Linux Kernel 2.2.20-pre10 Initial Impressions

Linux Kernel 2.2.20-pre10 Initial Impressions

[John L. Males]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello,

I am not on the kernel mailing list.  I would appreciate being copied
in on any replies.

Ok, I finially had a chance to compile the 2.2.20-pre10 Kernel and
run it though some basic paces.  I need to do more specific A vs b
(against the 2.2.19 Kernel), but it seems there are some performance
issues.  It is seems especially obvious with Netscape 4.78.  I also
had a odd Xfree error, that may have had some relationship to the
performance issue.  I have to say at this point the issue seems
selective and not a general one, but I need to do a bit more
checking.  I cannot forsee this checking happening until this
weekend.

I seem to also observe some iteresting memory management differences
with the 2.2.20-pre10 kernel vs the 2.2.19 kernel with the Open Wall
patch.  Agian not enough day to day use logged in to give a sense. 
Then if I do sense, the challence to how to quanitify this will
follow.

Just thought you like to know.  Oh, I keep forgetting to ask, is
there any issue related to the superblock work that has been going on
in tha last few version of the 2.4 kernel and a 2.2.19 or other 2.2.x
kernel?  Only asking as seemd to have some very interesting problems
show up back after switching back from  2.4.8 kernel to the 2.2.19
kernel.


Regards,

John L. Males
Willowdale, Ontario
Canada
18 September 2001 23:26
mailto:jlmales@softhome.net

-----BEGIN PGP SIGNATURE-----
Version: PGPfreeware 6.5.8 for non-commercial use <http://www.pgp.com>

iQA/AwUBO6geeOAqzTDdanI2EQJZywCg0yllJ8HWqhQuXdDWq6aSVu5B4EgAn0la
G1uAjzOiUjSd4goAUXRk1oLI
=4K5d
-----END PGP SIGNATURE-----



"Boooomer ... Boom Boom, how are you Boom Boom" Boomer 1985 - February/2000

Re: Linux Kernel 2.2.20-pre10 Initial Impressions

[Alan Cox]

> Ok, I finially had a chance to compile the 2.2.20-pre10 Kernel and
> run it though some basic paces.  I need to do more specific A vs b
> (against the 2.2.19 Kernel), but it seems there are some performance
> issues.  It is seems especially obvious with Netscape 4.78.  I also
> had a odd Xfree error, that may have had some relationship to the
> performance issue.  I have to say at this point the issue seems
> selective and not a general one, but I need to do a bit more
> checking.  I cannot forsee this checking happening until this
> weekend.

There are to all intents no VM changes of any kind between 2.2.19 and
2.2.20pre10, so it would be interesting to compare configure options
and see what else might be different

Re: Linux Kernel 2.2.20-pre10 Initial Impressions

[Tim Moore]

"John L. Males" wrote:
> ...
> Ok, I finially had a chance to compile the 2.2.20-pre10 Kernel and
> run it though some basic paces.  I need to do more specific A vs b
> (against the 2.2.19 Kernel), but it seems there are some performance
> issues.  It is seems especially obvious with Netscape 4.78.  I also
> had a odd Xfree error, that may have had some relationship to the
> performance issue.
> ...

FWIW, I've been using ns 4.78 since August 20 on the listed kernels with
no noticable change.  All kernels have Andre's IDE patch.

If you have a specific test script I can give it a run.

2.2.19pre17, 2.2.20pre{6,9,10}
XFree86-3.3.6-29 (SVGA driver)
ide.2.2.19.05042001.patch
communicator-v478-us.x86-unknown-linux2.2.tgz

System is Athlon 850 (CONFIG_M686=y) on an Abit KA7.

00:00.0 Host bridge: VIA Technologies, Inc.: Unknown device 0391 (rev
02)
00:01.0 PCI bridge: VIA Technologies, Inc.: Unknown device 8391
00:07.0 ISA bridge: VIA Technologies, Inc. VT82C686 [Apollo Super] (rev
22)
00:07.1 IDE interface: VIA Technologies, Inc. VT82C586 IDE [Apollo] (rev
10)
00:07.4 Host bridge: VIA Technologies, Inc. VT82C686 [Apollo Super ACPI]
(rev 30)

rgds,
tim.
--

Re: Linux Kernel 2.2.20-pre10 Initial Impressions

[Mike Fedyk]

