Subject: Re: [linux-audio-dev] some ide qs [was:Re: Interesting article on SCSI vs. IDE]
From: Frank Neumann (franky_AT_viona.de)
Date: Thu Nov 09 2000 - 11:30:16 EET
Hi,
Jörn wrote:
> does soft-raid on ide disks make sense when you have two disks per
> channel ?
> iirc, ide drives don't disconnect from the bus during seek times -
> don't they block the other drive then ?
On a similar vein, I made some tests yesterday because I was curious to see
how much two IDE devices on the same channel affect each other when accessed
in parallel.
Here's what I did: The first IDE channel of my work PC has got a harddisk as
the master and a DVD-ROM as the slave. The initial test was to see how much
throughput I get from the harddisk:
root_AT_tux:/dev> time dd if=/dev/hda of=/dev/null bs=1024 count=204800
204800+0 records in
204800+0 records out
real 0m16.177s
user 0m0.280s
sys 0m3.990s
That's 200MB in 16.177 seconds, or about 13MB/sec read speed. I think I have
not been fooled by the cache because this system has got only 128MB RAM,
and about 2/3 of that was used for X, Netscape etc. while testing.
Next, the DVD-ROM read speed, reading from an ISO9660 CD:
204800+0 records in
204800+0 records out
real 1m11.989s
user 0m0.390s
sys 0m5.590s
That's about 2.9MB/sec. Ok, now I have those two reads start in parallel; both
tests run for about the same time:
Harddisk:
real 1m11.395s
user 0m0.290s
sys 0m2.220s
DVD-ROM:
real 1m11.923s
user 0m0.250s
sys 0m2.680s
It's funny to notice here that "gkrellm" (a tiny&useful little system monitor)
shows me that the activity of hda is only about a third of what it was when
I read only from the harddisk.
So, this looks like the harddisk "adapts" its read speed to that of the
DVD-ROM. Is this typical behaviour for IDE? Unfortunately I could not easily
tests what numbers I would get if I had two harddisks on the same IDE channel.
By the way, reading from two IDE disks at the same time, one on the primary
and the other on the secondary IDE channel, they disturb each other just
slightly: When reading only from the IDE disk on the secondary channel alone,
I get:
real 0m14.219s
user 0m0.370s
sys 0m3.580s
(that's roughly 14.8MB/sec)
while reading from both simultaneously (both are masters), I get:
real 0m20.100s
user 0m0.410s
sys 0m5.260s
for the hda disk and
real 0m20.077s
user 0m0.260s
sys 0m4.790s
for the hdc disk. That's down to about 10.4MB/sec.
I did all these tests with DMA on (of course), but no further tweaking.
I also tried using a multcount of 16 and I/O support set to "32bit w/sync"
(hdparm -c 3 -w 16 /dev/hda, and the same for /dev/hdc), but this did not
improve numbers.
If there are other things to tweak, I'd love to hear about them.
Ok, I should probably add what system I tested this on: Pentium II-400MHz,
128MB RAM with these devices:
ide0: BM-DMA at 0xd800-0xd807, BIOS settings: hda:DMA, hdb:DMA
ide1: BM-DMA at 0xd808-0xd80f, BIOS settings: hdc:DMA, hdd:pio
hda: QUANTUM FIREBALL EX10.2A, 9787MB w/418kB Cache, CHS=1247/255/63, UDMA
hdb: TOSHIBA DVD-ROM SD-M1202, ATAPI 32X DVD-ROM drive, 256kB Cache
hdc: QUANTUM FIREBALLlct08 17, 16555MB w/418kB Cache, CHS=33637/16/63, UDMA
Regards,
Frank
This archive was generated by hypermail 2b28 : Thu Nov 09 2000 - 12:16:41 EET