Hi people, I am experiencing a poor sequental write performance with 3-disk RAID5 array on 3Ware 7500-8 - it is only 11-12 MB/s. Under Win2000 i get 30-35 MB/s from the same array. I have used different methods: dd, iometer and "copy /dev/zero tstfile" - they all give me the same low figures (11-12 MB/s). Has anybody managed to get good (> 30 MB/s) raid 5 write performance from 3ware 7500 on Linux and desktop (non-server) hardware - 32-bit 33MHz PCI ? And why does Linux give much worse performance than Win2k? It was disapointing for me to see this after i recommended 3ware cards for our NAS solution. 3ware use server motherboards and Windows 2k and for their benchmarks. I found several Linux benchmarks that showed good results but they used server motherboards: http://home.fnal.gov/~yocum/storageServerTechnicalNote.html http://pcbunn.cacr.caltech.edu/gae/3ware_raid_tests.htm I also found that another Russian guy had the same problem a year ago: http://bioinformatics.org/pipermail/bioclusters/2002-August/000364.html I run RedHat 9 with unchanged kernel and have updated firmware and drivers to 7.6 version 3Ware software. Write cache is enabled. Disks used are 80 GB ibm 120gxp (IC35L080AVVA07-0). I used 2 hardware configurations: 1) Via Epia-M motherboard with C3-933 Mhz, 256MB RAM 2) Asus A7V133 with Athlon 1Ghz, 512 MB RAM - they showed the same results except for CPU utilization: on Epia i get high numbers (50-70% for read, 90-100% for write) and on Athlon numbers are much lower(22-36% read, 7% for write). I also tested RAID0 3-disk array on Epia (just trying to understand what's wrong - i need RAID5 for my solution) and get the strange results: iometer: seq. read (64-1024KB blocks, 1-64 outst. I/O) - 34 МB/s, CPU total 65-70% seq. write (64-1024KB blocks, 1-64 outst. I/O) - 48 МB/s, CPU total 100% file copy (ext2 file system) seq. read (cp testfile /dev/null) - 54 МB/s, CPU total 45 % seq. write (cp /dev/zero testfile) - 60 МB/s, CPU total 93 % hdparm [root@RH9 3w]# hdparm -tT /dev/sda /dev/sda: Timing buffer-cache reads: 128 MB in 1.94 seconds = 65.98 MB/sec Timing buffered disk reads: 64 MB in 2.00 seconds = 32.00 MB/sec Under Win2k i get from this array 100 MB/s (66 MB/s at disk end) read and 61 MB/s write with much lower CPU usage. Thanks in advance for any help, sorry for my poor English. -- Best regards, Alex mailto:alexver5@xxxxxxx
