While RAID systems typically make use of mechanical hard disk drives, SSD RAID distributes redundant blocks of data across multiple flash-based SSDs. While this offers a much faster solution, it does make RAID data recovery more difficult.
Redundant array of inexpensive disks (RAID) – later changed to redundant array of independent disks – emerged in the 1980s, in an era where mechanical hard disk drives were the primary storage format. The purpose of the format was to provide the user with increased performance while providing fault tolerance. Fault tolerance refers to the capabilities of a computer system or network to deliver an uninterrupted service, even in the face of one or more of its components failing. In short, you get a faster drive, and your data is more secure, lessening the likelihood of needing to undertake data recovery.
In recent years, data storage vendors have started to produce RAID storage systems and servers that use NAND flash-based solid state drives (SSDs), for even higher performance with protection against data loss in the event of the array’s failure. RAID systems have started to move away from a whole-system level towards applying redundancy at a finer granularity. As with a typical hard disk drive-based RAID system, data in an SSD RAID can be divided at block level or distributed across multiple SSDs in a number of ways, also known as RAID levels. Different RAID levels offer different benefits in terms of speed, capacity and data security/ease of RAID data recovery.
Disk mirroring, also known as RAID 1, involves the replication of data across two or more disks, giving you higher levels of performance along with added data security, reducing the need to undertake potentially expensive RAID data recovery. Disk striping involves dividing data into blocks and spreading it across multiple drives, and is synonymous with RAID 0, meaning there’s a higher chance of data loss; if one drive fails, your data could be corrupt. Parity, or RAID 5, works by combining three or more drives, and logically combining them into a single entity. Striping with no redundancy or parity is used for increased performance, and striping with parity is used to strengthen data protection and to make RAID data recovery simpler.
HDD-based RAID was originally developed to increase performance; the operating system would see the drives in the array as a single logical storage unit, speeding up performance. With SSD-based RAID systems, however, the primary purpose isn’t increased performance, because flash-based storage is inherently faster anyway. Instead, SSD RAID systems are used to protect users’ data if a drive in the array fails. SSD-based RAID offers reduced access time – the time from the start of one access to the time when the next access can be undertaken – as well as superior I/O performance. When compared alongside a comparable HDD-based RAID system, as SSD RAID can significantly outperform it. Then there’s the reduced costs that are associated with using flash-based storage, perhaps not in the short term, but certainly in the long term. A typical NAND flash-based SSD consumes much less power than a traditional mechanical hard disk drive, due to it having no moving parts. When a large number of drives are combined, the power savings of an SSD RAID – from both electricity bills and simplified cooling systems – can translate to lower operating costs in the long-term.
On the other hand, there are some limitations of using SSD RAID, which are mostly synonymous with the drawbacks of using NAND flash outside of a RAID system. SSDs, while having increased performance, have a higher cost per gigabyte compared to traditional HDDs, and NAND flash chips wear out due to having a limited number of program/erase cycles they can perform.
