Hi, I have a problem with Zope speed. I've done benchmarking in two machines, with the same hardware. (K7 700 MHz, 128 MB, 20GB HD) Windows machine : W 2000 Professional with Zope for Windows Linux machine : Linux RedHat 6.1, with Apache Server 1.3.12 and Zope 2.2.1 Windows 2000----------------------------------------------------------- Server Hostname: windows Server Port: 80 Document Path: /index_html Document Length: 486 bytes Concurrency Level: 25 Time taken for tests: 1.807 seconds Complete requests: 1000 Failed requests: 137 (Connect: 0, Length: 137, Exceptions: 0) Non-2xx responses: 864 Total transferred: 660974 bytes HTML transferred: 473471 bytes Requests per second: 553.40 Transfer rate: 365.79 kb/s received Connnection Times (ms) min avg max Connect: 0 0 10 Processing: 11 42 356 Total: 11 42 366 ----------------------------------------------------------------------------------- Linux--------------------------------------------------------------------------- Server Hostname: linux Server Port: 8088 Document Path: /Contactos Document Length: 396 bytes Concurrency Level: 25 Time taken for tests: 7.688 seconds Complete requests: 1000 Failed requests: 0 Total transferred: 595000 bytes HTML transferred: 396000 bytes Requests per second: 130.07 Transfer rate: 77.39 kb/s received Connnection Times (ms) min avg max Connect: 0 8 224 Processing: 102 181 54 Total: 102 189 278 ------------------------------------------------------------------------------------------------------- There's a big difference in the number of requests per second, four times more in the Windows machine. Anyone knows why? Does it mean that Zope running in Windows is better than in Linux? /_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/ Aitor Grajal Crespo Director General e-mail: Aitor.Grajal@inetsysonline.com INETSYS, S.L. ETSI Informática Ctra. Colmenar, km. 15,500 28049 MADRID SPAIN Tfno : (34) 91 348 22 66 - Fax : (34) 91 348 22 66 http://www.inetsysonline.com /_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/
Hi all. Correct me, if i am wrong, but AFAIK ZServer (coming with Zope) is multi-threaded by default, whereas Apache/Zope uses PCGI or FCGI and is *not* multi-treaded. Best regards, Ansgar W. Konermann eMail: ak26@inf.tu-dresden.de WWW: http://www.inf.tu-dresden.de/~ak26
Why do you test the one with ZServer and the other with Apache? If you want correct results you've to test both with ZServer or both with Zope behind Apache. Not? Tom.
both of them go through the apache server, cos I have installed a proxy server (the reason is that windows is located on a different machine, and the proxy redirects to the windows zserver) thanks. *********** REPLY SEPARATOR *********** On 06/10/2000 at 17:27 Tom Deprez wrote:
Why do you test the one with ZServer and the other with Apache? If you want correct results you've to test both with ZServer or both with Zope behind Apache. Not?
Tom.
/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/ Aitor Grajal Crespo Director General e-mail: Aitor.Grajal@inetsysonline.com INETSYS, S.L. ETSI Inform�tica Ctra. Colmenar, km. 15,500 28049 MADRID SPAIN Tfno : (34) 91 348 22 66 - Fax : (34) 91 348 22 66 http://www.inetsysonline.com /_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/
To get ACCURATE timings, you need to create the situations almost identical, other than the operating system. So, that means: 1) Windows and Linux tunings should be similar; are you running extra daemons on the Linux box that Windows isn't (smtp, ftp, etc)? 2) The server architecture should be the same (identical hardware) 3) The software architecture should be the same; are you running both Zope and Apache on each machine, or are you spliting these services up somehow. A server with Apache+Zope, and then another server with just Zope will NOT match up. Just keep that in mind... Knight knight@phunc.com On Fri, 6 Oct 2000, Aitor Grajal Crespo wrote:
both of them go through the apache server, cos I have installed a proxy server (the reason is that windows is located on a different machine, and the proxy redirects to the windows zserver) thanks.
*********** REPLY SEPARATOR ***********
On 06/10/2000 at 17:27 Tom Deprez wrote:
Why do you test the one with ZServer and the other with Apache? If you want correct results you've to test both with ZServer or both with Zope behind Apache. Not?
Tom.
