Disk I/O optimization

Disk I/O optimization

1. Performance Testing

Our applications often require access to the disk system, and disk I/O is often time-consuming to determine if I/O is a bottleneck, and some parameter indicators can be referenced.

We can stress test the application to see if the system's i/owait indicators are normal. For example, if the test server has 4 CPUs, the ideal I/O wait parameter should be no more than 25%, and if it exceeds 25% then I/O is likely to be a performance bottleneck for the application. Linux operating system can be viewed through the iostat command.

[[Email protected] ~]# iostat Linux2.6. +-431.23.3. el6.x86_64 (IZ239YQA63MZ) -/ -/ -_x86_64_ (1CPU) Avg-CPU:%user% Nice%system%iowait%steal%Idle0.37    0.00    0.27    5.28    0.00   94.08Device:tps Blk_read/s blk_wrtn/s Blk_read Blk_wrtnxvda5.08       208.63         2.38  335106884    3815224Xvdb1.95        30.33         0.35   48718248     559072

Usually we look at another parameter when judging I/O performance. Is the IOPS, which is also related to the size and access of the data blocks stored on disk, but is primarily determined by the speed of the disk. The higher the disk speed, the higher the disk IOPS.

2. Improve I/O performance

The usual ways to improve disk I/O performance are:

1. Increase the cache and reduce the number of disk accesses.

2. Optimize the Disk Management system, design the optimal disk mode policy, has the disk's addressing strategy, which is considered at the level of the underlying operating system.

3. Design reasonable data storage blocks, and access to these data blocks of the strategy, these are at the application level, such as: we can give the data stored in the design index, through the addressing index to speed up and reduce the amount of disk access, but also in an asynchronous and non-blocking way to speed up disk access.

4. Apply a reasonable Raid

Disk I/O optimization