Wednesday, June 26, 2013

RAID System in SQL Server

RAID systems are arrays of disk drives that provide fault tolerance, more storage capacity, and better performance for the disk subsystem, depending on the configuration. Although RAID hardware systems are not part of the SQL Server configuration, they directly affect SQL Server performance. There are a variety of RAID levels each of which uses a different algorithm for fault tolerance. The most common RAID levels used with SQL Server are 0, 1, 5 and 10.
  • RAID 0 is also known as disk striping because it creates a disk file system called a stripe set. RAID 0 gives the best performance for read and writes operation because it spreads these operations across all the disks in the set. However, RAID 0 does not provides fault tolerance; if one disk fails, you lose access all the date on the stripe set.
  • RAID 1 also known as disk mirroring provides a redundant copy of the selected disk. RAID 1 improves read performance but can degrade the performance of write operations.
  • RAID 5 most popular RAID level, stripes the data across the disks of the RAID set as does RAID 0 , but it also adds parity information to provide fault tolerance. Parity information is distributed among all the disks. RAID 5 provides better performance than RAID1. However, when a disk fails, read performance decreases.
  • RAID 10 or RAID 1+0 , includes both striping without parity ad mirroring RAID 10 offer better availability and performance than RAID 5, especially for write-intensive applications.
  • The RAID configuration that is best for your database files depends on several factors, including performance and recover ability needs. RAID 10 is the recommended RAID system for transaction log, data and index files. If you have budget restrictions, keep transaction log files in a RAID 10 system, and store data and index files in a RAID 5 system.

No comments:

Post a Comment