Data Recovery Resource
RAID Recovery Glossary of Terms
Two or more disks grouped together to appear as a single disk to the host system.
A bus-based (usually PCI) hardware device -- such as an add-in card, group of motherboard ASICs, or a combination of both -- that converts the timing and protocol of a host's memory bus and an I/O bus. Usually used in entry-level servers, an array adapter also includes an on-board RAID co-processor to offload most of the RAID operations -- for example, secondary RAID 1 writes and RAID 5 parity calculations -- from the host CPU. This is in contrast to the microprocessor-based array controllers used in midrange and high-end servers, which also offload I/O commands. Array adapters improve performance over software RAID solutions embedded within network operating systems such as NetWare and Windows NT. These adapters provide the same connectivity functions as a standard host adapter.
Short for AT Attachment. A hard drive with an integrated controller. There are multiple levels of ATA standards including the base-level 16-bit IDE, ATA-2 (Enhanced IDE), Ultra ATA (ATA33), ATA66 and ATA100. A good explanation and tutorial is available at PC Guide.
ATA Packet Interface. Defines a set of commands supported through the ATA-2 interface for peripherals other than hard drives, such as CD-ROM, DVD-ROM, and tape drives.
An array which includes system disk files and allows a server to boot from the array while protecting the network operating system disk -- and other data on the array -- from drive failure.
Power must be switched off before the removal or insertion of a component.
Temporary storage of new write data or high-demand read data in solid state memory in order to accelerate performance. The cached data is later overwritten with newly cached data once it is either written to disk or deemed to be of low demand.
A performance caching technique in which the currently requested data is read in anticipation of another request before its allocated blocks are recycled. Reassignment of the blocks is done on the basis of least recently used (LRU).
Spreading data evenly over multiple disks to enhance performance. Sometimes referred to as RAID 0, data striping actually has no redundancy scheme and, therefore, does not provide any fault tolerance (data protection).
Synonym for disk, hard drive, hard disk, disk drive.
Mirroring across two host adapters. Used only with software-based RAID storage systems (usually the embedded network operating system RAID software such as NetWare and Windows NT).
ECC (Error Correction Code)
Refers to parity error detection and correction within memory or cache (for example, SIMM). Depending on the ECC SIMM, single-bit or double-bit parity errors can be detected but not corrected, or detected and corrected automatically.
Exclusive OR (XOR)
A process based on a mathematical algorithm that is used by RAID levels 2, 3, 4, and 5 to compare computer data (binary 0s and 1s) created by a write request or by a read request during a drive failure. The result of the XOR process is parity information that will be stored along with data for real-time recovery in the event of a disk failure.
External array controller
In contrast to bus-based array adapters and microprocessor-based array controllers, external array controllers reside in the external RAID storage enclosure. They connect to the host through a standard SCSI or serial (such as Fibre Channel) host adapter interface. These external controllers are similar to bus-based, microprocessor-based array controllers, in that they include an on-board microprocessor to offload all RAID functions (I/O commands and RAID operations) from the host CPU. They are usually used in midrange and high-end servers, especially in clustering environments.
A mode of reduced-performance operation that a disk array is in after a drive failure.
The automatic replacement of a failed system component with a properly functioning one. Most often used in the context of redundant external array controllers. If one of the controllers fails, failover enables the second controller to take over the failed controller's I/O load.
The ability of a system to continue to perform its functions, even when one or more components have failed.
A bus-based (PCI, EISA, ISA) hardware device, such as an add-in card or ASIC, that converts the timing and protocol of a host's memory bus and an I/O bus.
Hot spare RAID storage feature that allows a spare drive (or other component) to be configured for automatic (in contrast to hot-swap) replacement and reconstruction in the event of a disk failure. Users can remain on-line and continue to access data.
Hot swap A storage system's ability to allow the removal and replacement of a disk drive (or other component) while users are on-line and accessing data. In contrast to hot spare, this is a manual operation. Hot swap requires that the storage (or server) enclosure drive tray connectors be designed so that when a drive is removed, power is disconnected before the ground connection, and that the ground is restored before the power is reconnected upon reinsertion of the drive. This is usually accomplished by making the ground pin(s) in the drive tray connector slightly longer than the data pins.
