On Monday 06 September 2004 13.38, Dmitry Golubev wrote: > > PS: i wouldn't recommend software raid 5 if you care about performance. > > i am going to convert one of my raid-5 machines (4 x 80GB barracudas) > > to raid-1 (2 x 200GB barracudas) very soon because i'm unhappy with the > > performance(*)...if i had a spare approx $600AUD, i'd buy an IDE raid > > card with at least 32MB non-volatile cache memory and that would give > > me raid-5 with decent performance, but it's just not worth that much to > > me for a workstation. > > Any tests on RAID-5 being slower than RAID-1 (a problem in software-raid > implementation?) ? I have always thought RAID-5 is the fastest... But I
RAID5 does need more computation than RAID1, so if you have a CPU bottleneck RAID5 will always be slower (assuming RAID5 is computed on the main CPU.) For reading, RAID5 is very fast, since access can be spread over many disks. OTOH each read from RAID5 touches n - 1 disks, so concurrent reads tend to be not as fast as some may expect them to be. Big caches are mandatory here! For writing, RAID5 tends to be noticeably slower than RAID1, especially for writes smaller than stripe size, because a write actually is a read-recompute-write cycle. If you have battery backed RAM on your RAID controller, or you're just willing to risk it, RAID5 can profit a lot from a big write cache. (And even with read cache only, bigger is better for writing, too, as the non-written part of a stripe might just be in the cache.) For RAID 1, you can get quite close to the theoretical max bandwidth: 1 x disk speed on writing, and 2 x disk speed for reading. (Of course, available bus bandwidth etc. will limit this, and there is some minimal management overhead, but RAID1 is quite simple, after all.) -- Today is Prickle-Prickle, the 30th day of Bureaucracy in the YOLD 3170
pgpFjxn3o7XKh.pgp
Description: PGP signature
