RAID (Redundant Array of Inexpensive or Independent Disks) is an array of multiple independent hard disk drives that provide high performance and fault tolerance. A RAID disk subsystem improves I/O performance over a computer using only a single drive. The RAID array appears to the host computer as a single storage unit or as multiple logical units. I/O is expedited because several disks can be accessed simultaneously. RAID systems improve data storage reliability and fault tolerance compared to single-drive computers. Data loss because of a disk drive failure can be recovered by reconstructing missing data from the remaining data and parity drives.

RAID has gained popularity because it improves I/O performance and increases storage subsystem reliability. RAID provides data security through fault tolerance and redundant data storage.

Although disk drive capabilities have improved drastically, actual performance has improved only three to four times in the last decade. Computing performance has improved over 50 times during the same time period. Implementing RAID improves the performance of the disk subsystem.

The electromechanical components of a disk subsystem operate more slowly, require more power, and generate more noise and vibration than electronic devices. These factors reduce the reliability of data stored on disks. Implementing RAID on disk subsystems improve the reliability of data.

RAID products are either:

• software-based
• host-based
• external

A host-based RAID solution is a PCI adapter card that is installed in any available PCI expansion slot in a host system.

A host-based RAID product puts all of the RAID intelligence on an adapter card that is installed in a network server. A host-based RAID product provides much better performance than a software-based solution which runs on the host CPU. Host-based RAID also is a much better price/performance solution than external RAID.

The available sequential data transfer rate is determined by the following factors:

• the sustained data transfer rate on the motherboard PCI bus
• the sustained data transfer rate on the i960RN PCI to PCI bridge
• the sustained data transfer rate of the SCSI controller
• the sustained data transfer rate of the SCSI devices
• the number of SCSI channels
• the number of SCSI disk drives
• Host-based solutions must provide operating system-specific drivers

An external RAID product puts the RAID intelligence inside the RAID chassis and uses a plain Host Bus Adapter installed in the network server. The data transfer rate is limited to the bandwidth of the HBA channel.

A SCSI-to-SCSI RAID product puts the RAID intelligence inside the RAID chassis and uses a plain SCSI Host Adapter installed in the network server. The data transfer rate is limited to the bandwidth of the SCSI channel. A SCSI-to-SCSI RAID product that has two wide SCSI channels that operate at speeds up to 80 MB/s must squeeze the data into a single wide SCSI (40 MB/s) channel back to the Host computer.

In SCSI-to-SCSI RAID products, the hard drive subsystem uses only a single SCSI ID, which allows you to connect multiple drive subsystems to a single SCSI controller.

RAID (Redundant Array of Independent Disks) is a collection of specifications that describe a system for ensuring the reliability and stability of data stored on large disk subsystems. A RAID system can be implemented in a number of different versions (or RAID Levels) 0, 1, 5 and 10 (or 1+0).

Consistency Check

In RAID, check consistency verifies the correctness of redundant data in an array. For example, in a system with parity, checking consistency means computing the parity of the data drives and comparing the results to the contents of the parity drive.

Fault Tolerance

Fault tolerance is achieved through cooling fans, power supplies, and the ability to hot swap drives. Most RAID controllers provide hot swapping through the hot spare feature. A hot spare drive is an unused online available drive that instantly can use to rebuild a drive when an active drive fails.

After the hot spare is automatically moved into the RAID subsystem, the failed drive is automatically rebuilt on the spare drive. The RAID disk array continues to handle request while the rebuild occurs.

Disk Striping

Disk striping writes data across multiple disk drives instead of just one disk drive. Disk striping involves partitioning each drive storage space into stripes that can vary in size from 2 kilobytes (KB) to 128 KB. These stripes are interleaved in a repeated sequential manner. The combined storage space is composed of stripes from each drive. Most controllers support stripe sizes of 2 KB, 4 KB, 8 KB, 16 KB, 32 KB, 64 KB, or 128 KB.

