Date: Sat, 6 Nov 1999 04:33:49 -0800 From: John-Mark Gurney <gurney_j@efn.org> To: Rob Snow <RSnow@lgc.com> Cc: freebsd-questions@freebsd.org Subject: Re: writing much slower than reading... Message-ID: <19991106043349.28321@hydrogen.fircrest.net> In-Reply-To: <9F147E391A3FD111B9A800805F356C52E25974@lgcadev001.zycor.lgc.com>; from Rob Snow on Sat, Nov 06, 1999 at 05:48:38AM -0600 References: <9F147E391A3FD111B9A800805F356C52E25974@lgcadev001.zycor.lgc.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Rob Snow scribbled this message on Nov 6:
[-- Warning: windows-1252 is not compatible with your display.]
> Wow, I'm impressed! I bought the 10.1G 7200 last year for my Linux box, and
> I thought it was fast! Do me a favor, run Bonnie on it, with a 256M file. (
> bonnie -s 256 )
>
> Bonnie is in ports/benchmarks.
running now... the numbers will be significantly slower because of the
slow processor on the machine... I can not max out the drive when I'm
going through the file system because the overhead is just to great...
when going through the file system, I see 99% usage by the system...
the results:
# bonnie -s 256
File './Bonnie.4177', size: 268435456
-------Sequential Output-------- ---Sequential Input-- --Random--
-Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
Machine MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU
256 9333 97.1 15535 96.5 4591 34.7 9101 97.5 17984 92.3 123.2 4.8
as you notice, both of the block input/output still have 90+% cpu usage..
If I had a faster processor things would probably improve... but I'm
still quite a ways away from what rawio reports....
> Now, on to your problem, am I reading something wrong? It looks like your
> numbers are within 10% of each other.
I should of been more specific... the problem is the system time, not
the actual speed of the drive... notice that when writing it uses 8
times as much processor as when reading...
it's interesting... rawio and dd use about 26% of the cpu to write to
the disk, (plus or minus a percent), rawio when reading uses about the
same cpu (slightly less), but dd when reading uses almost NO cpu, 3.6%..
the system should not have to do anything special, the worse would be
to break it down to 4kb, but that isn't happening else I would get even
worse performance (~12meg/sec)... and at 4kb, the processor isn't pegged
and neither is the drive, is this an IDE limitation?? do the default IDE
drivers not enable UDMA/66 which I know the MVP3 and the IBM drive
support?
benchmarks:
# time rawio -c 128 -w /dev/rwd0s1g
Random read Sequential read Random write Sequential write
ID K/sec /sec K/sec /sec K/sec /sec K/sec /sec
anon 28706.9 438
38.42 real 0.64 user 9.47 sys
# time rawio -c 128 -r /dev/rwd0s1g
Random read Sequential read Random write Sequential write
ID K/sec /sec K/sec /sec K/sec /sec K/sec /sec
anon 29258.5 446
37.71 real 0.41 user 8.94 sys
# time dd if=/dev/zero of=/dev/rwd0s1g bs=64k count=16384
16384+0 records in
16384+0 records out
1073741824 bytes transferred in 50.254879 secs (21365922 bytes/sec)
50.30 real 0.05 user 13.40 sys
# time dd if=/dev/rwd0s1g of=/dev/null bs=64k count=16384
16384+0 records in
16384+0 records out
1073741824 bytes transferred in 45.897254 secs (23394468 bytes/sec)
45.90 real 0.03 user 1.61 sys
# time dd if=/dev/zero of=/dev/rwd0s1g bs=4k count=$((4*65536))
262144+0 records in
262144+0 records out
1073741824 bytes transferred in 85.742358 secs (12522887 bytes/sec)
85.74 real 0.47 user 28.09 sys
dmesg output:
wdc0 at 0x1f0-0x1f7 irq 14 flags 0xa0ffa0ff on isa
wdc0: unit 0 (wd0): <IBM-DPTA-372730>, DMA, 32-bit, multi-block-16
wd0: 26105MB (53464320 sectors), 53040 cyls, 16 heads, 63 S/T, 512 B/S
wd0: ATA INQUIRE valid = 0007, dmamword = 0007, apio = 0003, udma = 041f
> > John-Mark Gurney wrote:
> > >
> > > well, I am working on writing a capture program to do
> 640x480x12bpp@30fps
> > > to a raw disk, but writing to the raw device is SOOO slow... the reason
> > > I say it's slow is the fact that it takes 8 times the system time
> writing
> > > than reading...
> > >
> > > a bit about the system... k6/2-250, 100mhz system bus, pc100 64meg
> dimm,
> > > VIA MVP3 chipset (IDE DMA enabled), IBM-DPTA-372730 hard disk, Hauppauge
> > > WinCast/TV Model 61351 B226, 3.3-RELEASE...
[benchmarks duplicated above]
> > > now, why does it cost SOOO much more processing time to write than
> > > read?? are there plans to fix this slow down? is it possible? can't
> > > we just dma write out of userland since we are blocking on the write?
--
John-Mark Gurney Voice: +1 408 975 9651
Cu Networking
"The soul contains in itself the event that shall presently befall it.
The event is only the actualizing of its thought." -- Ralph Waldo Emerson
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-questions" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?19991106043349.28321>