Bryan-TheBS-Smith wrote:
>
> "D. Stimits" wrote:
> > I do not refer to logical append of disks as RAID0. RAID0 should be
> > stripped, and very good performance, splitting the load over all the
> > disks evenly (with no redundancy if anything fails). RAID5 will probably
> > make write performance worse, and read performance much better. RAID
> > definitions seem to have all kinds of different terminology depending on
> > who you talk to, but the linux software RAID docs I've seen always refer
> > to RAID0 as striped and high performance, not as merely a logical
> > append.
>
> Also note that some RAID-1 controllers don't do load balancing
> (especially those "trick BIOS" ones that are really doing it in
> software), and aren't any faster. Even many software RAID-1
> implementations are limited in what they can do because it is done at
> the CPU instead of at the ASIC or microcontroller. 3Ware seems to have
> best RAID-1 implementation in this case (see StorageReview.COM for
> tests).
I always use SCSI systems that allow combined writes. But the original
post said the drives were U160 scsi; I didn't catch whether this was
software RAID or using a hardware controller, but I think it said
hardware controller. So that leads to the question of whether the
hardware controller is able to spread out reads or writes. (my systems
were also all SMP, which might matter, but the 30 MB/sec spec on single
drives I mentioned is non-RAID with an intelligent hot swap backplane)
D. Stimits, stimits@xxxxxxxxxx
>
> -- TheBS
>
> --
> Bryan "TheBS" Smith mailto:b.j.smith@xxxxxxxx chat:thebs413
> Engineer AbsoluteValue Systems, Inc. http://www.linux-wlan.org
> President SmithConcepts, Inc. http://www.SmithConcepts.com
> ----------------------------------------------------------------
> "The [US] Constitution guarantees you Free, not Fair. 'Fair' is
> a socialist concept." -- Shawn McMahon
|