While this is the first time I've ever heard that 7.2k benchmark the same as 15k, I do know that 10k 2.5" benchmark nearly identically to 15k 3.5", mostly because of the smaller diameter reducing the rotational latency.
So, given that, if you're doing 24x10k 2.5" drives, you'll have no problem at all. I also agree that 24x10k will give you as much or better performance than 12x15k. However, a lot depends on the vendor. If you implement this as 24x10k drives and you do have problems, is the vendor going to support you, or just throw up their hands and tell you you're on your own because you didn't listen to them? -Adam On Wed, Feb 19, 2014 at 6:49 AM, Edward Ned Harvey (lopser) < [email protected]> wrote: > > From: [email protected] [mailto:discuss- > > [email protected]] On Behalf Of Ski Kacoroski > > > > We are putting in a very large application based on MS SQL and the > > vendor is insisting that we use 12 15k 600GB disks in a raid 10 > > configuration. Our standard is 10k disks so I would like to use 24 10k > > disks in raid 10 so I can use existing hot spares. This supplies more > > iops, more bandwidth; but the vendor is still insisting on 15k disks as > > they think SQL will have write stalls. > > > > Am I missing anything here where the 12 15K disks would be better then > > 24 10k disks? Could the latency be less with the 15k disks? I really > > am trying to understand if there is a valid reason the vendor is so set > > on the 15K disks other than 'we have always done it that way and it > > worked'. > > I generally use 7.2krpm disks, because I benchmarked them and found > there's virtually no difference. Nothing against higher rpm disks, but > it's not worth paying extra, if it's going to cost you extra. > > If you want something higher IOPS, go for SSD's. > > In a HDD, your access time is a combination of head seek & rotational > latency. (The electronic propagation delay is basically negligible.) The > head seek is around 9ms average. The rotational latency will average a > half a rotation, so at 10,000 rpm's, guess how much time that is. Sounds > like a 20th of 1ms. > > Even your sustainable throughput doesn't improve with higher rpm's. Your > sustainable throughput is determined primarily by the frequency response of > the head, and regardless of how good your disks are, that's approx > 1Gbit/sec. (Give or take something like 15%). > > In your situation, the thing that will matter more, is 24 disks vs 12 > disks. Your 24 disks will probably outperform the 12 disks, just by virtue > of the fact that you have a higher number of disks. > _______________________________________________ > Discuss mailing list > [email protected] > https://lists.lopsa.org/cgi-bin/mailman/listinfo/discuss > This list provided by the League of Professional System Administrators > http://lopsa.org/ >
_______________________________________________ Discuss mailing list [email protected] https://lists.lopsa.org/cgi-bin/mailman/listinfo/discuss This list provided by the League of Professional System Administrators http://lopsa.org/
