what performance do you get?

Hello,

we are currently struggling to increase our MODx websites performance. We already use opcode caching and gzip on apache/php server side level, which helps a lot. Also we use MODx caching where possible. Still, on  one of our dedicated servers where a MODx site homepage from the cache reaches the following stats:

Query Time = 0.0005 s
Query Count = 1
Parse Time = 0.0636 s
Total Time = 0.0641 s
Source = cache

we only reach about 27 request/second when we do a apache benchmark stress-test with 20 concurrent users for 30 seconds time. I understand that server, mysql and apache configuration/tuning are also important here, but i was wondering what other people reach for benchmarks..

to compare: on the same server a regular static HTML file reaches about 2500 requests/second, and a PHP script with a single database connect and loop through the resultset reaches about 500 request/second...

oh, we are testing on the server itself, so we have a theoretically unlimited bandwith

Olaf

ab -n100 -c20 [mymodxsite.com]

Gives me around 15 requests/sec on a dedicated server

As a comparison a php script running simple a db query on the same server yields 100 requests/sec

Hmm

Edit: the server is run by fasthosts.

Quote from: therebechips at Jun 05, 2009, 11:08 AM

ab -n100 -c20 [mymodxsite.com]

Gives me around 15 requests/sec on a dedicated server

As a comparison a php script running simple a db query on the same server yields 100 requests/sec

Hmm

Edit: the server is run by fasthosts.

did you enable MODx caching? And did you run the AB test directly on the webserver itself? otherwise you might be testing your internet connection speed
It seems with opcode-caching and MODx caching enabled the max a MODx site can reach on a dedicated server with apache and mysql and 4GB RAM is about 30 requests/second.

Olaf

did you enable MODx caching? And did you run the AB test directly on the webserver itself? otherwise you might be testing your internet connection speed Wink

I always thought you were meant to use AB on a local machine and target a remote server, thereby benchmarking real-world performance - am I wrong

I did have caching enabled on the tested page.

Well, I suppose if you’re testing for "real-world" performance, but if you’re comparing performance I think you’d get cleaner results locally. I mean, if I’m happily testing away and my kid comes home from school and logs into his facebook, starts to twitter at his friends, and logs in to his Warcraft account and his streaming radio account, my application’s performance will take a big hit. Or if my ISP takes a hit because 2,000 kids just got home from school...

Quote from: therebechips at Jun 09, 2009, 03:38 AM

did you enable MODx caching? And did you run the AB test directly on the webserver itself? otherwise you might be testing your internet connection speed Wink

I always thought you were meant to use AB on a local machine and target a remote server, thereby benchmarking real-world performance - am I wrong

I did have caching enabled on the tested page.

depends on what you want to measure. If you want to measure the maximum possible number of requests your server can handle with a theoretically unlimited bandwith, you should run the test on the server itself, otherwise the network-connection will problably be your bottleneck.

Olaf