Improved Performance and Fault Tolerance with RAIDs

You'll often want to give important data increased protection from drive failures. To do this, you can use RAID technology to add fault tolerance to your file systems. With RAID you increase data integrity and availability by creating redundant copies of the data. You can also use RAID to improve your disks' performance.

Different implementations of RAID technology are available. These implementations are described in terms of levels. Currently, RAID levels 0 to 5 are defined. Each RAID level offers different features. Windows Server 2003 supports RAID levels 0, 1, and 5.

• You can use RAID 0 to improve the performance of your drives .

• You use RAID 1 and 5 to provide fault tolerance for data.

Table 12-2 provides a brief overview of the supported RAID levels. This support is completely software-based.

Table 12-2. Windows Server 2003 Support for RAID

RAID Level	RAID Type	Description	Major Advantages
	Disk striping	Two or more volumes , each on a separate drive, are configured as a striped set. Data is broken into blocks, called stripes , and then written sequentially to all drives in the striped set.	Speed/performance.
1	Disk mirroring	Two volumes on two drives are configured identically. Data is written to both drives. If one drive fails, there's no data loss because the other drive contains the data. (Doesn't include disk striping.)	Redundancy. Better write performance than disk striping with parity.
5	Disk striping with parity	Uses three or more volumes, each on a separate drive, to create a striped set with parity error checking. In the case of failure, data can be recovered.	Fault tolerance with less overhead than mirroring. Better read performance than disk mirroring.

The most common RAID levels in use on servers running Windows Server 2003 are level 1 disk mirroring and level 5 disk striping with parity. Disk mirroring is the least expensive way to increase data protection with redundancy. Here, you use two identically sized volumes on two different drives to create a redundant data set. If one of the drives fails, you can still obtain the data from the other drive.

On the other hand, disk striping with parity requires more disks ”a minimum of three ”but offers fault tolerance with less overhead than disk mirroring. If any of the drives fail, you can recover the data by combining blocks of data on the remaining disks with a parity record. Parity is a method of error checking that uses an exclusive OR operation to create a checksum for each block of data written to the disk. This checksum is used to recover data in case of failure.

Real World

Although it's true that the upfront costs for mirroring should be less than the upfront costs for disk striping with parity, the actual cost per MB might be higher with disk mirroring. With disk mirroring, you have an overhead of 50 percent. For example, if you mirror two 36-GB drives (a total storage space of 72 GB), the usable space is only 36 GB. With disk striping with parity, on the other hand, you have an overhead of around 33 percent. For example, if you create a RAID-5 set using three 36-GB drives (a total storage space of 108 GB), the usable space (with one-third lost for overhead) is 72 GB.