stripe in array (stripe), stripe unit

Excerpt: http://blog.sina.com.cn/s/blog_4a362d610100aed2.html

In a disk array, the data is striped (stripe) across all hard disks on the disk array. The way this data is distributed can compensate for the insufficient amount of data that the OS reads.

1. Collect the stripe unit (stripe unit) from the first hard drive on the array to the last hard drive. It can be called a stripe (stripe). Sometimes the stripe element is also called the staggered depth. In optical fiber technology, a stripe element is called a segment.

2. The data is distributed in the form of a stripe on the hard disk in the array, which refers to the stored procedure of the data in all the hard disks in the array. The data in the file is segmented into small pieces of data that are stored sequentially on the hard disk in the array, and this smallest chunk is called a stripe unit .

The size setting of the stripe element has an impact on performance. If you select a stripe unit that is too small, the service application needs more data than the stripe unit size. Such a logical I/O can cause multiple physical I/O. This will definitely affect your performance.　If you select a stripe unit that is too large it can cause performance degradation because a larger than required hard disk I/O may continue to curt each demand. Will increase the time I/O operation will also read out the data that is not needed. Especially in RAID5, you must read the entire bar from the hard disk to generate a parity message, and use too large a stripe to read the extra data while also calculating the parity information.

This allows us to select the size of the stripe element as large as the size of the stripe unit and the average I/O request generated by the server application. That is, ideally the next time I/O requests produce a disk I/O operation. The best way to do this is to detect the I/O performance of the server application.

We use Iostat under Linux.

We use Perfmon under Windows.

stripe in array (stripe), stripe unit