RAID5 more/larger or fewer/smaller drives?

[JBruyet]

Hey all,

I'm getting ready to order some file servers (with hardware array controllers) and I have a question concerning a RAID5 array. Is there an advantage to using fewer/larger drives over more/smaller drives? I had heard that spreading data over more drives decreases the chance of data corruption. I can't really see how that is possible, but using more drives would create more heat and I CAN see where that could be a problem. I've looked around on the net and I can't find anything addressing this issue. Any ideas, suggestions, recommendations? Or maybe point me to a URL that deals with this issue?

Thanks,

Joe Brouillette

 

[bazzert]

If high capacity is first priority, the common option is RAID 5, which distributes data to all array members and adds rotating parity information, too. The net capacity of the whole array is the sum of all drive's storage minus the capacity of one single drive (so smaller hard drives are usually selected as you'll lose less of your total storage). However, this solution is not as ideal as it seems. Calculating the parity information as well as the real time reconstruction of data in case of a defective drive requires a lot of CPU performance and you either buy an expensive Raid controller or let the CPU cope with the calaulation which may nave an impact on performance.

 

[NickFerrar]

These days I'd just go with fewer/larger drives, there's little, if any, advantage to more/smaller drives. If performance is a big factor consider RAID0+1 or multiple RAID containers on separate channels etc.

 

[OUCATS]

I was in your position about 2 years ago and couldn't decide between more/smaller drives or less/larger ones. So I calculated out the price per GB and the smaller drives were cheaper. Now looking back I wish I would have opted for the larger drives. Currently I am running out of space on my RAID5 array and I'm stuck because I do not have any available HD slots left (because I have filled them all with smaller drives). Currently the server is working great (plenty of RAM and processing power) except for the drive space. So now I'm stuck with either purchasing and external array of drives (SAN, NAS or SCSI) to connect to the server or purchasing another server. If I had purchased the larger drives, then I'd still have 3-4 HD slots to add more storage.

 

[JBruyet]

Thanks all. I'm going to go with the 72 gig drives and get the servers ordered.

Thanks again,

Joe Brouillette

 

[technome]

With raid 5 having more than 3 drives does increase throughput significantly up to the point a raid adapter's scsi channels are saturated ,or the Pci bus is saturated (bit simplified). Greater than 5 ,possible 6 u320 drives are the point at which little or any speed increase is gained by adding more drives to a 2 channel raid adapter.My personal server tops out at 5 drives, on a Lsi u320-2x raid adapter.

Would have pulled more links but my line is slow right now...

http://www.storagereview.com/guide2000/ref/hdd/perf/raid/concepts/perfStripe.html

If you have a bit of time, I suggest you cruise over to 2cpu.com, search for posts by FemmeT (tests raids big time), and read the topics he posts in; he discusses ,or others have discussed the throughput of multiple drives over the bear minimum of 3.
FYI, generally, very large drives >140Gig have somewhat lower speed specs than smaller drives, 72 Gig is a good capacity. As far as heat, insure your array's temperature maintain a few degrees above room temperature (within 10 F), even if you must add fans to the server; most of my server maintain drives at < 6 degrees F over room temperature, optimal room temp is around 75 F. Spreading data over more drives has no bearing on data corruption.

My comments are related to hardware raid, I would not even consider OS implemented raid 5 .

 

[karmic]

Are you worried just about corruption? about price? about the space???

What i've been offering to clients lately is multiple SATA mirrors... Not as fast as the raid 5 mind you but cheaper.

2 or 3 SATA mirrors are cheap, loads of space, fast enuf if you're running multiple servers and offers separate drives for OS and data drives...

I've had clients that bought one honking big server, dual xeon, dual raid 5 (140 gig total) and a bunch of other bells and whistles for up to $12-15000 then bought the less expensive entry level servers with the sata mirrors... Works great.

Just something to think about....

~ K.I.S.S - Don't make it any more complex than it has to be ~