Re: Sparc has slow disk access
On Thu, 23 Aug 2001, Andrew Sharp wrote:
> Several things. One, you left out a ton of information, as this
> could be the result of many, many things. Where to start?
>
> What kind of disk/controller for each machine/disk?
Hmm, I quote from my previous mail:
> > E250:
> > ...
> > 4x36 GB SCSI
> > PC:
> > ...
> > 20 GB IDE
I´m not really sure if it is UDMA/66 or UDMA/100. (How to check it under
Linux (I have only online documentation preformated for the wrong OS :-(.)
What else information do you need?
> What is the results you get from running bonnie?
E250:
Version 1.01d ------Sequential Output------ --Sequential Input- --Random-
-Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
bse 4G 4482 99 17120 19 7767 22 4296 97 18639 15 293.5 3
------Sequential Create------ --------Random Create--------
-Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
16 348 99 1225 99 10544 99 381 99 10207 100 2138 99
bse,4G,4482,99,17120,19,7767,22,4296,97,18639,15,293.5,3,16,348,99,1225,99,10544,99,381,99,10207,100,2138,99
PC:
Version 1.01d ------Sequential Output------ --Sequential Input- --Random-
-Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
wr-linux02 496M 10889 96 14988 10 5219 3 8076 70 18197 6 89.9 0
------Sequential Create------ --------Random Create--------
-Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
16 569 99 +++++ +++ 28067 97 479 81 +++++ +++ 4486 99
wr-linux02,496M,10889,96,14988,10,5219,3,8076,70,18197,6,89.9,0,16,569,99,+++++,+++,28067,97,479,81,+++++,+++,4486,99
I don´t know why the E250 used 4G Size instead of 496M for the PC because
I used bonnie without any options for both machines. Anyway the result should
reproduce what I noticed in my quite simple test.
> Can you give the output of hdparm for each disk?
PC:
# hdparm /dev/hda
/dev/hda:
multcount = 0 (off)
I/O support = 1 (32-bit)
unmaskirq = 0 (off)
using_dma = 1 (on)
keepsettings = 0 (off)
nowerr = 0 (off)
readonly = 0 (off)
readahead = 8 (on)
geometry = 1240/255/63, sectors = 19932192, start = 0
Sparc:
# hdparm /dev/sda
/dev/sda:
readonly = 0 (off)
geometry = 34732/64/32, sectors = 71132959, start = 0
> Also notice the the x86 box you are talking about is a considerably
> faster processor/memory combination, and no, I'm not talking about
> bogomips.
Yes, but regarding to harddisk IO should it be no such big difference
because here is the disk speed the limiting factor and not the processor
speed in my opinion.
> Try the solaris test again, Binkey, and this time use a filesystem
Sorry, what is Binkey???
> other than /tmp for the output. Try setting up a ufs-log file
> system and see if the performance doesn't exceed that of linux.
Could you in short describe how to do that. Sorry, I didn´t used Solaris
before and I can´t imagine a reason why /tmp should slower than any
other dirs / filesystems should be faster. Shouldn´t /tmp be the fastest
filesystem???? I did not any modifications on solaris install - just
default filesystems.
> That's your homework for today. Tomorrow we'll get to advanced disk
> performance considerations. ~:^)
Sorry for beeing late in solving my homework but I left my place just
after sending my last E-Mail and just beeing back now. I´m looking
foreward to the next lession ;-).
Kind regards
Andreas.
Reply to: