Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

I come from a web developer background and have been humming along building my PHP app, using the CakePHP framework. The problem arose when I began the ab (Apache Bench) testing on the Amazon EC2 instance in which the app resides. I'm getting pretty horrendous average page load times, even though I'm running a c1.medium instance (2 cores, 2GB RAM), and I think I'm doing everything right.

I would run:

ab -n 200 -c 20 http://localhost/heavy-but-view-cached-page.php

Here are the results:

Concurrency Level:      20 
Time taken for tests:   48.197 seconds 
Complete requests:      200 
Failed requests:        0 
Write errors:    0 
Total transferred:      392111200 bytes 
HTML transferred:       392047600 bytes 
Requests per second:    4.15 [#/sec] (mean) 
Time per request:       4819.723 [ms] (mean) 
Time per request:       240.986 [ms] (mean, across all concurrent requests) 
Transfer rate:          7944.88 [Kbytes/sec] received

While the ab test is running, I run VMStat, which shows that Swap stays at 0, CPU is constantly at 80-100% (although I'm not sure I can trust this on a VM), RAM utilization ramps up to about 1.6G (leaving 400M free). Load goes up to about 8 and site slows to a crawl.

Here's what I think I'm doing right on the code side:

  • In Chrome browser uncached pages typically load in 800-1000ms, and cached pages load in 300-500ms. Not stunning, but not terrible either.
  • Thanks to view caching, there might be at most one DB query per page-load to write session data. So we can rule out a DB bottleneck.
  • I have APC on.
  • I am using Memcached to serve the view cache and other site caches.
  • xhprof code profiler shows that cached pages take up 10MB-40MB in memory and 100ms - 1000ms in wall time.

Pages that would be the worst offenders would look something like this in xhprof:

Total Incl. Wall Time (microsec):   330,143 microsecs
Total Incl. CPU (microsecs):    320,019 microsecs
Total Incl. MemUse (bytes): 36,786,192 bytes
Total Incl. PeakMemUse (bytes): 46,667,008 bytes
Number of Function Calls:   5,195

My Apache config:

KeepAlive On
MaxKeepAliveRequests 100
KeepAliveTimeout 3

<IfModule mpm_prefork_module>
    StartServers           5
    MinSpareServers        5
    MaxSpareServers       10
    MaxClients            120
    MaxRequestsPerChild  1000
</IfModule>

Is there something wrong with the server? Some gotcha with the EC2? Or is it my code? Some obvious setting I should look into? Too many DNS lookups? What am I missing? I really want to get to 1,000 concurrency capacity, but at this rate, it ain't gonna happen.

share|improve this question
    
You'll really have to look more at what's going on. Are the CPUs maxed? Is there evidence of memory shortage? (Like the disk cache being squeezed) Is one core maxed and the other idle? How much disk I/O is there? –  David Schwartz Sep 8 '12 at 5:11
    
OK, I added vmstat results under the ab test results. –  AMF Sep 9 '12 at 14:03
    
Would be useful with a reference number. How many pages with phpinfo(); can you deliver locally? –  3molo Sep 9 '12 at 18:22
add comment

1 Answer

You're transferring a lot of data. 8mbytes/s, you're maxing out your network connection.

share|improve this answer
    
That's a load test, not the actual data usage... –  Chris S Mar 15 '13 at 6:13
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.