Solid State Hard drives vrs 15K RPM SAS.

[mjradik]

I have setup and manage many VMWare ESXi servers. Most SATA RAID 10, but some SAS RAID 10 w/ 15K drives. Now I’m wondering more and more about solid state hard drives. I use them in my desktop and laptop, and love the speed.

I’m getting ready to build another system, and am most likely going to use (8) 300GB 15K RPM SAS drives, RAID 10 and on LSI or Dell Perc 6i RAID controller, like before. However, 128GB SSD hard drives are about the same price, and wondering in a high random IO environment, if SSD drives will be better? I don’t need a lot of space, but use a lot of drives for better performance.

All the research and googleing I’ve done, I just turn up similar topics from like 6 months to a year ago. SSD drives have gotten better and cheaper since then.

This particular VM host server is going to be mainly hosting a SQL server and Exchange server, among other less used servers.

Do you think I’ll get better and faster/better performance with SSD or 15K RPM drives? I’ve seen bench mark tests that give (8) SSD drives in RAID close to 1GB per second transfer rate, but that’s sequential, not Random IO.

Thanks.

[mcowger]

SSDs are exceptionally good at random IO, certainly better than almost any disk.

Be very careful about lifetime - the door wear out VERY quickly if hit wrong, so make sure you have decent redundancy in there (RAID6 at least I would want) and a good warrenty :smileyhappy:

--Matt

VCP, vExpert, Unix Geek, Storage Nerd

[mjradik]

You would choose RAID 6 over RAID 10? Same redundancy about. (all dependons on which ones fail in the RAID 10) Isn't RAID 10 perfered for SQL type servers because of the way the blocks line up or what ever? Or does it not matter with SSD?

[mcowger]

SSD is generally fast enough that it wouldn't really matter (assuming you have a good controller).

I'd choose RAID6 because ANY 2 drives could fail without dataloss - in RAID 10, if you had bad luck and lost 2 drives in oneof the mirrors (very likely they would wear out at the same time because they are doing the same workload!), you'd lose your whole array.

--Matt

VCP, vExpert, Unix Geek, Storage Nerd

[J1mbo]

On the other hand, the rebuild time for RAID-6 is very much longer vs disk-to-disk clone for RAID-10. RAID-10 with a hotspare, or RAID-6 without...., I don't know the answer to that.

Re wear, ensure the alignment is right to minimise wear. Recent Windows OS's align correctly anyway.

http://blog.peacon.co.uk

Please award points to any useful answer.

[Jackobli]

On the other hand, the rebuild time for RAID-6 is very much longer vs disk-to-disk clone for RAID-10. RAID-10 with a hotspare, or RAID-6 without...., I don't know the answer to that.

Good point, but there it will be "just" 128 GB disks, given a fast controller, that should not matter so much?

On the other hand, aren't there still performance problems with SSD when the disks become full(er) (read - zap -write opposite to just write on regular disks)?