Who are you choose (on) RAID level
Which of RAID 012345 is suitable for you, not just cost problems, fault-tolerance and transmission performance considerations, and future expansion should comply with the needs of the application.
RAID's application in the market is not a fresh thing, many people have a little about RAID's basic concepts, as well as distinguish between different RAID Level. But in the actual application, we have found that there are many users who choose a suitable RAID Level, which is still unable to master, especially for RAID 0 1 (10), RAID 3, RAID 5 choice. It is even more lifting.
This article will be aimed at the working principle and characteristics of RAID 0 1/10, RAID 3, and RAID 5 to list applications that are suitable for these different RAID classes, and I hope that you can have principled help.
RAID strip "Striped" access mode
In the RAID system of the data string Data Stripping, it can be divided into two ways to the member disk drive:
Parallel access ﹝ Paralleled Access ﹞
Independent access ﹝ Independent Access ﹞
RAID 2 and RAID 3 are taken in parallel access mode.
RAID 0, RAID 4, RAID 5 and RAID 6 are independent access modes.
Parallel access mode
In parallel access mode, it is a precise control of all the spindle motor of all disk drives, so that each disk is synchronized with each other, and then makes a very short I / O data transmission for each disk drive, so Each I / O instruction from the host is all distributed to each disk drive.
In order to achieve parallel access, each of the RAID must have almost identical specifications: the speed must be the same; the magnetic head search speed ﹝ Access Time must be the same; Buffer or Cache's capacity and access speed should be consistent The speed of the CPU processing instruction is the same; the speed of I / O Channel is also the same. All in all, you should use parallel access mode, all member disk drives in RAID should use the same label, the same model of disk drive.
Basic working principle of parallel access
Suppose there are four disk drives of four identical specifications in RAID, which are disk drives A, B, C and D, and we are divided into T0, T1, T2, T3, and T4, respectively.
T0: RAID Controller Transmits the first data to a buffer, disk drive B, C and D Buffer are empty, waiting
T1: RAID Controller Transmits the second data to B buffer, and a begins writing data in the buffer into a sector, and the buffer of disk drive C and D is empty, waiting in the waiting
T2: RAID Controller Transmits the third data to C's buffer, b begins writing data in the buffer, A has completed the write action, the buffer of disk drive D and A is empty, waiting in the waiting
T3: RAID controller Transmits the fourth data to D's buffer, c Start writing data in buffer into sectors, b has completed writing actions, and the buffer of disk drives A and B is empty, waiting in the waiting
T4: RAID Controller Transfer the fifth data to a buffer, d Start writing data in the buffer into a sector, C has completed the write action, the buffer of disk drives B and C are empty, waiting in the waiting
So has been looping, until the processing from the host is complete, the RAID controller will be processed the next I / O instruction. The focus is that when any disk drive is ready to write the data into a sector, the destination sector must just turn to the head. At the same time, the RAID controller is only transmitted to a disk drive data length, but also must just be good, with the speed of the disk drive, otherwise the Miss, RAID performance is greatly reduced. Best application for parallel access RAID
The architecture of the RAID is used in parallel, with its fine motor control and distribution data transmission, the performance of each of the arrays per disk drive is maximized, while taking advantage of the frequency width of the Storage Bus, it is especially suitable for applications in large, data continuous Archive access application, for example:
Image, video archive server
Data warehousing system
Multimedia database
E-library
Pre-print or negative output file server
Other large and continuous file servers
Due to the characteristics of the RAID architecture, the RAID controller can only handle an I / O requirement at a time, unable to perform the multitasking of overlapping, so it is very uncomfortable to use frequent I / O times, data is random access, each data transmission A small environment. At the same time, because parallel access is unable to execute overlapping multitasking, there is no way to "hide" disk drive to search for the time of SEEK ﹞, and wait for the first disk drive at each I / O first data transmission. Rotation Delay ﹝ Rotational Latency ﹞, averages the time of rotating a half circle, if a 10,000-turn disk drive is used, you will need to wait 50 USEC. So mechanical delay time is the biggest problem of parallel access architecture.
Independent access mode
Compared with the parallel access mode, the independent access mode does not rotate synchronously with the member disk drive, and its access to each disk drive is independent and has no order and time-sensitive limits, while each transferred data. The amount is relatively large. Therefore, the independent access mode can use the overlapping multitasking, tagged command queuing, etc., "Hide" The Machine Time Delay of the above-mentioned disk drive ﹝ Seek and Rotational Latency ﹞.
Since the stand-alone access mode can make Overlapping multitasking, it can handle different I / O Requests from multiple hosts at the same time, and can play the greatest performance in multi-host environments such as Clustering ﹞.
Best application of independent access RAID
Since the stand-alone access mode can accept multiple I / O Requests simultaneously, it is particularly suitable for systems that are frequent in data access, and have a smaller amount of data. E.g:
Online trading system or e-commerce application
Multi-user database
ERM and MRP system
Small file file server