Server disk RAID0,RAID1,RAID10,RAID5,RAID50,RAID6,RAID60 analysis and case

1,RAID0 features: With split, the data will be split on several disks. Data is divided into chunks, and each chunk is written to a different disk. As a result, each disk's workload is reduced, which helps speed data transfer. RAID-0 makes the disk more responsive, especially for e-mail, database, and Internet applications. A minimum of two drives is required to implement RAID-0. Advantage: Improve system performance by distributing I/O load across multiple drives. Simple implementation. It is important to note that RAID-0 does not have data protection capabilities and is not suitable for critical data.

2,RAID1 Features: RAID-1 is implemented by disk mirroring and is primarily used to ensure data reliability. The same data will be replicated to different disks, and if a disk fails, the data can be found on one of the disks in the array, making it easy to recover. Mirroring not only creates redundant data and provides high availability, it also keeps critical applications up and running. Advantage: Data read performance is improved, and data write performance is no different than a single disk. 100% data redundancy means that there is no need to rebuild data in the event of a disk failure. It is important to note the inefficient use of disk capacity-the highest cost for all RAID types (100%).

3,RAID10 Features: RAID-10 is a combination of RAID-1 and RAID-0. This configuration requires at least 4 hard drives, with the best performance, protection, and capacity in all RAID levels. The RAID-10 contains paired mirrored disks whose data is stripped across the array. In most cases, the RAID-10 can withstand the failure of multiple disks, so that the system can be more properly run. Its data is least likely to be lost. Advantage: The same redundancy as RAID-1 (mirror) is the ideal choice for data protection. It is important to note that the price may be high, related to the mirrored disk array.

4,RAID5 Features: RAID-5 maintains data redundancy through a technique called parity. When data is stripped on multiple disks, the parity data is also included and distributed across all the disks in the array. Parity data is used to maintain the integrity of the data and rebuild when a disk fails. If one of the disks in the array fails, the lost data can be rebuilt based on the parity bit data on the other disks. RAID-5 configuration requires at least 3 hard drives. Advantage: More efficient use of disk capacity for all redundant RAID configurations. Maintain good read and write performance. It is important to note that a disk failure can affect the throughput rate. The time to reconstruct the information after the failure is longer than the mirror configuration.

5,RAID50 Features: RAID-50 is a combination of RAID-5 and RAID-0. This configuration makes the stripping of data, including parity information, on each disk of the RAID-5 sub-disk group. Each RAID-5 sub-disk group requires three hard disks. The RAID-50 has higher fault tolerance because it allows one disk in a group to fail without causing data loss. And because the odd and even bits are on the RAID-5 sub-disk group, the reconstruction speed is greatly improved. Advantage: Higher fault tolerance with the potential for faster data read rates. It is important to note that a disk failure can affect throughput. The time to reconstruct the information after the failure is longer than the mirror configuration.

Features of the 6,RAID6:

Performance of RAID6:

(1) RAID6 Random Read performance: Very good (when using large data blocks).

(2) RAID6 Random Write performance: poor, because not only to write the checksum on each hard disk and to write the data on the special check hard disk.

(3) RAID6 Continuous Read performance: Good (when small data blocks are used).

(4) Persistent write performance for RAID6: General.

(5) The advantages of RAID6: Fast read performance, higher fault tolerance.

(6) Disadvantages of RAID6: Slow write speeds, RAID controllers are more complex and cost-efficient to design.

Features of the 7.RAID60:

Higher fault tolerance, support for both HDD failure repair, and higher read performance. There are still some problems in technology, not mature enough, and few users at present.

RAID 50:4 disks make up a single set of RAID 5, then two groups of RAID 5 then compose RAID 0 and finally get 8 disk RAID 50

650) this.width=650; "src=" Http://image.lxway.com/upload/b/b2/bb2c4c61e1de392a2999718397eaa00a_thumb.jpg "alt=" Consider using RAID 50 to replace RAID ten (SSD article) "title=" consider using RAID 50 to replace raid ten (SSD) "style=" Border:0px;vertical-align:middle;height:auto; />

Case Analysis:

The fault-tolerant probability of the 1th disk is 100%, and the most basic ability of the disks array is fault tolerance, but the data protection ability of different level array can be different;

Starting with the 2nd disk except that RAID 6 provides 100% failure availability, other than RAID 10 offers the perfect solution. At the same time we can find the fault tolerance of RAID 5 is the worst of the four, but to achieve the same capacity is the minimum number of disks required, if the fault operation is more timely, RAID 5 is a cost-effective solution, Otherwise, after the first disk failure to the array repair period if a second disk failure will lead to the entire array failure (data loss), this is the risk;

RAID 6 is certainly a more reliable solution, but it has to sacrifice both disk capacity and poor performance (followed by performance test instructions), so the tradeoff between performance and availability;

Of course, the focus is on RAID 10 and 50: We found 50 the probability of the second disk failure and 10 is closer, because only two groups of RAID 5 in the above case can only provide fault tolerance to up to two disks, if the number of RAID 5 group more can fault tolerance of the number of disks will be more, and the probability of availability will be higher.

