RAID principle and classification

A. Raid principle

With the rapid development of computer and network technology, the requirements for storage performance and data reliability are increasing, and individual disk devices are not able to meet the requirements of speed and security.

RAID (standalone disk array device): multiple independent physical hard disks are combined in different ways to form a virtual hard disk

How does RAID implement a disk mix?

1. If you want a lot of disks to be combined, how do you manage these disks in a unified way?

Such as:

650) this.width=650; "title=" raid principle. png "src=" http://s5.51cto.com/wyfs02/M02/79/1A/wKiom1aI98uixuvWAAA45_ V85zq464.png "alt=" Wkiom1ai98uixuvwaaa45_v85zq464.png "/>

In the diagram, the controller on the motherboard is connected to the control chip, each control chip is connected to a different disk under the control chip to manage these combined disks, so the CPU in the communication with the disk through the control chip, the control chip to the CPU want to talk to the hard disk to understand, to achieve the CPU and hard disk communication

2. We all know that the data on the hard disk is stored as a sector, so the data is scattered on different disks?

Stripe: Divides the whole part of the data into different parts, which are stored on separate disks: if the data is scattered on different disks

650) this.width=650; "title=" striped. png "src=" http://s4.51cto.com/wyfs02/M01/79/18/wKioL1aI-tDz7sYZAABL__ 81jpe696.png "alt=" Wkiol1ai-tdz7syzaabl__81jpe696.png "/>

3. As shown above data stripe after the complete data is scattered on different disks, then there is a problem, if a disk is broken, then the entire data will be lost, the risk of data loss than using a disk greatly increased, how to do?

There are two ways of doing this:

Disk Mirroring technology:

The complete information is stored on each disk:

650) this.width=650; "title=" Disk mirroring technology. PNG "src=" http://s2.51cto.com/wyfs02/M00/79/18/wKioL1aI--_ Japtmaaboquqhnue379.png "alt=" Wkiol1ai--_japtmaaboquqhnue379.png "/>

This approach ensures the security of the data, but it's not hard to see how much disk space is wasted. So how do you mirror to improve utilization? Then there are the following improvement methods

Mode 1:

650) this.width=650; "title=" Mirror 01.png "src=" http://s4.51cto.com/wyfs02/M01/79/1A/wKiom1aI_ Liiubcxaabvezhj6iy680.png "alt=" Wkiom1ai_liiubcxaabvezhj6iy680.png "/>

This is the combination of RAID01: (first stripe and then mirror), you can see whether it is a mirrored group or a disk group of data storage one of the groups of disk corruption is not lost, but because it is not known that the data of each disk is mirrored to that disk, if both groups are bad one, then the data is lost , the entire disk group is hung up, and reaching therefore has the following mirroring method

Mode 2: One-to-one mirroring

650) this.width=650; "title=" Mirror Raid10.png "src=" http://s3.51cto.com/wyfs02/M00/79/18/wKioL1aI_ W2cu3gvaabbp-rbdzq984.png "alt=" Wkiol1ai_w2cu3gvaabbp-rbdzq984.png "/>

The combination of the above is a combination of RAID10, in this way can allow two disks in different groups at the same time, the data fault tolerance increased

Check code technology: Use a disk specifically as a calibration dial

650) this.width=650; "Title=" Check code. png "Src=" http://s2.51cto.com/wyfs02/M01/79/1A/wKiom1aI_ Bxx5eeeaabpgud3csc573.png "alt=" Wkiom1ai_bxx5eeeaabpgud3csc573.png "/>

Check code is the combination of the use of RAID5, check code in any of the disk error can be launched on other disk data, such as: Disk 1 Storage data 1, disk 2 storage data 2, disk 3 storage data 3, disk 4 in the check code 1+2+3, if the disk 2 bad data loss, We can also roll out the data in Disk 2 based on other disks, but the data will be lost if two disks are broken at the same time.

By the above check code way to store data, every access to other disk data can access the check code disk, then check the code disk becomes a performance bottleneck then how to do:

Workaround: Turn all disks into a check disc

JBOD: Combine different small disks and use them in Hadoop

650) this.width=650; "title=" Jbod.png "src=" http://s5.51cto.com/wyfs02/M02/79/19/wKioL1aJAP_ Akjcoaaaaeoxpjpm225.png "alt=" Wkiol1ajap_akjcoaaaaeoxpjpm225.png "/>

RAID level:

Note: Levels are not prioritized, but are not grouped in the same way

Raid:

Level: only means the disk organization is different, there is no upper and lower points;
0: Strip
Performance improvement: Read, write
Redundancy capability (fault tolerance): none
Space utilization: NS
At least 2 pieces of disk
1: Mirror
Performance: Write performance degradation, read performance improvement
Redundancy capability: Available
Space utilization: 1/2
At least 2 pieces of disk
5:
Performance: read, write promotion
Redundancy capability: Available
Space utilization: (n-1)/n
Need at least 3 bucks.
10:
Performance: Read and write promotion
Redundancy capability: Available
Space utilization: 1/2
Need at least 4 bucks.
01:
Performance: Read and write promotion
Redundancy capability: Available
Space utilization: 1/2
Need at least 4 bucks.
50:
Performance: Read and write promotion
Redundancy capability: Available
Space utilization: (n-2)/n
Need at least 6 bucks.
Jbod
Performance: No lift
Redundancy capability: None
Space utilization: 100%
Need at least 2 bucks.

RAID principles and classifications