The similarities and differences between RAID10 and RAID5 _ routers, switches

From a common application, the requirements of the storage system has good IO performance also requires the protection of data security work, so raid10 and RAID5 should be our focus of attention.

Simply talking about the advantages of different RAID levels is impractical, and raid is only a complete array with disk binding.

The following from the IO performance, data refactoring and the impact on the system performance, data security, etc., combined with the status of the disk to analyze the differences between the two technologies.

IO performance:

Read operation Raid10 and RAID5 are quite,

For RAID-5 in some very small data writes (such as smaller data than each stripe) requires 2 read, 2 write, and 2 XOR operations, for a single user write operations, the new data must be used before the application of the old data from the calibration disk removed, the whole process is this: read out the old data, Old data and new data to do XOR, and create an immediate value, read the old data of the checksum, will be the time and validation data xor, and finally write down the new checksum information. To reduce the impact on the system, most RAID5 read and write the entire stripe (including the checksum strip) to the cache, performs 2 XOR operations and then emits parallel writes (usually to the entire stripe), and even with the above optimizations, the system still needs additional read and XOR operations for this write operation. A small number of write operation difficulties make RAID-5 technology rarely used in intensive write-operation occasions, such as rollback field and redo comrades. Of course, you can also define the stripe size of the storage system to match the data size of the regular read and write action, but this limits the flexibility of the system and does not apply to other applications in the enterprise.

For RAID10, because there is no data validation, each write operation is simply a write operation. This raid10 better than RAID5 in write performance.

Data refactoring:

For RAID10, when a disk fails, the process of data refactoring is only a copy of a new disk, and if the disk's capacity is assumed to be 250G, the amount of data replicated is 250G.

For RAID5 storage arrays, you need to read data from each disk, recalculate the amount of data on a hard drive, and if RAID5 is 4+1, and the capacity of each disk is 250G, You need to read out a total of 1000G of data in the remaining 4 disks to calculate the 250G data. From this point of view, RAID5 's workload and time spent on data reconstruction should be much larger than RAID10, and load-changing generals affect performance during refactoring, which means that the likelihood of data corruption is greater again.

Data security:

RAID10 system in the case of a disk failure, only the corresponding mirror disk of the failure disk also invalidated, will cause data loss. No data loss occurs with other disk failures.

RAID5 system in the case of a disk failure, as long as any one disk failure, will result in data loss.

From the comprehensive point of view, the RAID10 and RAID5 system after the failure of a disk, the data reconstruction, RAID5 time to spend than raid10 long, while the system load during the reconstruction RAID5 higher than the RAID10, At the same time, the probability of RAID5 data loss is higher than RAID10, so the reliability of RAID5 system is much lower than that of RAID10 during the data reconstruction.

RAID5 is higher than RAID10 on disk space rate, RAID5 space utilization is (N-1)/n (n is the number of disks in the array), and RAID10 disk space utilization is only 50%.

However, considering the disk, the quality of the ATA hard drives produced by today's hard disk vendors can already assume enterprise-class applications, and the capacity is increased by a considerable amount, and the storage capacity of a single disk 400G can now be achieved. SCSI drives use small diameter platters because of high speed, and the capacity increases relatively slowly. The ATA disk has a smaller cost of ownership relative to the SCSI disk.

In this context, in some applications with very high IO requirements, RAID5 with SCSI disk is a better choice, other applications with large ATA hard drive combined with RAID10, both reduce the RAID10 to obtain a certain amount of storage space must adopt double disk space cost of ownership, Also avoids the RAID5 relative raid10 each kind of shortcoming.

In enterprise applications, RAID10 with ATA disks means a better choice.