Messages in this thread |  | | From | "John L. Males" <> | Date | Sun, 23 Sep 2001 20:00:53 -0500 | Subject | Re[05]: Linux Kernel 2.2.20-pre10 Initial Impressions |
| |
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Hi Alan,
I have completed some testing. Some as I am not sure what other testing I need to do as will become evident below. I have not copied in Andrea Arcangeli. Based on the results it would be my (limited Linux Kernel knowledge) opinion whatever I am observing is not likely VM related.
The summary of the results are as follows:
Test Case One: ****************
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) - - 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 were instant response from Netscape. Did the test for each kernel twice; and yes that means each of the 6 tests were done from a new boot of the PC.
This is the expected result, but until I did this test I was not aware of any instant Netscape response to this action. My "usual" personal sequence of starting things was different, but traditional as will be noted later. I came up with this test based on some non-browser experiences the latter part of this week.
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 when one clicks the noted button. Did the test for each kernel twice; and yes that means each of the 6 tests were done from a new boot of the PC.
This is sort of the expected results, but until I did this test I was not aware of how long the waint was in absolute terms. My "usual" personal sequence of starting things was different, but traditional as will be noted later. I came up with this test as I wanted to see if the IPChains (2.2 kernels) or IPTables (2.4 kernel) was having an effect on the delays I experienced.
Although I had never know how long it took for this URL list to appear, my memory from my expereinces (have many as been happening for long time) was that with the 2.2.20-pre10 kernel things were really much much longer. Clearly from the tests I have done that is not the case. This test is not exactly how I usually start things up. Normally I have kppp start the IPChains script, then I start Netscape. This test clearly shows some relationship to my ppp0 connection being established and how long Netscape took to respond to viewing a URL list. Note we are not actually selecting a site to go to, hence why test one was done and the very big surprize discovering there was no delay in the action. the test is just to click the list button to see the list, no more than that.
Test Case Three: *******************
Steps used:
- - After the second test with Kernel 2.4.9-ac10 of Test Case Two - - kppp is still up and connected to ISP from Test Case Two - - Closed Netscape completely. - - 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 2.4.9-ac10 as only Kernel did this test with:
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.
A special note for the second test done in this test case. After the first time did test for this test case, did an e2fsck and mount for a mount point for HDA7. The other mount points are all SCSI and started at boot time. The HDA device is not being auto mounted mostly to allow me to work around the various problems with booting, related lilo, but also exist with grub. Again I decided on this varient of the test as I do in fact start the system at times without the HDA drive mounted or the IDE interface disabled when I need to do lilo activities.
Test Case Four: ******************
Steps used:
- - After the second test with Kernel 2.4.9-ac10 of Test Case three - - kppp is still up and connected to ISP from Test Case Two - - Closed Netscape completely. - - Started my IPTables based script that was also set to do a "iptables -Lv" at end to see if that altered the time of the delay as this would often match a suitation I have had in past few weeks on my Linux system. - - 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 2.4.9-ac10 as only Kernel did this test with:
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: *****************
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.) - - Started my IPChains based script that was also set to do a "ipchains -L" at end to see if that altered the time of the delay as this would often match a suitation I have had in past few weeks on my Linux system. - - 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 only was this test done. First time the delay was 2 minutes 3 seconds, the second time it was 2 minutes 15 seconds.
Other Comments *******************
All the kernels were complied using the same gcc compiler supplied with SuSE 6.4. The SuSE 6.4 I am using is up to date as of about April/2001 as far as security fixes and like. It is still using KDE 1.1.2, and XFree 3.3.6. And yes the necessary elements to meet the requirements of the 2.4 kernel are in place. I managed to work the SuSE 7.2 versions of those tools into the SuSE 6.4 system with some nurturing :). Other than that it is basically a SuSE 6.4 system witth the odd application at a much higher level than supplied with SuSE, but these are either console or X based apps. Obviously no KDE 2 apps on this system.
Also of interesting note, during the time delay experienced there was not much, if any CPU activity, so the delay does not seem to be "processing" related. For all intents and purposes I could not observe any internet activity. That is not to say there was none, but not really observable.
