[TEST] 2.4 vs 2.6.2 vs 2.6.2-mm1 vs 2.6.2-rc3-mm1

[TEST] 2.4 vs 2.6.2 vs 2.6.2-mm1 vs 2.6.2-rc3-mm1

[Samium Gromoff]

Here are the tests i`ve promised, and sorry for the delays.

The test machine was a pIII-600/192M RAM/10krpm SCSI drive.

There was three different loads.

the test app whose run time was measured was:

time find / -xdev | \
	bzip2 --compress | bzip2 --decompress | \
	bzip2 --compress | bzip2 --decompress | \
	bzip2 --compress | bzip2 --decompress | \
	cat > /dev/null

the loads were:

Load 1:
	boot options: mem=32M init=/bin/bash
	swapon -a
	run the test

Load 2:
	boot options: mem=48M init=/bin/bash
	swapon -a
	run the test

Load 3:
	boot options: mem=48M
	usual X session, with lots of terminals, emacs and stuff
	the test was run from one of the x terminal emulators

the kernels were:
	2.4.20-pre9, 2.6.2 -- no comments
	2.6.2-rc3-mm1 -- that one didn`t include the Namesys VM patches
	2.6.2--mm1 -- that one _did_ include the Namesys VM patches

results:


		2.4.20-pre9	2.6.2		2.6.2-mm1	2.6.2-rc3-mm1

Load 1
  run1		6.27		9.14		9.42		10.52

Load 2
  run1		3.29		4.42		3.40		3.45
  run2		3.28		4.37		3.39		3.45

Load 3
  run1		4.42		8.39		18.26


short summary:

	2.4 is faster.


regards, Samium Gromoff

Re: [TEST] 2.4 vs 2.6.2 vs 2.6.2-mm1 vs 2.6.2-rc3-mm1

[Samium Gromoff]

At Mon, 09 Feb 2004 17:57:46 +0300,
Samium Gromoff wrote:
[snip]
> the kernels were:
> 	2.4.20-pre9, 2.6.2 -- no comments
> 	2.6.2-rc3-mm1 -- that one didn`t include the Namesys VM patches
> 	2.6.2--mm1 -- that one _did_ include the Namesys VM patches
Small correction -- these are not exactly the Namesys patches, they are
from Nick Piggin (http://www.kerneltrap.org/~npiggin/vm/), and some are
certainly based on the Namesys folks` work...

[snip]
> regards, Samium Gromoff

Re: [TEST] 2.4 vs 2.6.2 vs 2.6.2-mm1 vs 2.6.2-rc3-mm1

[Nick Piggin]

Samium Gromoff wrote:

>Here are the tests i`ve promised, and sorry for the delays.
>
>The test machine was a pIII-600/192M RAM/10krpm SCSI drive.
>
>There was three different loads.
>
>the test app whose run time was measured was:
>
>time find / -xdev | \
>	bzip2 --compress | bzip2 --decompress | \
>	bzip2 --compress | bzip2 --decompress | \
>	bzip2 --compress | bzip2 --decompress | \
>	cat > /dev/null
>
>the loads were:
>
>Load 1:
>	boot options: mem=32M init=/bin/bash
>	swapon -a
>	run the test
>
>Load 2:
>	boot options: mem=48M init=/bin/bash
>	swapon -a
>	run the test
>
>Load 3:
>	boot options: mem=48M
>	usual X session, with lots of terminals, emacs and stuff
>	the test was run from one of the x terminal emulators
>
>the kernels were:
>	2.4.20-pre9, 2.6.2 -- no comments
>	2.6.2-rc3-mm1 -- that one didn`t include the Namesys VM patches
>	2.6.2--mm1 -- that one _did_ include the Namesys VM patches
>
>results:
>
>
>		2.4.20-pre9	2.6.2		2.6.2-mm1	2.6.2-rc3-mm1
>
>Load 1
>  run1		6.27		9.14		9.42		10.52
>
>Load 2
>  run1		3.29		4.42		3.40		3.45
>  run2		3.28		4.37		3.39		3.45
>
>Load 3
>  run1		4.42		8.39		18.26
>
>
>short summary:
>
>	2.4 is faster.
>
>

What are the units? minutes.seconds?

The test is interesting, I'll have to try it. Does it
resemble a workload you're interested in?

It looks like the -mm kernels might have something other
than Nikita's and my VM patches that is affecting times.

Your Load 3 looks quite bad. Does it give decent results?
Is it possibly because the other stuff is getting better
treatment, do you think?

Thanks
Nick

Re: [TEST] 2.4 vs 2.6.2 vs 2.6.2-mm1 vs 2.6.2-rc3-mm1

[Samium Gromoff]

At Tue, 10 Feb 2004 23:53:36 +1100,
Nick Piggin wrote:
> Samium Gromoff wrote:
[snip]
> >		2.4.20-pre9	2.6.2		2.6.2-mm1	2.6.2-rc3-mm1
> >
> >Load 1
> >  run1		6.27		9.14		9.42		10.52
> >
> >Load 2
> >  run1		3.29		4.42		3.40		3.45
> >  run2		3.28		4.37		3.39		3.45
> >
> >Load 3
> >  run1		4.42		8.39		18.26
> >
> >
> >short summary:
> >
> >	2.4 is faster.
> >
> >
> 
> What are the units? minutes.seconds?
Right.

> The test is interesting, I'll have to try it. Does it
> resemble a workload you're interested in?
Basically, yes.

> It looks like the -mm kernels might have something other
> than Nikita's and my VM patches that is affecting times.
> 
> Your Load 3 looks quite bad. Does it give decent results?
> Is it possibly because the other stuff is getting better
> treatment, do you think?
Everything else was basically idle.
The test script generated something like 15 rows of text,
which hardly could be considered noticeable...

> Thanks
> Nick

regards, Samium Gromoff