On Sun, Sep 23, 2001 at 08:00:53PM -0500, John L. Males wrote:
Content-Description: Mail message body
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> Hi Alan,
> 
> Test Case One:
> ****************
> 
> - - Start Netscape 4.78 from desktop ICON
> - - Click the pull down arrow on far right side of the "Go To:" (AKA
> URL line) of Netscape 4.78.
> 
> Results:
> All were instant response from Netscape.  Did the test for each

> Test Case Two:
> *****************
> 
> Steps used:
> 
> - - Boot System
> - - Log onto non-root user id.
> - - "startx" (KDE desktop started)
> (applications prestart from last log off include kppp, kpppload,
> xosview, gnome terminal, khrono, 3 kfm's)
> - - Had kppp "connect" to my ISP (dial up model 56KB, really 49,300
> most of time, some connects 48,000, note kppp is configured to start
> up ntpdate to sync with some time servers.)
> - - Start Netscape 4.78 from desktop ICON
> - - Click the pull down arrow on far right side of the "Go To:" (AKA
> URL line) of Netscape 4.78.
> 
> Results:
> 
> For Kernel 2.2.19OWL, 2.2.20-pre10, 2.4.9-ac10
> 
> All times were 2 minutes 3 seconds to get the URL list that appears
> 
> 

> Test Case Three:
> *******************
> 
> For the two times this test case was also done, still a delay of 2
> minutes 3 seconds from the time clicked the button until the URLs
> appeared.

> Test Case Four:
> ******************
> 
> For the one time I did this test case, a delay of 2 minutes 19
> seconds from the time clicked the button until the URLs appeared.
> 
> 

> Test Case Five:
> *****************
> For Kernel 2.2.19OWL only was this test done.  First time the delay
> was 2 minutes 3 seconds, the second time it was 2 minutes 15 seconds.
> 
> 

I think you're right, it's not a kernel issue.  You should run "tcpdump -qni
ppp0" in a xterm before you start netscape.

You will probably see dns requests going out.  You should check to make sure
that you are not blocking incomming udp ports (1024-5000 for bind, not sure
about resolver...) as that would lengthen the response time considerably if
only a few are open, and completely stop you if all are blocked.

Mike

Re[07]: Linux Kernel 2.2.20-pre10 Initial Impressions

[John L. Males]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Mike,

Thanks for your reply to my posting of test results.  I guess in the
future I had better keep placing at the start of my eMail note I am
not on the kernel mailing list.  therefore be appreciated if any
reply also copies me in as well.

Ok, you make a good point in your suggestion re:

You will probably see dns requests going out. You should check to
make sure that you are not blocking incomming udp ports (1024-5000
for bind, not sure about resolver...) as that would lengthen the
response time considerably if only a few are open, and completely
stop you if all are blocked. 


Although I would not of thought of that specifically, I did do a test
where I had no firewall or filtering started or in place previously. 
Yes, not how one should connect to the internet these days, but just
done for the purposes of the test and had a limited time window it
was done in.  So give that type of test, I really be interested in
your thoughts on how this point could be a factor.  I am not saying
it cannot be, as I really do not have enough knowledge in this area
to know.  The purpose of the test was to determine if either IPChains
or IPTables/Netfilter had any relationship to the delay.  Based on
the results I would think not, but again just my extremely limited
knowledge of the details of IP/TCP/UDP/DNS and kernel matters.

I will try the "tcpdump -qni ppp0" suggestion you made and see if it
help shed some light on the matter.  If the information is beyond my
understanding I will post it so you can look at it and offer your
analysis of the tcpdump information.

My schedule is a bit tight this week, so I may not get to the test
until the weekend if I do not have to work the weekend.  Failing that
I will try later in the week once my current assignment is complete,
or a week this weekend if the assignment I am on gets extended again
for a few days/week.

Again Mike thanks kindly for taking the time to reply and your
thoughts and observations.  You have given me a few more things to
chew on in my thoughts.  I will post any additional thoughts, tests
results, aor the solution if I determine it on my own.


Regards,

John L. Males
Willowdale, Ontario
Canada
25 September 2001 23:43
mailto:jlmales@softhome.net


=========================================================
=========================================================

From: Mike Fedyk (mfedyk@matchmail.com)
Date: Tue Sep 25 2001 - 04:41:57 EST 

     Next message: Daniel Phillips: "Re: Linux VM design" 
     Previous message: Alan Cox: "Re: Burning a CD image slow down my
connection" 
     Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] 



On Sun, Sep 23, 2001 at 08:00:53PM -0500, John L. Males wrote: 
Content-Description: Mail message body 
> -----BEGIN PGP SIGNED MESSAGE----- 
> Hash: SHA1 
> 
> Hi Alan, 
> 
> Test Case One: 
> **************** 
> 
> - - Start Netscape 4.78 from desktop ICON 
> - - Click the pull down arrow on far right side of the "Go To:"
> (AKA  URL line) of Netscape 4.78. 
> 
> Results: 
> All were instant response from Netscape. Did the test for each 

> Test Case Two: 
> ***************** 
> 
> Steps used: 
> 
> - - Boot System 
> - - Log onto non-root user id. 
> - - "startx" (KDE desktop started) 
> (applications prestart from last log off include kppp, kpppload, 
> xosview, gnome terminal, khrono, 3 kfm's) 
> - - Had kppp "connect" to my ISP (dial up model 56KB, really 49,300
>  most of time, some connects 48,000, note kppp is configured to
> start  up ntpdate to sync with some time servers.) 
> - - Start Netscape 4.78 from desktop ICON 
> - - Click the pull down arrow on far right side of the "Go To:"
> (AKA  URL line) of Netscape 4.78. 
> 
> Results: 
> 
> For Kernel 2.2.19OWL, 2.2.20-pre10, 2.4.9-ac10 
> 
> All times were 2 minutes 3 seconds to get the URL list that appears
>   
> 
> 

> Test Case Three: 
> ******************* 
> 
> For the two times this test case was also done, still a delay of 2 
> minutes 3 seconds from the time clicked the button until the URLs 
> appeared. 

> Test Case Four: 
> ****************** 
> 
> For the one time I did this test case, a delay of 2 minutes 19 
> seconds from the time clicked the button until the URLs appeared. 
> 
> 

> Test Case Five: 
> ***************** 
> For Kernel 2.2.19OWL only was this test done. First time the delay 
> was 2 minutes 3 seconds, the second time it was 2 minutes 15
> seconds.   
> 
> 

I think you're right, it's not a kernel issue. You should run
"tcpdump -qni 
ppp0" in a xterm before you start netscape. 

You will probably see dns requests going out. You should check to
make sure 
that you are not blocking incomming udp ports (1024-5000 for bind,
not sure 
about resolver...) as that would lengthen the response time
considerably if 
only a few are open, and completely stop you if all are blocked. 

Mike 
- - 
To unsubscribe from this list: send the line "unsubscribe
linux-kernel" in 
the body of a message to majordomo@vger.kernel.org 
More majordomo info at http://vger.kernel.org/majordomo-info.html 
Please read the FAQ at http://www.tux.org/lkml/ 



     Next message: Daniel Phillips: "Re: Linux VM design" 
     Previous message: Alan Cox: "Re: Burning a CD image slow down my
connection" 
     Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] 



This archive was generated by hypermail 2b29 : Tue Sep 25 2001 -
21:00:15 EST 

-----BEGIN PGP SIGNATURE-----
Version: PGPfreeware 6.5.8 for non-commercial use <http://www.pgp.com>

iQA/AwUBO7Fc6uAqzTDdanI2EQJcagCeKuKqtviLg7RHwwqxYZAzahZZfdoAn3mZ
RTrGXUQphk/o91ywxyNzRL5S
=Tbl/
-----END PGP SIGNATURE-----



"Boooomer ... Boom Boom, how are you Boom Boom" Boomer 1985 - February/2000

Re: Linux Kernel 2.2.20-pre10 Initial Impressions

[Tim Moore]

"John L. Males" wrote:
> ...
> Ok, you make a good point in your suggestion re:
> 
> You will probably see dns requests going out. You should check to
> make sure that you are not blocking incomming udp ports (1024-5000
> for bind, not sure about resolver...) as that would lengthen the
> response time considerably if only a few are open, and completely
> stop you if all are blocked.
> ...

strace will instantly validate or eliminate DNS lookup delays as an
issue wo having to sort through tcpdump output.  In my case nscd was a
lifesaver and saved me from running local bind server.

rgds,
tim.
--