Objectives
After completing this section, you should be able to describe basic RAID architecture to understand the correlation between application writes and RAID sizing alignment.
Reviewing RAID
The Redundant Array of Inexpensive Disks (RAID) is a virtualized data storage technology, combining storge drives as a single logical storage unit to support increased performance, redundancy, and fault tolerance. RAID is an alternative to storage systems based on single, large and expensive disks. RAID supports the failure of a RAID disk without data downtime or data loss.
A RAID breaks the data in small chunks which are written to different disks of the array. Red Hat Enterprise Linux 8 defines a chunk size of 512 KB by default. The data is distributed among the RAID disks to meet one or more of the following data strategies: data striping, data mirroring, and data parity.
Data striping divides the data in stripes that are distributed among several disks inside of the RAID array. This strategy increases throughput because data is retrieved from several disks instead of a single disk.
Data mirroring provides redundancy copying the stripes to at least two different RAID disks, so in case of a disk failure, the RAID array can provide data with other data copies available in the array with no data downtime or loss. Read throughput of RAID 1 is slower than the fastest drive. Write throughput is always slower because every drive is updated, and the slowest drive determines the write performance.
Data parity supports data redundancy by adding parity to the stripes so other stripes related to the same data volume can be reconstructed using the working data and its associated parity bits.
A RAID array supports one or more of these strategies on different RAID levels. The following table includes some of the most used RAID levels.
![](https://s2.save4k.ru/pic/oWyoHSq0DFU/maxresdefault.jpg)