/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/
Aitor Grajal Crespo Director General e-mail: Aitor.Grajal@inetsysonline.com
INETSYS, S.L. ETSI Inform�tica Ctra. Colmenar, km. 15,500 28049 MADRID SPAIN Tfno : (34) 91 348 22 66 - Fax : (34) 91 348 22 66 http://www.inetsysonline.com
/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/
_______________________________________________ Zope maillist - Zope@zope.org http://lists.zope.org/mailman/listinfo/zope ** No cross posts or HTML encoding! ** (Related lists - http://lists.zope.org/mailman/listinfo/zope-announce http://lists.zope.org/mailman/listinfo/zope-dev )
On Fri, 6 Oct 2000, Tom Deprez wrote:
Why do you test the one with ZServer and the other with Apache? If you want correct results you've to test both with ZServer or both with Zope behind Apache. Not?
Tom.
Exactly, or atleast both ZServer. You can definately attribute some latency due to the Apache integration... (minimal, but still needs to be noted!) Knight knight@phunc.com
Concurrency Level: 25 Time taken for tests: 1.807 seconds Complete requests: 1000 Failed requests: 137 (Connect: 0, Length: 137, Exceptions: 0) Non-2xx responses: 864 Total transferred: 660974 bytes HTML transferred: 473471 bytes Requests per second: 553.40 Transfer rate: 365.79 kb/s received
Linux--------
Concurrency Level: 25 Time taken for tests: 7.688 seconds Complete requests: 1000 Failed requests: 0 Total transferred: 595000 bytes HTML transferred: 396000 bytes Requests per second: 130.07 Transfer rate: 77.39 kb/s received
There's a big difference in the number of requests per second, four times more in the > Windows machine. Anyone knows why? Does it mean that Zope running in Windows is better than in Linux?
No - look at your results. Every request failed on the win32 box (and they all succeeded on linux). You have some sort of problem in your windows setup, and you'll always get higher throughput for errors than you can for completed requests... Brian Lloyd brian@digicool.com Software Engineer 540.371.6909 Digital Creations http://www.digicool.com
Don't you all Zope mailinglist participants think that Aitor Grajal owe us all an excuse in the form of a NEW TEST with the Win32 ZServer running properly? ;-) Especially to all Linux supporters for the badwill of the emails title. I would really be interested in its correct results. Cheers everyone
Failed requests: 137
No - look at your results. Every request failed on the win32 box (and they all succeeded on linux). You have some sort of problem in your windows setup, and you'll always get higher throughput for errors than you can for completed requests...
Brian Lloyd brian@digicool.com Software Engineer 540.371.6909 Digital Creations http://www.digicool.com
_______________________________________________ Zope maillist - Zope@zope.org http://lists.zope.org/mailman/listinfo/zope ** No cross posts or HTML encoding! ** (Related lists - http://lists.zope.org/mailman/listinfo/zope-announce http://lists.zope.org/mailman/listinfo/zope-dev )
Hi all! Peter Bengtsson wrote:
Don't you all Zope mailinglist participants think that Aitor Grajal owe us all an excuse in the form of a NEW TEST with the Win32 ZServer running properly?
Definitely, yea!
I would really be interested in its correct results.
Me too. :-) Cheerio! Ansgar W. Konermann eMail: ak26@inf.tu-dresden.de WWW: http://www.inf.tu-dresden.de/~ak26
Does anyone have a simple example of creating and setting the properties of a node within an xml document, using xmldocument and external methods. thanks Matt
On Sat, 07 Oct 2000 01:57:40 +0200, "Ansgar W. Konermann" <konermann@inf.tu-dresden.de> wrote:
Definitely, yea!
I would really be interested in its correct results.
We have been stress testing our Zope application on NT and Linux. Our conclusion is that OS is not a factor in Zope performance. Toby Dickenson tdickenson@geminidataloggers.com
On Sat, 07 Oct 2000 01:57:40 +0200, "Ansgar W. Konermann" <konermann@inf.tu-dresden.de> wrote:
Definitely, yea!
I would really be interested in its correct results.
We have been stress testing our Zope application on NT and Linux. Our conclusion is that OS is not a factor in Zope performance.
Toby Dickenson tdickenson@geminidataloggers.com
Can we see some profiling data from that to support your statement? Regards, Knight knight@phunc.com
We have been stress testing our Zope application on NT and Linux. Our conclusion is that OS is not a factor in Zope performance.
Can we see some profiling data from that to support your statement?
Its all application-specific, and theres nothing I can release at the moment. Essentially the timings were all within a 15% band after a little tweaking. Personally I don't think this result is at all suprising: In a loaded Zope system the processor spends all its time executing the python VM, not OS code. Toby Dickenson tdickenson@geminidataloggers.com
On Mon, Oct 09, 2000 at 01:39:41PM +0100, Toby Dickenson wrote:
We have been stress testing our Zope application on NT and Linux. Our conclusion is that OS is not a factor in Zope performance.
