On 19/09/12 09:40, Roman Mamedov wrote: > On Wed, 19 Sep 2012 16:25:57 +1000 > NeilBrown <neilb@xxxxxxx> wrote: > >> RAID5/RAID6 write speed is never going to be brilliant, but there probably is >> room for improvement. Hopefully one day I figure out how to effect that >> improvement. > "/sys/block/md*/md/stripe_cache_size" is absolutely the first thing one should > look at when facing a problem with md RAID5/6 write performance. It is > understandable that the default value can't be high (as it consumes a lot of > RAM on large disk counts), but increasing it if you have the RAM can increase > the write speed by 200 to 400%: > > http://peterkieser.com/2009/11/29/raid-mdraid-stripe_cache_size-vs-write-transfer/ > Thank you for your answer. I did try this, it had zero effect. And Neil, thank you for your answer too, I have now put the disks into the other box, and resynced with the 8th disk, and things are fast now. How fast i cannot exactly say, as the filesystem is put on top of dm-crypt, which caps me at about 100MB/s on this cpu, but that would seem to indicate that things are functioning as they should. -- Kasper Sandberg Sandberg Enterprises +45 51944242 http://www.sandbergenterprises.dk -- 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
