RAID5 performance is *highly* dependent upon how fast your controller is, and how much offloading it does to your CPU. This is because it requires parity information to be calculated for every write. Most INTEL/AMD/PROMISE/HIGHPOINT controllers are to a varying degree fakeraid, and are very poor at raid5. Especially those integrated into mainboards. RAID 5 doesn't *have* to be killer slow, but the pricepoint most people are willing to pay means they get poor raid5 performance. Justin Krejci wrote: > I concur. We did some I/O benchmarking using a 24 drive server, I think > it was this server > http://supermicro.com/products/chassis/4U/846/SC846E1-R900.cfm with 24 > 15K SAS drives. > > > > The benchmarking was random reads and random writes of varying sizes > (multi-gig to make sure it was beyond the total cache available). > > > > RAID0 = fastest (just for fun and to set the watermark) > > RAID10 = second fastest (about 65% of RAID0) > > RAID5 = a very very distant last (about 20% of RAID0) > > > > These % numbers are rough and from memory of several weeks ago. It is > combined read and write performance. > > > > We also had two 4-gig iram drives in a RAID0, wow that is fast!! > > > > ------------------------------------------------------------------------ > > *From:* tclug-list-bounces at mn-linux.org > [mailto:tclug-list-bounces at mn-linux.org] *On Behalf Of *Chris Barber > *Sent:* Wednesday, May 28, 2008 3:58 PM > *To:* tclug-list at mn-linux.org > *Subject:* Re: [tclug-list] OT - Hardware Advice > > > > It would be insane to hook up 12 drives in a RAID5. If you lose a > drive, it's gonna take a day to rebuild the array and what happens if > you lose a second drive while you're rebuilding? 4 x 250GB drives in > RAID 5 took about 10 hours to rebuild. It is probably quicker to format > and re-install. I hope you have backups. > > I use RAID 1+0 for database servers. For a app, file, or mail server, I > do RAID 1 unless I needed the space, then I'd do RAID 6. I don't run > RAID 5 on any of my servers anymore because I need redundancy beyond > losing 1 drive and the write performance blows. > > I should note that you can get sufficient performance from a RAID 5/6 by > using a controller with 256MB of cache and allocated most of it to > write-back cache. Also, you should have a lot of RAM to cache database > pages in RAM and cache queries to avoid hitting the disk. > > -Chris > > > Bret Baptist wrote: > > On Wednesday 28 May 2008 1:47:32 pm Josh Paetzel wrote: > > > >> On Wednesday 28 May 2008 11:14:50 am Justin Krejci wrote: >> >>> http://download.intel.com/support/motherboards/server/srcsas18e/sb/axxrpc >>> m2 _ tps_10.pdf >>> >>> Benefits are identified in this PDF. >>> Data caching (write-back cache can greatly improve write performance) >>> >>> Busy databases servers commonly need lots of I/O >>> >>> Also consider running RAID10 if you have drive availability (4 drive >>> minimum) as you will get much higher I/O performance with that as well >>> especially with writes. Though if you need the capacity, RAID5 will give >>> you one drive more of capacity. RAID10 can also give you a smaller chance >>> of data loss due to drive failures as you can potentially lose up to half >>> of your drives and still operate whereas using RAID5 and losing 2+ drives >>> = disaster. >>> >> Somewhere a DBA just rolled over in his grave at the mention of a database >> using RAID 5. If you're ever going to care about performance at all don't >> use RAID 5. It's particularly slow at the sorts of write I/O database >> systems typically generate. >> > > > For the most part this is true, however on a lot of modern RAID controllers if > > you hook 12 drives up in RAID-5 you are going to see amazing performance. > > > > Here is an article with a very thorough review of 9 SATA RAID cards: > > http://tweakers.net/reviews/557/26/comparison-of-nine-serial-ata-raid-5-adapters-pagina-25.html > > The issue here is that they do not do a RAID-10 test with 12 drives. > > > > I don't know what they are doing on the Coraid SR 1521 to make RAID-5 faster > > than RAID-10, but when you get up to 14 drives in the chassis you get much > > better throughput, now mind you this is not random I/O, just another thing to > > think about: > > http://coraid.com/support/sr/ANSR002.pdf > > > > For a large number of drives in a RAID-5 you get really good performance and > > much higher capacity. > > > > Not that this is really relevant to the original poster. > > > > > > Thanks. > > > > > > > ------------------------------------------------------------------------ > > _______________________________________________ > TCLUG Mailing List - Minneapolis/St. Paul, Minnesota > tclug-list at mn-linux.org > http://mailman.mn-linux.org/mailman/listinfo/tclug-list