Its all application-specific, and theres nothing I can release at the moment. Essentially the timings were all within a 15% band after a little tweaking.
Personally I don't think this result is at all suprising: In a loaded Zope system the processor spends all its time executing the python VM, not OS code.
...which takes us to the obvious question - howto make python faster? Has anyone tried diffrent compiler flags for python etc. to make it faster? -- Riku Voipio riku.voipio@tietoenator.com 09-862 60764
I feel zserver somewhat dumber on windows 95 than on linux. a p200 machine with windows 95 is considerably slower inrentering pages than a p166 with linux. Toby Dickenson wrote:
On Sat, 07 Oct 2000 01:57:40 +0200, "Ansgar W. Konermann" <konermann@inf.tu-dresden.de> wrote:
Definitely, yea!
I would really be interested in its correct results.
We have been stress testing our Zope application on NT and Linux. Our conclusion is that OS is not a factor in Zope performance.
Toby Dickenson tdickenson@geminidataloggers.com
_______________________________________________ Zope maillist - Zope@zope.org http://lists.zope.org/mailman/listinfo/zope ** No cross posts or HTML encoding! ** (Related lists - http://lists.zope.org/mailman/listinfo/zope-announce http://lists.zope.org/mailman/listinfo/zope-dev )
-- Manuel Amador (Rudd-O)
Don't you all Zope mailinglist participants think that Aitor Grajal owe us all an excuse in the form of a NEW TEST with the Win32 ZServer running properly? ;-) Especially to all Linux supporters for the badwill of the emails title.
I would really be interested in its correct results.
Cheers everyone
Especially with all this press I've seen in publications over the last month or two regarding Windows 2000 is outperforming *nix. (over my dead _dead_ body) Knight
Failed requests: 137
No - look at your results. Every request failed on the win32 box (and they all succeeded on linux). You have some sort of problem in your windows setup, and you'll always get higher throughput for errors than you can for completed requests...
Brian Lloyd brian@digicool.com Software Engineer 540.371.6909 Digital Creations http://www.digicool.com
_______________________________________________ Zope maillist - Zope@zope.org http://lists.zope.org/mailman/listinfo/zope ** No cross posts or HTML encoding! ** (Related lists - http://lists.zope.org/mailman/listinfo/zope-announce http://lists.zope.org/mailman/listinfo/zope-dev )
_______________________________________________ Zope maillist - Zope@zope.org http://lists.zope.org/mailman/listinfo/zope ** No cross posts or HTML encoding! ** (Related lists - http://lists.zope.org/mailman/listinfo/zope-announce http://lists.zope.org/mailman/listinfo/zope-dev )
which publications? it sounded like .... dunno, if there is a publication that dares to say that, they ARE lying. knight wrote:
Don't you all Zope mailinglist participants think that Aitor Grajal owe us all an excuse in the form of a NEW TEST with the Win32 ZServer running properly? ;-) Especially to all Linux supporters for the badwill of the emails title.
I would really be interested in its correct results.
Cheers everyone
Especially with all this press I've seen in publications over the last month or two regarding Windows 2000 is outperforming *nix. (over my dead _dead_ body)
Knight
Failed requests: 137
No - look at your results. Every request failed on the win32 box (and they all succeeded on linux). You have some sort of problem in your windows setup, and you'll always get higher throughput for errors than you can for completed requests...
Brian Lloyd brian@digicool.com Software Engineer 540.371.6909 Digital Creations http://www.digicool.com
_______________________________________________ Zope maillist - Zope@zope.org http://lists.zope.org/mailman/listinfo/zope ** No cross posts or HTML encoding! ** (Related lists - http://lists.zope.org/mailman/listinfo/zope-announce http://lists.zope.org/mailman/listinfo/zope-dev )
_______________________________________________ Zope maillist - Zope@zope.org http://lists.zope.org/mailman/listinfo/zope ** No cross posts or HTML encoding! ** (Related lists - http://lists.zope.org/mailman/listinfo/zope-announce http://lists.zope.org/mailman/listinfo/zope-dev )
_______________________________________________ Zope maillist - Zope@zope.org http://lists.zope.org/mailman/listinfo/zope ** No cross posts or HTML encoding! ** (Related lists - http://lists.zope.org/mailman/listinfo/zope-announce http://lists.zope.org/mailman/listinfo/zope-dev )
-- Manuel Amador (Rudd-O)
participants (10)
-
Aitor Grajal Crespo -
Ansgar W. Konermann -
Brian Lloyd -
knight -
Manuel Amador (Rudd-O) -
matt -
Peter Bengtsson -
Riku Voipio -
Toby Dickenson -
Tom Deprez