RAID IN INDUSTRIAL COMPUTER SYTEMS
This white paper by CP Technologies discusses RAID. Redundant array of independent disks (RAID), a group of hard drives and one or more controller cards, and embedded software that increases reliability and redundancy in data storage on hard drives. Many RAID options offer better performance and/or greater data reliability. RAID-5, for example, does not represent the number of drives involved. RAID implementations that are most commonly used include 0, 1, and 5. You can implement RAID with or without hot-swapping a drive.
Virtually all of the major drive controller manufacturers, such as Adaptec or DPT, offer plug-in controllers that allow RAID implementation. These controllers interface with SCSI drives, and they are available in ISA or PCI configurations. The PCI format provides the best throughput. RAID support is not available for IDE drives.
A RAID controller can be used with any SCSI drive. Different drive sizes and throughputs may be used on the same bus. For more information, you should contact the controller manufacturer. RAID controllers can also be used as generic drive controllers to interface with CD and tape drives, as well as external accessories like scanners.
RAID drives can be mounted permanently in a chassis or removable 5-1/4″ carriers. External drive bays allow for quick access and replacement in case of drive failure. Even in situations with non-removable drives, full RAID protection is possible. The RAID system provides data security and allows for the removal of the drive to be replaced. This is a cheaper and more reliable alternative to expensive removable drive carriers. Kingston Technology Data Express or JMR Wildcat are good options for removable media.
RAID LEVEL DEFINITIONS
RAID 0 Striping
To improve access performance, the data is written to multiple drives. Data redundancy is not possible. A 4Meg file could be written on 4 drives in one-megabyte chunks. The data will be lost if one drive fails. This is because the throughput is significantly higher.
This provides 100% redundancy and a mirror or exact copy of the primary drive. If one drive fails, data on the secondary drive will still be accessible. Simultaneous reads will not improve performance. It is important to note that you must purchase twice as many drives. One controller can mirror across one bus, but two controllers are possible to provide both controller and drive redundancy.
Adaptec offers extensive online discussions on RAID controllers and controllers, particularly in their Array Guide.
RAID 10, or 0/1 Striping, and Mirroring
Combination of RAID 0, 1, and 2. For improved performance, data is divided across multiple drives. Each drive is also mirrored to ensure redundancy. It is important to note that you will need twice the number of drives.
Thinking Machines, Inc. has patented a proprietary array where data is divided on a bit level between several drives. Additional drives provide parity information. This requires large numbers of drives. It is not usually implemented.
RAID 3 Striping With Parity
Redundancy and improved performance. Data is shared between multiple drives. An additional drive provides parity information. Data stripping improves performance, but it requires simultaneous reads when the array is accessed. If one of the data drives fails, the drive with the parity information is used to rebuild the data. This drive is usually used with three data drives and one parity drive. Small random writes are usually slow because the parity drive must also be accessed for each one.
RAID4 Striping with Dedicated Parity Disk
Similar to RAID 3, but larger data blocks can be striped. Each access does not require participation from each drive. Each data access requires the parity drive to be accessed.
RAID 5 Striping and Parity
This is the most popular RAID implementation. The data and parity information are striped across multiple drives, with each drive holding both the data and the parity information. If one drive fails, all remaining drives will contain enough information to recover. This provides complete redundancy and improved performance. For the smallest RAID 5 implementation, three drives are required. However, more drives can be added to improve performance.
There is no real definition. Different vendors may use different terms.
Proprietary Storage Technology, Inc. is an alternative to RAID 4, with caching and a proprietary operating system to manage the array.
Hot Swapping is the ability to remove an array drive while it is powered up. To prevent data glitches, the signal pins and power connector pins must be longer than those on the drive tray. There are many removable drive carriers available. It is important to ensure that they can support hot-swapping and not removable media.
Warm Swapping can be used to block drive access until a drive is removed. This is usually a software function that suspends drive activity. This configuration can be used with a low-cost removable drive carrier that does not require a hot swap.
A hot spare is an array’s backup drive that automatically comes online in case of failure of any other drives. An array is usually able to tolerate one drive failure without data loss, so a hot spare drive decreases the chance of total failure.
SMART is a predictive failure analytics system that allows the drive to perform self-analysis and communicate any potential failures to its controller. This allows for the early replacement of potentially faulty drives before they fail.
Dynamic sector repair allows you to use a RAID system for finding faulty sectors on drives and transparently fixing the data. You can also flag bad sectors to block future access.