't often get a chance to test your backups. You don't want to find out the hard way that your only backup solution doesn't work.

Traditionally, hard drives were expensive and tape was cheap, so you backed up your drives onto tape. Hard drives have dropped in price so much, that they're a very practical backup solution. An extra drive on your server makes for a very fast and convenient backup device.

While an onboard backup device is convenient, an external backup solution is essential for recovery from catastrophic failures. The easiest way to do external or offsite backups is to backup to another server.

By setting up a second server to mirror your primary, you could create a RAIS (Redundant Array of Independent Servers). In the event that you primary server fails, you could configure your router or name server to redirect traffic to your backup server or you could reconfigure your backup server to replace your primary server.

List of backup solutions on Wikipedia

Further resources on storage devices

Categories
Hardware

Please give us your feedback.