the extra spindles aren't being used because the benchmark isn't set up to perform the test on 3 different theads each of which process one third of the file concurrently at 3 different offsets where the data is distributed on 3 different platters. This illustrates the danger of using load generating benchmarks instead of a system snapshot running your apps with your data... garbage in = garbage out :) -----Original Message----- From: "Nat Makarevitch" <nat@xxxxxxxxxxxxxxx> Subj: Re: Linux MD RAID 5 Benchmarks Across (3 to 10) 300 Gigabyte Veliciraptors Date: Wed Jun 11, 2008 12:06 pm Size: 1K To: "linux-raid@xxxxxxxxxxxxxxx" <linux-raid@xxxxxxxxxxxxxxx> Justin Piszcz <jpiszcz <at> lucidpixels.com> writes: > Ever wonder what kind of speed is possible with 3 disk, 4,5,6,7,8,9,10-disk RAID5s? > Here are the bonnie++ results: > http://home.comcast.net/~jpiszcz/20080607/raid5-benchmarks-3to10-veliciraptors/veliciraptor-raid.html Why does the amount of spindles has nearly no effect on the amount of seeks per second? 3 disks: 713.9 seeks/s (AFAIK the Raptor works at 10000 rpm, getting 230+ seeks/s is astonishing) 10 disks: 705.5 seeks/s (same as 3 disks?!) Did I miss something? Or did you use a very large stripe size (to the point of forbiding the 16 GB file used to span over all spindles?)? Or is it some glitch in the RAID code (I don't think so, on a RAID10 with 10 low-end disk I obtained ~1000 IOPS: http://www.makarevitch.org/rant/raid/#3wmd)? -- To unsubscribe from this list: send the line "unsubscribe linux-raid" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html -- To unsubscribe from this list: send the line "unsubscribe linux-raid" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