In terms of the two 2.2 Kernels, the 2.2.20-pre-10 kernel was build using the config of the 2.2.19 kernel via oldconfig. The only difference was when running oldconfig it asked about one item that was not in the 2.2.19 kernel. My reply was to make the module, even though I had no use for the module.
The 2.2.19 Kernel is the base kernel from kernel.org plus the Openwall patch:
http://www.openwall.com/linux/linux-2.2.19-ow1.tar.gz
Of course there was no Openwall patch made to the 2.2.20-pre10 kernel.
I have attached a screen image of the URL history list that was the subject of this test to give a better understanding of what was being selected to display.
This is not the only area where there are unexpected delays. I am just most familar with this one as it is the one I happen to do almost every day. Now although I am sure there were shorter and different delays, I cannot duplicate it at moment the shorter ones for the activity I do every day. I will keep my wits about me to better observe and see if I can identify what may be related to the amount of time in a delay.
As for the impact of this observation for the 2.2.20-pre10 Kernel, I am in no position to really say given my very lack of familarity with the process, not to mention the kernel itself. As a QA/Testing person I can say that from what I have experienced in the past, and see with the 2.2.20-pre10 kernel on the issue I am experiencing there is likey no change. At this point I would not be able to say if the matter has some kernel element of impact or not.
It is possible I may have omitted some important point(s) in this eMail note, or even omitted in my testing. If so feel free to comment. I will be busy over the wekk and suggestion has it I will need to work next weekend. Be that the case, please be patient if I am very slow in replying to any additional information or clarification that may be required.
Regards,
John L. Males Willowdale, Ontario Canada 23 September 2001 20:00 mailto:jlmales@softhome.net
Subject: Re: Re[03]: Linux Kernel 2.2.20-pre10 Initial Impressions To: jlmales@softhome.net Date sent: Sat, 22 Sep 2001 20:00:06 +0100 (BST) Copies to: alan@lxorguk.ukuu.org.uk (Alan Cox), linux-kernel@vger.kernel.org From: Alan Cox <alan@lxorguk.ukuu.org.uk>
> > Understood, but I actually took my 2.2.19 .config and ran "make > > oldconfig", then "make xconfig" making no changes, just saved it > > Excellent. That makes the data so much more valuable > > > I am likely to do the benchmark tonight to get hard numbers on > > the difference I sense. I am a QA/Testing Specialist, so I am > > all to aware of the importance of keeping the variables all the > > same. My initial background was with assembler back in the real > > core > > memory/keypunch days where I disassembled and heavily modified > > the OS, compiler, assembler, system utilities and wrote a new way > > to load the OS, compilers, etc from scratch to a new disk. Ony > > advising you so you have a sense of my mindset and level of > > understanding. Not current with intimate x86 details or > > assembler, but will someday now that "falt" memory is back! :)) > > > > I will let you know what I find. If in meantime you feel there > > are other things needed or for me to check please let me know and > > I will be most happy to assist. > > I look forward to the results. Can you cc Andrea Arcangeli on them > if they look VM related as Andrea is the 2.2.19 VM person (and now > the 2.4.10pre one)
-----BEGIN PGP SIGNATURE----- Version: PGPfreeware 6.5.8 for non-commercial use <http://www.pgp.com>
iQA/AwUBO66FquAqzTDdanI2EQLaBACg0SD0omLkTiCHBmfZMXXXCKsdjVcAoOi7 dYuaMbqUOnO1/1Fr108/ttbf =dRDV -----END PGP SIGNATURE-----
"Boooomer ... Boom Boom, how are you Boom Boom" Boomer 1985 - February/2000 The following section of this message contains a file attachment prepared for transmission using the Internet MIME message format. If you are using Pegasus Mail, or any another MIME-compliant system, you should be able to save it or view it from within your mailer. If you cannot, please ask your system administrator for assistance.
---- File information ----------- File: ns478.png Date: 23 Sep 2001, 13:40 Size: 12884 bytes. Type: Unknown [unhandled content-type:application/octet-stream] |  |