Re: IDE/ATAPI

From: Michal Jaegermann (michal@ellpspace.math.ualberta.ca)
Date: Wed Apr 19 2000 - 17:52:38 EST


Mark Hahn writes:

> summary: current IDE disks sustain 15-30 MB/s with minimal overhead,
> with cost asymptotic to $US 8/GB. current SCSI disks are even faster,
> and certainly superior for many-device systems, but have both higher
> entry cost (controllers), and cost 2-3x per GB

I only know that on a system with old SCSI drives and a very current
IDE disk, and with IDE support patches with latest Al Viro patches,
I can do disk intensive operations, like writing 650 MB of iso
image from a file system on the same machine, to a SCSI disk without
batting an eye. The same operation makes this box practically useless
for any other purpose when a target is located on an IDE drive.

I can also use speed 6 without any problems when writing CD from an
image on a SCSI disk and, on the same system and with the same CD
writer, I have to drop that speed to 4 (with rewritables to 2) and keep
my fingers crossed when I am doing that from IDE. No, no problems with
fifo not beeing full enough and I actually wrote many CDs from NSF
mounted images so disk speed is obviously a non-issue.

This IDE disk is "modern" but the motherboard is a bit older. So yes, I
can get a good performance from IDE (where I mean by that real tasks and
not hdparm numbers) if I spent money to keep all hardware on the leading
edge, and I was lucky enough to buy a board without horrible bugs, and
this IDE disk is kind enough to cooperate with that board and the moon
phase is right. I have some chances to achieve that if I keep in my
head all IDE specs and the latest quirks and what versions works with
what version but otherwise it is a crapshot.

This from a point of view of someone who is not breathing IDE night and
day. (Not mentioning this small detail that I have a 10 years old
WREN still in an active use and none of IDE drives I have see so far
lasted even half, to be very generous, of that :-).

  Michal

-
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/



This archive was generated by hypermail 2b29 : Sun Apr 23 2000 - 21:00:16 EST