lkml.org 
[lkml]   [1998]   [Nov]   [26]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
SubjectRe: Absolutely horrid IDE performance...
Date
Another piece of *very* telling information was found on a very different
machine with a different model IBM UDMA drive. There seems to be a huge
overhead for each sequential read operation. If we increased the block size
and left the dd bytes at 16megs total, the performance started increasing
quickly. Reading one 16meg chunk gave close to 4meg/sec (max on the drive
the way it is setup). 256 64k reads gave half the performance.

Brian Macy

----- Original Message -----
From: George <greerga@nidhogg.ham.muohio.edu>
To: Brian Macy <bmacy@sunshinecomputing.com>
Cc: <linux-kernel@vger.rutgers.edu>
Sent: Wednesday, November 25, 1998 9:58 AM
Subject: Re: Absolutely horrid IDE performance...


>On Tue, 24 Nov 1998, Brian Macy wrote:
>
>>On /dev/hda:
>>hdparm gives 6.68meg/sec
>>time dd if=/dev/hda of=/dev/null bs=4096 count=16384 gives well under
>>1meg/sec (the more transfered the slower)
>
>I have the same drive and under DMA mode2 (HX chipset PIIX3) I get:
> Timing buffered disk reads: 64 MB in 6.67 seconds = 9.60 MB/sec
> Timing buffered disk reads: 64 MB in 7.15 seconds = 8.95 MB/sec
> Timing buffered disk reads: 64 MB in 6.64 seconds = 9.64 MB/sec
>
>>/var/log/dmesg
>>ide: i82371 PIIX (Triton) on PCI bus 0 function 57
>> ide0: BM-DMA at 0xffa0-0xffa7
>>hda: IBM-DTTA-371010, 9641MB w/465kB Cache, CHS=1024/255/63, UDMA
>
>PIIX3: IDE controller on PCI bus 00 dev 39
>PIIX3: not 100%% native mode: will probe irqs later
> ide1: BM-DMA at 0xf008-0xf00f, BIOS settings: hdc:pio, hdd:pio
>hdc: IBM-DTTA-371010, 9641MB w/465kB Cache, CHS=19590/16/63, (U)DMA
>
>Except I don't use LBA and I'm not in UDMA mode and this is 2.1.129.
>
>> Model=BI-MTDAT3-1710 0 , FwRev=7TO77AB0, SerialNo=
>> Config={ HardSect NotMFM HdSw>15uSec Fixed DTR>10Mbs }
>> RawCHS=16383/16/63, TrkSize=0, SectSize=0, ECCbytes=34
>> BuffType=3(DualPortCache), BuffSize=465kB, MaxMultSect=16, MultSect=off
>> DblWordIO=no, maxPIO=2(fast), DMA=yes, maxDMA=2(fast)
>> CurCHS=16383/16/63, CurSects=16514064, LBA=yes, LBAsects=19746720
>> tDMA={min:120,rec:120}, DMA modes: sword0 sword1 sword2 mword0 mword1
>>mword2
>> IORDY=on/off, tPIO={min:240,w/IORDY:120}, PIO modes: mode3 mode4
>
>Same here. It also does UDMA 0-2 which your hdparm doesn't show.
>
>>#time dd if=/dev/hda of=/dev/null bs=4096 count=16384
>>0.01user 0.38system 2:13.61elapsed 0%CPU (0avgtext+0avgdata 0maxresident)k
>>0inputs+0outputs (85major+11minor)pagefaults 0swaps
>>#time dd if=/dev/sda of=/dev/null bs=4096 count=16384
>>0.00user 0.20system 0:13.79elapsed 1%CPU (0avgtext+0avgdata 0maxresident)k
>>0inputs+0outputs (81major+11minor)pagefaults 0swaps
>
>I use: hdparm -m16 -c1 -d1 -u1 -X34 -W1 /dev/hd[ac]
>(-m16 is redundant though)
>
>0.05user 58.40system 1:15.72elapsed 77%CPU (0avgtext+0avgdata
0maxresident)k
>0inputs+0outputs (84major+12minor)pagefaults 0swaps
>
>Not sure about that, but a bonnie comparison against a 1GB Maxtor (hda):
> -------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
>IBM 100 846 95.0 10626 48.1 3535 37.9 6463 70.4 14288 54.4 1622.1
30.8
>Maxtor 100 3871 41.8 3863 15.5 1543 12.9 4032 43.5 6188 20.9 798.0
14.6
>
>-George
>


-
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 13:45    [W:0.038 / U:0.032 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site