Hello, RAID 1 is mirror, RAID 0 is strip.
We are using sw raid - mirror (2 same disc - one partition, no lvm). Faster reading, little slow write (but better security - you have 2 copies of data). If you can use raid 5 (with 3 hdd or with 5 hdd - better with 5, or use raid 10 (mirror strip) with 4 hdd) We are using dual core intel with 2G ram, running qmail + vpopmail + shupp toaster + clamav. 1M emails per months. There is also web server + db server (mysql). Load sometimes goes over 5 (mostly when accessing large mailboxes). Over day there is about 15 - 25 incoming connections per second, so most of cpu eats spamd. 2007/9/6, Joey Novak <[EMAIL PROTECTED]>: > Hey Jeff, > > I don't have much info specific to your question, but I wanted to chime in > here. I don't think you will find a lot of performance increase by using > RAID for the queue, as data is read and written a LOT and Raid 0 (Mirroring) > (correct me if I am wrong) usually only makes reads faster... We have found > that most of the bottleneck on our mail server was spamd. > > I have NOT setup a single toaster with RAID, but we implemented a modified > version of Bill's ISP setup, and you may find some of our results > interesting. > > This page: > http://www.chemistry.wustl.edu/~gelb/castle_raid.html shows > the results of some tests with hardware vs. software raid, and show that in > most situations, the performance increase of hw vs. software RAID is small > (unless it is a very expensive raid card). > > We have a 4 node mail cluser, where there are 4 boxes that run Bill's > toaster, all of them store their mail on the same NFS server, which has > Seven drives in a software Raid 5 Array. ALL of our cluster nodes are > almost ALWAYS at 100% CPU (Except for a few hours each night when they > finaly clear their queues completly and can rest a little). Here is the > output of a current "top" > > top - 12:28:05 up 50 days, 16:30, 1 user, load average: 0.50, 1.15, 1.20 > Tasks: 325 total, 1 running, 324 sleeping, 0 stopped, 0 zombie > Cpu(s): 10.9% us, 1.7% sy, 0.0% ni, 62.6% id, 24.2% wa, 0.0% hi, 0.7% si > Mem: 1035284k total, 1021516k used, 13768k free, 299584k buffers > Swap: 2096472k total, 388k used, 2096084k free, 444052k cached > > PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND > 4237 mysql 15 0 139m 40m 4096 S 11.6 4.0 7008:38 mysqld > 452 root 10 -5 0 0 0 S 0.7 0.0 86:34.96 md0_raid5 > 22301 root 15 0 2088 1104 760 R 0.3 0.1 0:00.55 top > 1 root 15 0 1692 552 472 S 0.0 0.1 0:01.61 init > 2 root 34 19 0 0 0 S 0.0 0.0 0:02.23 ksoftirqd/0 > 3 root RT 0 0 0 0 S 0.0 0.0 0:00.00 watchdog/0 > 4 root 10 -5 0 0 0 S 0.0 0.0 0:00.01 events/0 > 5 root 20 -5 0 0 0 S 0.0 0.0 0:00.00 khelper > 6 root 16 -5 0 0 0 S 0.0 0.0 0:00.00 kthread > 50 root 10 -5 0 0 0 S 0.0 0.0 0:37.66 kblockd/0 > 51 root 20 -5 0 0 0 S 0.0 0.0 0:00.00 kacpid > 181 root 10 -5 0 0 0 S 0.0 0.0 0:00.48 ata/0 > 182 root 20 -5 0 0 0 S 0.0 0.0 0:00.00 ata_aux > 183 root 10 -5 0 0 0 S 0.0 0.0 0:00.00 ksuspend_usbd > 186 root 10 -5 0 0 0 S 0.0 0.0 0:00.00 khubd > 188 root 10 -5 0 0 0 S 0.0 0.0 0:00.01 kseriod > 207 root 10 -5 0 0 0 S 0.0 0.0 0:00.00 kapmd > 215 root 10 -5 0 0 0 S 0.0 0.0 31:22.55 kswapd0 > 216 root 20 -5 0 0 0 S 0.0 0.0 0:00.00 aio/0 > 362 root 10 -5 0 0 0 S 0.0 0.0 0:00.00 scsi_eh_0 > 363 root 10 -5 0 0 0 S 0.0 0.0 0:00.00 scsi_eh_1 > 364 root 10 -5 0 0 0 S 0.0 0.0 0:00.00 scsi_eh_2 > 365 root 10 -5 0 0 0 S 0.0 0.0 0:00.00 scsi_eh_3 > 379 root 10 -5 0 0 0 S 0.0 0.0 0:00.00 scsi_eh_4 > 380 root 10 -5 0 0 0 S 0.0 0.0 0:00.00 scsi_eh_5 > 387 root 10 -5 0 0 0 S 0.0 0.0 0:00.00 scsi_eh_6 > 388 root 10 -5 0 0 0 S 0.0 0.0 0:00.00 scsi_eh_7 > 405 root 11 -5 0 0 0 S 0.0 0.0 0:00.00 kpsmoused > > As you can see, the software raid (md0_raid5) takes almost no cpu power, > and in fact, most of the cpu power goes to MySQL. You can also see that the > "wa" percentage (which shows how much cpu time is spent waiting for io > operations, frequently disk io), is pretty low. > > Joey > > > On 9/6/07, Jeff Koch <[EMAIL PROTECTED]> wrote: > > > > > > Has anyone successfully setup Bill's toaster with SATA RAID? A year or two > > ago we setup a toaster with a two drive 3ware IDE RAID mirroring setup and > > the performance was awful. Maybe it was because we didn't have write > > caching enabled on the RAID controller or should have tweaked the kernel > > settings. > > > > I looked at Bill's proposed setup for an ISP but we're just trying to do > > this for a single server setup. The only solution we've been able to come > > up with in the past is to have a single small drive for booting, > /var/qmail > > and /var/logs and run SATA RAID for /home/vpopmail and everything else. > But > > we'd really like to have RAID running for the qmail queue since that's > what > > beats the hell out of a hard disk. > > > > Any recommendations or experiences anyone? > > > > > > > > Best Regards, > > > > Jeff Koch > > > > > > > > -- > --- > http://www.joeynovak.com > > C) 803-409-9969 (Work Cell) > W) 757-233-0834 > "Very funny, Scotty. Now beam down my clothes." > > Be nice to nerds. Chances are you'll end up working for one. > --Bill Gates > > Your most unhappy customers are your greatest source of learning. > --Bill Gates > > Cope with Life, go buy a slurpee! > http://www.slurpee.com/games.html
