Server Fault is a question and answer site for system and network administrators. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

This question already has an answer here:

Assume a single disk has a baseline of 1W & 1R performance, how does the different RAID configurations behave?

My initial guess:

Raid 0: 2W 2R 
Raid 1: 1W 2R
Raid 5:  ?  ?
Raid 6:  ?  ?
Raid 10: ?  ?



  • Using hardware RAID
  • With min. number of disks required, e.g.3 disks for Raid5
share|improve this question

marked as duplicate by ewwhite, RobM, Chopper3, Jenny D, mdpc Apr 20 '13 at 19:16

This question has been asked before and already has an answer. If those answers do not fully address your question, please ask a new question.

You did not tell us whether you are considering software of hardware raid in your question. Also, different raid controllers may implement things differently. – Khaled Apr 20 '13 at 7:22
@Khaled, why do you care whether it would be hard- or software? – poige Apr 20 '13 at 7:35
@Khaled Updated my questions. – Howard Apr 20 '13 at 7:44
up vote 1 down vote accepted
  • R0: roughly double sequential performance
  • R1: no or slight negative impact on writes, 1-1.7 (estimate) factor on random reads
  • R10: roughly double everything ... (using linux mdadm and some tuning you can triple sequential reads)
  • R5 3drive: same as R0 2 drive(assuming your controller is powerful enough)
  • R6 4drive: same as R0 2 drive(assuming your controller is powerful enough)

both R5 and R6 can greatly degrade performance on small (random) writes because of the read-modify-write operations needed to write changes.

there is a lot of information, you are looking for on the Wikipedia article Standard RAID Levels

share|improve this answer
This is the answer I need, thanks! – Howard Apr 20 '13 at 9:55
worth mentioning that write hole is not a performance, but a durability factor – DukeLion Apr 20 '13 at 10:28
@DukeLion thanks for the correction. Mistaken the term. Edited my answer. – Fox Apr 20 '13 at 10:41

With raid 0 you can theoretically get twice the read and write performance, but in practice that really depends on the workload. Raid 1 will not give you additional write performance, and I'm not sure whether it will give you extra read performance (it may read from both disks, but I don't know if it does).

When it comes to raid 5/6, that depends on a lot more than just the number of disks. Each read or write will result in at least two (three for raid 6) disks being used, and you also need computing resources (CPU or raid card) to calculate and verify the parity.

And all this leaves out (battery backed) write cache on the raid controller. That can cause an enormous performance increase as well (in fact, many of my databases wouldn't run without one!), and this increase is not related very much to raid level or number of disks.

So the only way to find an actual answer to your question is to try it out, tune the cache settings and tune your filesystem. No theoretical answer has any possibility of being correct, as most of the behaviour depends heavily on the application and configuration.

share|improve this answer
Raid 1 - depends on controllre /programming. High end ones do. Heck, adaptec has a hybri raid (ssd, disc) and then redirects all reads to the SSD. – TomTom Apr 20 '13 at 9:42

It's non-linear, it depends on number of disks (RAID-5 can be ≥ 3 of them), stripe size, workload and so on. So this is completely wrong way to have performance estimation.

But what usually is taken into consideration is number of R/W-operations involved by single write operation on logical disk.

share|improve this answer
Yes, but just want to rough estimate. – Howard Apr 20 '13 at 7:44

Not the answer you're looking for? Browse other questions tagged or ask your own question.