JBOD Just a bunch of drives. Refers to an array of drives without data redundancy.
MTBF Mean time between failure. Used to measure computer component average reliability/life expectancy. MTBF is not as well-suited for measuring the reliability of array storage systems as MTDL, MTTR or MTDA (see below) because it does not account for an array's ability to recover from a drive failure. In addition, enhanced enclosure environments used with arrays to increase uptime can further limit the applicability of MTBF ratings for array solutions.
MTDA Mean time between data access (or availability). The average time before non-redundant components fail, causing data inaccessibility without loss or corruption.
MTDL Mean time to data loss. The average time before the failure of an array component causes data to be lost or corrupted.
MTTR Mean time to repair. The average time required to bring an array storage subsystem back to full fault tolerance.
Member (disk) A disk that is in use as a member of a disk array.
Mirroring Also known as RAID 1 or duplexing (when using two host bus adapters). Full redundancy is obtained by duplicating all data from a primary disk on a secondary disk. The overhead of requiring 100 percent data duplication can get costly when using more than two drives.
PCI - Peripheral Component Interconnect Industry-standard specification that refers to a high-speed (133 MB/sec) host bus commonly used for host adapters, Ethernet adapters, and video cards.
Parity A form of data redundancy used by RAID levels 2, 3, 4, and 5 to recreate the data of a failed drive in a disk array.
Pre-fetching Intelligent gathering of data from disks prior to requests from the operating system.
RAID Redundant array of inexpensive disks. The term coined in 1987 by researchers at the University of California at Berkeley to describe a series of redundant architectures used in fault-tolerant disk arrays.
RAID levels Numbered 0 through 5, RAID levels refer to different array architectures that offer various advantages in terms of data availability, cost, and performance.
Read-ahead cache A performance caching technique in which data is anticipated and read into the cache before it is actually requested.
Redundant A duplicate disk or component that provides a recovery path in case of a failure.
SCSI Small computer system interface (pronounced scuzzy). The fast, intelligent input/output parallel bus used by high-performance peripherals.
S.M.A.R.T. - Self-Monitoring, Analysis and Reporting Technology Drives equipped with this feature report predicted failures based on threshold values determined by the manufacturer. This allows the network manager to replace a drive before it fails.
Software-based array An array in which all management functions including parity calculation (XOR) are performed by the host server CPU. These products are low priced but have high CPU utilization and limited fault-tolerant features. High-performance, low-cost array adapters are quickly replacing these inferior software-based arrays.
System disk The disk (or array) on which a system's operating system is stored and from which it is initially loaded into system memory.
Usable storage capacity Disk array capacity that is usable for data storage (vs. for mirroring or parity data). For example, under mirroring (RAID 1 and 0/1), usable storage remains a constant fifty percent (half of storage is always used for redundancy). This is in contrast to other RAID levels such as RAID 5, in which usable storage capacity is determined by the formula of "n-1". "n" is the total number of disk drives and "1" is the number of disks worth of capacity used for parity (redundancy) overhead. So, as the number of disks in the array grows, the usable storage capacity percentage increases in relation to parity (redundancy) information.
Warm swap The ability to remove and replace a disk drive while the power is on. All bus activity must be paused (usually done through a utility within the array management software) to maintain data integrity during removal or replacement. Typically used when hot swap is not supported by the server or storage enclosure drive tray.
Write-back cache A performance caching technique in which the completion of a write request is signaled as soon as the data is in cache; Actual writing to the disk occurs at a later time. Since the operating system is "fooled" into thinking that the write has actually been written to disk, there is a risk of losing or corrupting data in cache should an error or power failure occur. Therefore, use of a battery-backed cache is recommended to prevent such data loss/ corruption.
Write-through cache A caching technique in which the completion of a write request is not signaled until data is safely stored on disk. Performance of write-through is essentially the same as in non-cached systems.
Courtesy of adaptec