Re: [PHP-DEV] Simple Apache 1.3 vs 2.0 benchmarks
Uh, but did you benchmark straight cgi's between the two servers? How about straight static pages? Why would you automatically attribute the difference in performance to DSO vs. static issues? It is pretty well known that Apache2 is currently slower than Apache1 for stuff like this because of the added overhead of the bucket brigade implementation and other issues. -Rasmus On Mon, 25 Feb 2002, August wrote: #include as usual, all normal benchmark disclaimers. This is -dev, so let's assume folks have a grasp of the gap between performance benchmarking and the real world. Before flaming please give credit to folks who are able to use this information in a larger and broader context, while recognizing that few of us have the time to implement fantastic benchmarks. Apache 1.3 with php *statically* compiled in gets 1436 reqs/s Apache 2.0 with php as DSO gets 1114 req/s. Apache 2 performs similarlity to Apache 1.3 with php as a DSO. Compiling statically appears helpful, and currently it is not possible to do so with apache 2. Apache 2 supports static modules generally. # ./bin/ab -c 15 -n 10 /hello.php This is ApacheBench, Version 2.0.32 $Revision: 1.87 $ apache-2.0 Server Software:Apache/2.0.32 Server Port:80 Document Path: /hello.php Document Length:12 bytes Time taken for tests: 89.706050 seconds Complete requests: 10 Failed requests:0 Write errors: 0 Requests per second:1114.75 [#/sec] (mean) Time per request: 0.013 [ms] (mean) Transfer rate: 268.89 [Kbytes/sec] received Percentage of the requests served within a certain time (ms) 50% 12 95% 20 # ./bin/ab -c 15 -n 10 /hello.php This is ApacheBench, Version 2.0.32 $Revision: 1.87 $ apache-2.0 Server Software:Apache/1.3.23 Document Path: /hello.php Document Length:12 bytes Complete requests: 10 Failed requests:0 Write errors: 0 Requests per second:1436.89 [#/sec] (mean) Time per request: 0.010 [ms] (mean) Transfer rate: 259.59 [Kbytes/sec] received Percentage of the requests served within a certain time (ms) 50% 10 95% 14 -- PHP Development Mailing List http://www.php.net/ To unsubscribe, visit: http://www.php.net/unsub.php -- PHP Development Mailing List http://www.php.net/ To unsubscribe, visit: http://www.php.net/unsub.php
RE: [PHP-DEV] Simple Apache 1.3 vs 2.0 benchmarks
Hey Rasmus, Well, I'd done a similar set of benchmarks between DSO and static with apache 1.3.23, which showed the DSO variant getting around 1100 req/s. So it wasn't an automatic assumption, I just noticed that the Apache 2 variant had some similar numbers. Unfortunately, the hardware I've been working with is giving me trouble (aka segfaulting during compiles) but I'll be happy to run some apache 1.3.23 DSO/static numbers when I get things fixed up. I was actually just trying to make the point you make about performance with Apache 2 being similar or slightly worse then Apache 1. There was some discussion along the lines of I need the blazing Apache 2 speed and wanted to show that was unlikely to happen, made more obvious because Apache 2 doesn't support static builds. There was a discussion on slashdot about some of this with some various perspectives: http://slashdot.org/comments.pl?sid=28250cid=3035903 Also benchmarked various MPM options, with little noticeable difference. Its's such a narrow workload however I'm not sure it really is meaningful. Also have benchmarked php footprint/speed for all the various options, as well as apache footprint/speed for all the various options. One lesson from that was that enabling or disabling the smaller modules makes little/no footprint/speed differences especially on the apache side. Anyways, hoping to get this box back in shape so I can get some work done. - AZ -Original Message- From: Rasmus Lerdorf [mailto:[EMAIL PROTECTED]] Sent: Monday, February 25, 2002 10:05 PM To: August Cc: [EMAIL PROTECTED] Subject: Re: [PHP-DEV] Simple Apache 1.3 vs 2.0 benchmarks Uh, but did you benchmark straight cgi's between the two servers? How about straight static pages? Why would you automatically attribute the difference in performance to DSO vs. static issues? It is pretty well known that Apache2 is currently slower than Apache1 for stuff like this because of the added overhead of the bucket brigade implementation and other issues. -Rasmus -- PHP Development Mailing List http://www.php.net/ To unsubscribe, visit: http://www.php.net/unsub.php
RE: [PHP-DEV] Simple Apache 1.3 vs 2.0 benchmarks
Which platform are you on? On Mon, 25 Feb 2002, August wrote: Hey Rasmus, Well, I'd done a similar set of benchmarks between DSO and static with apache 1.3.23, which showed the DSO variant getting around 1100 req/s. So it wasn't an automatic assumption, I just noticed that the Apache 2 variant had some similar numbers. Unfortunately, the hardware I've been working with is giving me trouble (aka segfaulting during compiles) but I'll be happy to run some apache 1.3.23 DSO/static numbers when I get things fixed up. I was actually just trying to make the point you make about performance with Apache 2 being similar or slightly worse then Apache 1. There was some discussion along the lines of I need the blazing Apache 2 speed and wanted to show that was unlikely to happen, made more obvious because Apache 2 doesn't support static builds. There was a discussion on slashdot about some of this with some various perspectives: http://slashdot.org/comments.pl?sid=28250cid=3035903 Also benchmarked various MPM options, with little noticeable difference. Its's such a narrow workload however I'm not sure it really is meaningful. Also have benchmarked php footprint/speed for all the various options, as well as apache footprint/speed for all the various options. One lesson from that was that enabling or disabling the smaller modules makes little/no footprint/speed differences especially on the apache side. Anyways, hoping to get this box back in shape so I can get some work done. - AZ -Original Message- From: Rasmus Lerdorf [mailto:[EMAIL PROTECTED]] Sent: Monday, February 25, 2002 10:05 PM To: August Cc: [EMAIL PROTECTED] Subject: Re: [PHP-DEV] Simple Apache 1.3 vs 2.0 benchmarks Uh, but did you benchmark straight cgi's between the two servers? How about straight static pages? Why would you automatically attribute the difference in performance to DSO vs. static issues? It is pretty well known that Apache2 is currently slower than Apache1 for stuff like this because of the added overhead of the bucket brigade implementation and other issues. -Rasmus -- PHP Development Mailing List http://www.php.net/ To unsubscribe, visit: http://www.php.net/unsub.php