We analyze 9 disk RAID 50 (3 RAID 5) as an example (10 disks are required to reach the same capacity of RAID 12):

650) this.width=650; "src=" Http://image.lxway.com/upload/b/e2/be2e59f09a87790e8b698b6fe4ed52f9_thumb.jpg "alt=" Consider using RAID 50 to replace RAID ten (SSD article) "title=" consider using RAID 50 to replace raid ten (SSD) "style=" Border:0px;vertical-align:middle;height:auto; />

Take 12 Disk RAID 50 (4 RAID 5) as an example (10 disks are required to reach the same capacity of RAID 16):

650) this.width=650; "src=" Http://image.lxway.com/upload/0/20/020d43bd4438008a86ff1a4047568d6b_thumb.jpg "alt=" Consider using RAID 50 to replace RAID ten (SSD article) "title=" consider using RAID 50 to replace raid ten (SSD) "style=" Border:0px;vertical-align:middle;height:auto; />

RAID 50 in the above comparison has been able to tolerate the 3rd or even 4th disk failure, but the availability is lower than raid 10, but neither of them can achieve the perfect 100%, so the tradeoff between availability and cost of RAID 50 is a considerable advantage.

Next look at performance, in order to be able to analyze performance well, we followed the first set of comparison scenarios for performance analysis:

650) this.width=650; "src=" Http://image.lxway.com/upload/c/27/c27ce534da687656b583f291b65e50ff_thumb.jpg "alt=" Consider using RAID 50 to replace RAID ten (SSD article) "title=" consider using RAID 50 to replace raid ten (SSD) "style=" Border:0px;vertical-align:middle;height:auto; />

Random Read analysis:

In theory, R10 performance is the best, real test data show 4K, 8K data block under R5, R50, R6 performance is better than R10;

When the data block increases to 16K, 32K, R10 's multi-disk advantage is gradually reflected.

Random Write Analysis:

4K because of the R50, R5 because of a large number of calibration calculations to a certain extent, the impact of performance, but as the data block gradually increased, the advantages of the number of disks also appear. When the data block reaches and exceeds 8K, the R50 performance surpasses the R10 completely;

R10 due to the existence of R1 write synchronization problem, so only 4 disk in support concurrent write, with the increase of data block, R50 and R5 multi-disk performance advantage began to play.

Mixed Random Read-write analysis:

Thanks to multi-disk and non-calibration calculations, the hybrid read-write R10 lead; R50 second, and R10 difference 27%, performance is also relatively close, R5 and R50 performance is linear, R6 performance is the worst.

Sequential Read Analysis:

Because there is no calibration calculation, sequential read performance is basically determined by the number of disks; R50 and R10 performance is also relatively close, the same disk number of R6 and R50 performance, while the smaller number of R5 performance than the first three weaker, in line with the expected. As to why R10 performance is not linearly increased, mainly due to the performance limitations of the array card itself.

Sequential Write Analysis:

Sequential write R5 is optimized optimally; R50 because of the need to simultaneously calculate two times the checksum therefore loses some performance, and R10 performance is equal, when the data block reaches 512K, the multi-disk superiority further manifests, has opened the gap with the R10; R6 because the implementation of the checksum calculation is more complex, the sequential write performance is also the worst.

Performance Test Conclusion:

Performance test shows that the same capacity of the R50 and R10 performance close to: the small block file of random read R50 to be better than R10, random write 4K although R50 and R10 gap in 28%, but the block increased after R50 to overall better than R10. In sequential reading and writing, R50 and R10 are very close.

Fault tolerance, R50 close to R10: The second disk fault tolerance rate R50 very close to R10, the difference of 30%. R10 advantage is mainly in the probability of providing a third, or even the fourth disk fault tolerance, but considering not 100% fault tolerance, so from a fault-tolerant point of view R50 although and R10 some gaps, but also has shown a better fault tolerance rate, at least better than R5. and R50 with flexibility, can even specify 3 sets of R5 to achieve the maximum fault tolerance of 3 disks;

In terms of cost, R50 has a great advantage: by this configuration The R50 is only 3/4 of the R10.

Summarize:

RAID 50 offers close to raid 10 performance, availability and proximity to RAID 5 cost, with a better overall cost-performance advantage, so consider using RAID 50 to replace RAID 10!

This article is from the "@ Fast unblocked @" blog, make sure to keep this source http://speediness.blog.51cto.com/760841/1783323

Server disk RAID0,RAID1,RAID10,RAID5,RAID50,RAID6,RAID60 analysis and case