Chinaunix首页 | 论坛 | 博客
  • 博客访问: 1755204
  • 博文数量: 335
  • 博客积分: 4690
  • 博客等级: 上校
  • 技术积分: 4341
  • 用 户 组: 普通用户
  • 注册时间: 2010-05-08 21:38
个人简介

无聊之人--除了技术,还是技术,你懂得

文章分类

全部博文(335)

文章存档

2016年(29)

2015年(18)

2014年(7)

2013年(86)

2012年(90)

2011年(105)

分类: 服务器与存储

2013-03-07 13:22:21

Different types of RAID protection

These descriptions are based on the original RAID definitions from the Berkeley paper by Patterson, Gibson and Katz. RAID originally stood for Redundant Array of Inexpensive Disks, but the disk vendors did not like that, as it had cost implications. They changed it to mean Redundant Array of Independent Disks.

Now this page has turned out to be a lot more popular that I ever thought it would, and needs a bit more explanation, as a lot of people are coming in from the home PC angle. I'm from a big systems background, IBM mainframes, big Unix servers, Windows and Netware clusters, that sort of stuff and that biases my opinions on RAID. If you want to put RAID onto your home PC, then in my opinion, RAID1 is the best way to go. It's simple. it works and it only needs two disks. It will even perform if it is a software implementation. 
If you run big storage systems with gigabytes of cache and hundreds of physical disks, then I would definitely go for RAID5. Why? It is cheaper because it uses fewer disks for a given capacity and it performs just as good as RAID1. If you have eighty 500GB disks, you can only store 20 Terabytes of data on them with RAID1, but you will get 35 TB on them in a 7+1 RAID5 implementation. That's why I claim that RAID5 is cheaper than RAID1. It is for big systems, but not for small systems, say less than a couple of terabytes. 
I had an animated discussion (which is one way of describing it) with a DBA last year who insisted that Oracle databases had to have RAID1 or they would not perform. We bought a DMX and ran some tests with the same database on RAID1 and RAID5, and the RAID5 setup actually performed better, I suspect, because it was pulling the data off more spindles. 
However, I would never touch a software implementation of RAID5 as the write penalty will kill performance. 
So there you go, PCs and small systems; RAID1, big systems RAID5 but at the end of the day it is your money.

RAID can be implemented by software in the host, but this is not usually successful. It is best implemented by microcode in the storage subsystem controller. The various types of RAID are explained below. In the diagrams, the square box represents the controller and the cache.

Parity is a means of adding extra data, so that if one of the bits of data is deleted, it can be recreated from the parity. For example, suppose a binary halfword consists of the bits 1011. The total number of '1's in the halfword is odd, so we make the parity bit a 1. The halfword then becomes 10111. Suppose the third bit is lost, the halfword is then 10?11. We know from the last bit that there should be an odd number of '1's, the number of recognisable '1's is even, so the missing but must be a '1'. This is a very simplistic explanation, in practice, disk parity is calculated on blocks of data using XOR hardware functions. The advantage of parity is that it is possible to recover data from errors. The disadvantage is that more storage space is required.

  • RAID0 is simply data striped over several disks. This gives a performance advantage, as it is possible to read parts of a file in parallel. However not only is there no data protection, it is actually less reliable than a single disk, as all the data is lost if a single disk in the array stripe fails.

    RAID0 principles

  • RAID1 is data mirroring. Two copies of the data are held on two physical disks, and the data is always identical. RAID1 has a performance advantage, as reads can come from either disk, and is simple to implement. However, it is expensive, as twice as many disks are needed to store the data.

    RAID1 principles

  • RAID2 is a theoretical entity. It stripes data at bit level across an array of disks, then writes check bytes to other disks in the array. The check bytes are calculated using a Hamming code. Theoretical performance is very high, but it would be so expensive to implement that no-one uses it.
  • RAID3 A block of data is striped over an array of disks, then parity data is written to a dedicated parity disk. Successful implementations usually require that all the disks have synchronised rotation. RAID3 is very effective for large sequential data, such as satellite imagery and video.

    RAID3 principles

    In the gif above, the right hand disk is dedicated parity, the other three disks are data disks.

  • RAID4 data is written in blocks onto the data disks (i.e. not striped), then parity is generated and written to a dedicated parity disk.

    RAID4 principles

    In the gif above, the right hand disk is dedicated parity, the other three disks are data disks.

  • RAID5 data is written in blocks onto data disks, and parity is generated and rotated around the data disks. Good general performance, and reasonably cheap to implement. Used extensively for general data.

    RAID5 principles

    The gif below illustrates the RAID5 write overhead. If a block of data on a RAID5 disk is updated, then all the unchanged data blocks from the RAID stripe have to be read back from the disks, then new parity calculated before the new data block and new parity block can be written out. This means that a RAID5 write operation requires 4 IOs. The performance impact is usually masked by a large subsystem cache. 
    As Nat Makarevitch pointed out, more efficient RAID-5 implementations hang on to the original data and use that to generate the parity according to the formula new-parity = old-data XOR new-data XOR old-parity. If the old data block is retained in cache, and it often is, then this just requires one extra IO to fetch the old parity. Worst case it will require to read two extra data blocks, not four.

    RAID5 write overhead

    RAID 5 often gets a bad press, due to potential data loss on hardware errors and poor performance on random writes. Some database manufactures will positively tell you to avoid RAID5. The truth is, it depends on the implementation. Avoid software implemented RAID5, it will not perform. RAID5 on smaller subsystems will not perform unless the subsystem has a large amount of cache. However, RAID5 is fine on enterprise class subsystems like the EMC DMX, the HDS USP or the IBM DDS devices. They all have large, gigabyte size caches and force all write IOs to be written to cache, thus guaranteeing performance and data integrity.

    Most manufactures will let you have some control over the RAID5 configuration now. You can select your block stripe size and the number of volumes in an array group. 
    A smaller stripe size is more efficient for a heavy random write workload, while a larger blocksize works better for sequential writes. A smaller number of disks in an array will perform better, but has a bigger parity bit overhead. Typical configurations are 3+1 (25% parity) and 7+1 (12.5% parity).
    ref:

阅读(1545) | 评论(0) | 转发(2) |
给主人留下些什么吧!~~