Hi, Happy new year! I have a question on raid 1 performance. I setup a simple two disk raid1 on a p4 pc and used iometer to measure the random write performance. *the disks i used are two same seagate 18gb 15k rpm scsi disks. disks are connected to different channels in one aic7xxx controller. *the raid1 device is md0. *the iometer workload is 100% random 100% write 4k request size, outstanding request number 1. i run test with 1 hours. *i changed the raid1 code slightly. in raid1_make_request(), i change mirror = conf->mirrors + raid1_read_balance(conf, bh); to mirror = conf->mirrors; so there is no read balance and raid1 only read from the first scsi disk. i ran iometer on /dev/sda, the IO per second is 260. i ran iometer on /dev/md0, the IO per second is 290. can anybody tell me why md0 is faster than sda? i had assumed that md0 can not be faster than sda at this situation. since regularly raid1 can improve read performance but not write performance. so the raid1 write performance should be the same with the slowest device i used. but here sth different happened. and i noticed that even i set to 100% write, there are still many read activities on first scsi disk shown by iostat. one test result is Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn dev3-0 0.15 1.60 1.42 37170 33082 dev8-0 15.60 13.76 64.24 320452 1496416 dev8-1 8.70 0.00 64.24 48 1496416 i think i already choose the request size to be 4k, which is page size, but why here still so many reads? ming - To unsubscribe from this list: send the line "unsubscribe linux-raid" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
