Or I would say, will SSD kill RAID in the near future? Let start by the beginning, with some simple definitions.
What is RAID? RAID stands for Redundant Array of Independent Disks, originally coined at Berkley University in 1987. RAID is a method of combining several hard drives into one unit. It offers fault tolerance (except fo RAID0) and higher throughput levels than a single hard drive or group of independent hard drives. RAID levels 0,1, 10 and 5 are the most popular.
Why do we need RAID? To achieve high levels of storage reliability from low-cost and less reliable PC-class disk-drive components, via the technique of arranging the devices into arrays for redundancy.
Why are PC-class disk-drive components less reliable? Because of their moving parts, platters, heads, actuator arms, spindle motor, head motor… Although modern drives make extensive use of Error Correcting Codes (ECCs), particularly Reed–Solomon error correction, that doesn’t protect from failure of a moving part, and there are quite a lot of them as you can see.
What is SSD? From Wikipedia.org, a solid-state drive (SSD) is a data storage device that uses solid-state memory to store persistent data. An SSD emulates a hard disk drive interface, thus easily replacing it in most applications. The usage of the term “solid-state” has been adopted to distinguish solid-state electronics from electromechanical devices. With no moving parts, solid-state drives are less fragile than hard disks and are also silent (unless a cooling fan is used); as there are no mechanical delays, they usually enjoy low access time and latency.
So there is no moving parts? Nope, nada, nicks, nothing.
So if there is no moving parts, the MTBF is much higher than a regular hard disk, is it? Well surprisingly, it is not… Look at 2 examples here:
- A WD VelociRaptor 300GB 10k rpm hard disk’s MTBF -> 1.4 million hours!
- An Intel X-25M 80GB Solid State Drive’s MTBF -> 1.2 million hours!
How do you calculate MTBF? Wow this one is very tricky, there are different models, the two most popular models are MIL-HDBK-217 and Telcordia (formerly Bellcore) SR332. There are so many parameters that you can have different MTBF results with one model, and with another model you could have very dissymmetric results. Basicaly we can say that MTBF of a drive is obtained by multiplying a large quantity of the drives (thousands?) with the number of hours running before experiencing a failure in the batch. For example, when a disk manufacturer batch tested 1500 units of hard disk and achieved an average of 30 days operation out of the batch between each individual unit failure, then the MTBF of the disk is 1500 x 30 x 24 hours = 1 million hours. Unfortunately hard disk vendors are not keen to explain how they do to calculate the MTBF 😦
So now that we understand a bit more the whole picture, that we saw that the MTBF is almost the same between a regular hard disk and a SSD one, do you think we can rely on SSD and stop using RAID to protect your data? I don’t! RAID will still be there for quite some times, and actually by setting a RAID you could achieve even better results in terms of bandwidth, IOPS, and … MTBF!