RAID is an abbreviation for the redundant array of separate disks. This term is the most used in the papers of the array configuration and application of the University of California, Berkeley, California.
In the past, computer systems were often limited to write information to a single disk. This disk is usually expensive and it is very easy to fail. The hard drive has always been the most vulnerable link in a computer system because these devices are unique mechanical components in systems that are fully electronized in other components. The disk drive contains many movable mechanical parts running at high speed. The problem is not whether the hard drive will malfunction, but when a fault occurs.
RAID is intended to thoroughly change the computer management and access to data from data in a large-capacity memory by providing an inexpensive and redundant disk system. It has been called a cheap disk redundant array (RAID). RAID writes the data into multiple cheap disks instead of writing a single large capacity expensive (SIED). Initially RAID represents a redundant array of cheap disk, but now it has changed to a separate disk redundant array.
RAID basic principle
RAID achieves its redundant and fault tolerance. Targets by stripping storage and parity. Strip storage means writing files into multiple disks in a manner that writes a data block at a time. Strip storage techniques write data to multiple drives, increasing data transfer rates and shortening the total disk processing total time. This system is ideal for transaction processing, but reliability is very poor because the reliability of the system is equal to the reliability of the worst a single drive.
The parity checks the validity of the data by redundantly checking all data after transmission. With parity, other disks can rebuild the fault disk when a disk of the RAID system fails. In both cases, these functions are transparent to the operating system. Bringing storage and parity control by the Disk Array Controller (DAC).
Assembly
The main component of RAID is a disk array controller (DAC) and a queue consisting of 5 disks. The data is stored by the strip on all 5 disks, and the failed disk is recovered with parity. RAID has multiple different levels. Some RAID levels are used to increase speed, some of which are used to provide protection, while RAID-5 combines two advantages. We will discuss them one by one.
Strip storage data
Previously, the computer only writes the file into a disk. Strip storage allows you to split files and write different snipps simultaneously to multiple disks. If your file has 5 blocks and stores them into 5 disks, each data block will be written to their respective disks. If you have 5 OLTP transactions, each less than a data block, you can handle 5 different deals at the same time.
Most RAID levels are stripped in the data block level, but RAID can also strip storage in position or byte level. The size of the data block is determined by the system administrator and is called the depth of the baseband bar.
In order to maximize the transaction capability of the disk array subsystem, the data must be written to multiple drivers at the same time or read from multiple drives. To achieve this, the user data block is stored on the entire drive array by the strip. A baseband strip includes a column sector (512 bytes per sector) that are located on the same position on each disk in the array. The baseband strip depth (ie the number of sectors in each data block) is defined by the subsystem software.
The depth of the baseband strip has a direct impact of performance because the depth is too shallow, and the system execution needs more I / O commands than actual needs. If the specified depth is too large, the multitasking capabilities of the processor and the many benefits brought by the multi-drive may be offset.
In an ideal trading environment, each request from the host involves only one drive, which can implement multiple concurrent transactions of multiple drives.
Strip the data strip stored to the array drive to solve the problem of one system driver overloaded in the previously described and another idle problem. The data strip-cranked storage avoids the use of a dedicated drive and ensures the average allocation of the data processing load, and improves performance by simultaneously writting multiple data blocks simultaneously.
Parity
People often confuse parity and mirror (or images). The image involves making a copy of the disk. The image is technique for writing data to both drives. Because any of the two drivers can be completed, these systems have excellent reliability and achieve excellent transaction processing results. But the price is the capacity that must only be purchased for two drivers and only a drive. The mirror is expected to be 100%, or double disk space. If a disk is faulty, the mirror disk will be running. The parity provides the same general protection as mirror, but there is less overhead. If a user has an array consisting of 5 disks, four of which are used to store data and 1 for parity. Its overhead is only 20%, which is a big advantage when considering costs.
The computer only uses 0 and 1 to represent data. Different or (xor) is a method of making parity. A bit (0 and 1) is removed from each disk and add it. If the same is an even number, the parity is set to 0; if the odd number, the parity bit is set to 1.
Depending on the RAID level, the parity can be saved on a disk and can be assigned to all disks. When you use 5 disks, each way the parity checks up 1/5 or 20% of the disk space. When using 3 disks, 1/3 or 33%.
RAID configuration level
At present, the industry is recognized with 6 RAID configuration levels, and they are specified as RAID0 to RAID5. Each RAID level is designed for speed, protection or binding. The RAID level includes:
RAID 0 - data stripping storage array
RAID 1 - Mirror Disk Array
RAID 2 - Parallel array, Hamming code
RAID 3 - Parallel arrays with parity
RAID 4 - Disk array with dedicated parity drive
RAID 5 - Disk array, all drives include parity checks
The most common RAID level is RAID-0, RAID-2 and RAID-5. The following is a detailed description
RAID-0 data stripping storage array
RAID-0 Store data strip to all drives, but does not use parity. If one of the disks fails, the data must be resembled from the backup to all 5 disks. This RAID is intended to increase speed, the fastest speed in all RAIDs, but offers the least protection.
RAID-1 transparent or strip storage mirror
RAID-1 technology requires that each raw data disk has a mirror disk. The original disk and the image of the image are exactly the same. RAID-1 provides the best data protection, but the speed is not as good as RAID0 and 5.
Also written on the original disk on the original disk is also written to the mirror disk. This mirroring process is invisible for users. Therefore, RAID-1 is also known as a transparent image. The user can set RAID-1 to write the data to one disk and mirror the disk; or can also store it on a plurality of disks, each strip-cracked disk has a mirror copy. This is called strip storage mirror, RAID1 0, RAID10, which is also referred to as RAID6 in some cases.
RAID-5 disk array, parity quilt
RAID-5 uses data to strip storage in the form of data blocks, while also using parity. With RAID-5 technologies, user information and parity can be merged into each disk in the array. Independent and / or parallel data read / write operations can be performed. This RAID is the most widely used in all RAIDs. RAID-5 does not have RAID-0 so fast, nor does it provide much protection like RAID-1 mirroring. But RAID-5 provides good speed and protection. This is why it tends to be a RAID level selected by people.
RAID disk array components
The main components of the RAID disk array are disk array controllers, 5 SCSI channels, and one or more disk array columns. There is usually two disk array controllers (DACs) work as a group. This implementation in the past is often fails by an active DAC and a passive DAC, and the other will control all of the disk columns. There are two DACs in the figure below to control 4 disk column together. You can configure the disk into any supported RAID level. You can even disrupt disks and configure multiple RAIDs in a column. Built-in / external disk array
Previously disk arrays are dedicated to the host through a cable and are always placed in an external chassis. The SCSI length of the external disk array is limited to about 80 inches or 25 meters. It can be extended by 25 meters using a repeater, but this will lose 5% performance.
Many computers currently use built-in RAID. The CPU communicates with the disk, but the basic principles are still the same. Whether it is built-in or external, disk arrays have one or two disk array controller-controlled disk queues. Keep in mind points
RAID is used to improve data performance, reliability, and availability.
Strip storage, parity and mirror are three main functions of the RAID system.
The function of RAID is transparent to the operating system.
The system is typically defined by a disk column, and each disk queue includes 5 disks and connects to one or two disk array controller.
Different grade RAIDs provide different speeds and different degrees of data protection.
