lkml.org 
[lkml]   [1998]   [Dec]   [13]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: Dumb question: Which is "better" SCSI or IDE disks?
On Wed, 9 Dec 1998, Gregory Maxwell wrote:
> What? Join the real world but. Raid-1 doubles read throughput and the
> slowdown of writes is negligable.

Sorry, but I don't find that in actual benchmarks. Perhaps it is true
with some RAID controllers, but definitely not with the ICP-Vortex GDT
6517. I get the same throughput from RAID1 on that setup that I get
from the bare Seagate Cheetah on that setup (no RAID). The only way I
can figure that they could make it faster would be to operate dual
elevators (i.e. on one drive reading from the bottom half, on the
other drive reading from the top half), which would reduce
random-access seek times but which would not improve streaming
performance one whit. They can't do striping because the drives have
to be identical (mirrored), meaning meaning that the head would have
to move to access the next track of data no matter which drive it is
coming from. There could be some improvement from read-ahead (i.e.,
starting the other head moving while we're reading from the first
head), but in real life that's a minimal improvement.

On the other hand, you are correct that the slowdown of writes is
neglible. Even with the write cache turned off the GDT seems capable of
issuing two writes in quick succession without slowing down the
main processor. (The GDT has an i960 on-board, so the main processor isn't
involved once it gets the request to the GDT). With the write-cache
turned on there's no slowdown at all. I suspect that this would be
different if the SCSI bus was heavily loaded, but it's not -- a 40mb/sec
SCSI bus will feed two Cheetahs at full speed without flinching.

--
Eric Lee Green eric@linux-hw.com http://www.linux-hw.com/~eric
"Linux represents a best-of-breed UNIX, that is trusted in mission
critical applications..." -- internal Microsoft memo


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 13:46    [W:0.133 / U:0.316 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site