For example, in a four-disk system using only disk striping (as in RAID level 0), segment 1 is written to disk 1, segment 2 is written to disk 2, and so on. Disk striping enhances performance because multiple drives are accessed simultaneously; but disk striping does not provide data redundancy.

Stripe width is the number of disks involved in an array where striping is implemented. For example, a four-disk array with disk striping has a stripe width of four.

The stripe size is the length of the interleaved data segments that the RAID controller writes across multiple drives. Most RAID controllers support stripe sizes of 2 KB, 4 KB, 8 KB, 16 KB, 32 KB, 64 KB, or 128 KB.

Disk Mirroring

With mirroring (used in RAID 1), data written to one disk drive is simultaneously written to another disk drive. If one disk drive fails, the contents of the other disk drive can be used to run the system and reconstruct the failed drive. The primary advantage of disk mirroring is that it provides 100% data redundancy. Since the contents of the disk drive are completely written to a second drive, it does not matter if one of the drives fails. Both drives contain the same data at all times. Either drive can act as the operational drive.

Disk mirroring provides 100% redundancy, but is expensive because each drive in the system must be duplicated.

Parity

Parity generates a set of redundancy data from two or more parent data sets. The redundancy data can be used to reconstruct one of the parent data sets. Parity data does not fully duplicate the parent data sets. In RAID, this method is applied to entire drives or stripes across all disk drives in an array. The types of parity are:

If a single disk drive fails, it can be rebuilt from the parity and the data on the remaining drives.

RAID level 5 combines distributed parity with disk striping. Parity provides redundancy for one drive failure without duplicating the contents of entire disk drives, but parity generation can slow the write process. A dedicated parity scheme during normal read/write operations is shown below:

Hot Spares

A hot spare is an extra, unused disk drive that is part of the disk subsystem. It is usually in standby mode, ready for service if a drive fails. Hot spares permit you to replace failed drives without system shutdown or user intervention.

Most RAID controllers implement automatic and transparent rebuilds using hot spare drives, providing a high degree of fault tolerance and zero downtime. The RAID management software allows you to specify physical drives as hot spares. When a hot spare is needed, the RAID controller assigns the hot spare that has a capacity closest to and at least as great as that of the failed drive to take the place of the failed drive.

Disk Rebuild

You rebuild a disk drive by recreating the data that had been stored on the drive before the drive failed.

Rebuilding can be done only in arrays with data redundancy such as RAID level 1, 5, and 10.

A hot spare can be used to rebuild disk drives in RAID 1, 5 or 10 systems. If a hot spare is not available, the failed disk drive must be replaced with a new disk drive so that the data on the failed drive can be rebuilt. A rebuild will occur automatically when a drive is replaced (by hot-swapping it in the same drive bay.)

RAID controllers automatically and transparently rebuilds failed drives with user-definable rebuild rates. If a hot spare is available, the rebuild starts automatically when a drive fails. Most RAID controllers automatically restarts the system and the rebuild if the system goes down during a rebuild.

The rebuild rate is the fraction of the compute cycles dedicated to rebuilding failed drives. A rebuild rate of 100 percent means the system is totally dedicated to rebuilding the failed drive.

The rebuild rate can be configured between 0% and 100%. At 0%, the rebuild is done only if the system is not doing anything else. At 100%, the rebuild has a higher priority than any other system activity. The default rebuild rate is 30%.

A RAID array is a collection of physical disk drives governed by the RAID management software. A RAID array appears to the host computer as one or more logical drives.

Logical Drive

A logical drive is a partition in a physical array of disks that is made up of contiguous data segments on the physical disks. A logical drive consists of an entire physical array.

Hot Swap

A hot swap is the manual replacement of a defective physical disk unit while the computer is still running. When a new drive has been installed, a rebuild will occur automatically if it is placed in the same drive bay as the failed drive it is replacing.

SCSI Drive States

Logical Drive States

Disk Array Types