Benchmarking s9y?

[TKS]

I was wondering if anyone ever benchmarked serendipity? The reason being...I've been doing some myself using apache bench and plan on giving the php a benchmark as well.


The reason I ask is this...I've seen steady growth in the amount of traffic my blog has had...for instance, the first month in existence I had around 10k hits all month long...mainly due to a syndicated article I wrote. Now that my blog has been up for 7 months or so...I've been syndicated and linked to more and had over 200k hits this month so far.

What I fear is that I'll get something akin to a slashdotting and the blog will crumple. I'm moving to a host to combat the hardware side...now what I need to know is the software side. Has anyone benchmarked s9y before? If so, would you share the results?


I'll share one of mine...what concerns me is that there are close to 934 error requests...I installed zend optimization and it dropped to 480...but that is odd considering if I do the same benchmark on a phpBB forum...there is 0 errors. Anyone know why that would be this way? That's a bit extreme...of course, I would have chalked it up to being a settings thing had phpBB not completely took the bench and done great with it.

As promised, the bench below:

Code: Select all

%>  ab -n 1000 -c 200 http://linuxblog.sytes.net/index.php?/archives/74-Experiment-Update-Installed-Ubuntu-5.04-Hoary.html


Server Software:        Apache
Server Hostname:        linuxblog.sytes.net
Server Port:            80

Document Path:          /index.php?/archives/74-Experiment-Update-Installed-Ubuntu-5.04-Hoary.html
Document Length:        30581 bytes

Concurrency Level:      200
Time taken for tests:   164.314321 seconds
Complete requests:      1000
Failed requests:        934
   (Connect: 0, Length: 934, Exceptions: 0)
Write errors:           0
Total transferred:      13545284 bytes
HTML transferred:       13088749 bytes
Requests per second:    6.09 [#/sec] (mean)
Time per request:       32862.863 [ms] (mean)
Time per request:       164.314 [ms] (mean, across all concurrent requests)
Transfer rate:          80.50 [Kbytes/sec] received

Connection Times (ms)
              min  mean[+/-sd] median   max
Connect:        0 1938 8051.4      0   92997
Processing:  2043 23744 20871.1  15577  151285
Waiting:     1933 23558 20655.1  15519  151274
Total:       2061 25683 24209.0  15707  160284

Percentage of the requests served within a certain time (ms)
  50%  15707
  66%  31222
  75%  35796
  80%  38812
  90%  53616
  95%  79462
  98%  107644
  99%  110537
 100%  160284 (longest request)

With a different concurrancy level (100 instead of 200)

Code: Select all

%>  ab -n 1000 -c 100 http://linuxblog.sytes.net/index.php?/archives/74-Experiment-Update-Installed-Ubuntu-5.04-Hoary.html

Server Software:        Apache
Server Hostname:        linuxblog.sytes.net
Server Port:            80

Document Path:          /index.php?/archives/74-Experiment-Update-Installed-Ubuntu-5.04-Hoary.html
Document Length:        33014 bytes

Concurrency Level:      100
Time taken for tests:   257.979613 seconds
Complete requests:      1000
Failed requests:        48
   (Connect: 0, Length: 48, Exceptions: 0)
Write errors:           0
Total transferred:      32082140 bytes
HTML transferred:       31570600 bytes
Requests per second:    3.88 [#/sec] (mean)
Time per request:       25797.961 [ms] (mean)
Time per request:       257.980 [ms] (mean, across all concurrent requests)
Transfer rate:          121.44 [Kbytes/sec] received

Connection Times (ms)
              min  mean[+/-sd] median   max
Connect:        0    2   9.0      0      80
Processing:   244 24731 10039.6  24135  126324
Waiting:      243 24233 9827.4  23942  126323
Total:        244 24733 10038.8  24135  126333

Percentage of the requests served within a certain time (ms)
  50%  24135
  66%  25854
  75%  27672
  80%  28888
  90%  34705
  95%  39083
  98%  44959
  99%  48018
 100%  126333 (longest request)

Notice that the errors drop considerably. This leads me to believe that more connections would cause more errors. If you bench this in phpBB...there is no change. Both concurrancies have 0 errors.

Odd. Is there a way to improve this?

[nohn]

To gain up to 400%-500% better performance use lighttpd, fastcgi, eaccelerator.

Apache is slow, mod_php sucks.

[oxygenws]

nohn, most web servers are Apache and using mod_php.
and phpBB works on them, without any error (as TKS said )

p.s.: phpBB is a very bad optimized system!!!

[garvinhicking]

I've just written up last week a simple "full cache" plugin which staticizes all pages in situations like yours.

You'll use some dynamic flexibility (as with any caching system) of course, but it would hand out pages pretty fast.

Other than that, I can only agree to nohn - lighthttpd and putting RAM on the machine + DB caching, opcode optimizers is pretty neat, as well as using either the plugin mentioned above or the entryproperties caching facility.

Regards,
Garvin

[cyberscribe@php.net]

I have the same problem with a s9y blog under heavy load ... where can I get your code to staticize my pages?

[garvinhicking]

The plugin can be found here: http://cvs.sourceforge.net/viewcvs.py/p ... achesimple

Also make sure to make use of PHP opcode caches, and if available try out LightHTTPD. ALso make sure your webserver has enough RAM and MySQL query caching activated.

Regards
Garvin