LinuxQuestions.org

LinuxQuestions.org (/questions/)
-   Linux - Server (https://www.linuxquestions.org/questions/linux-server-73/)
-   -   When testing VPS server with Siege is 50 concurrent users good enough? (https://www.linuxquestions.org/questions/linux-server-73/when-testing-vps-server-with-siege-is-50-concurrent-users-good-enough-932953/)

deskt0plinux 03-06-2012 12:34 AM

When testing VPS server with Siege is 50 concurrent users good enough?
 
This question is not technical, rather about what thumbrules to use and what experiences are common.

Using the siege testing tool, my e-commerce site currently peaks at 45-50 concurrent users, after which performance starts going down, but remains available, although slow.

It is a LAMP application with PHP in CGI mode and the server is a VPS Xeon with 4GB RAM alloted to my site. I know CGI restricts performance, however, I have read a lot of warnings about using PHP as an Apache Module for stability - one thread error taking down the server, it seems.

The application itself is being tuned but I dont see a 100-500% performance boost coming from that tweaking without a major overhaul.

I've tweaked MySQL parameters to get the best possible results, run Mysqltuner.pl and taken its suggestions, rewritten some queries and so on.

Is there a comparative listing of what backend can sustain what kind of concurrent users - VPS/Xeon/4G -- 50 users, Dedicated/4G -- 1000 users, and so on.

What basic changes could be done to handle more concurrent users?

Any tools that give more insight into this scaling-up activity?

I do Google around a lot but I also want some anecdotes and experience if possible.

If you could just point me to a site or group where such discussions for beginners are common that too would be great.

Thanks in advance.

elfenlied 03-06-2012 05:38 AM

When you are testing load on your site are you accurately simulating a user? That's probably the hardest part because unless you are what ever metrics you're using as a basing won't matter if you later find a bottleneck that your load test didn't pickup due to it not simulating traffic correctly.

How are you running apache? prefork or worker? See http://httpd.apache.org/docs/2.0/mpm.html

In a lot of cases changing apache to use the worker MPM will make the biggest difference, the next thing I'd look at are things such as your database connection pools. Are they enough? Do you need more? Are they being closed in a timely fashion etc.

Hope this helps.


All times are GMT -5. The time now is 04:27 PM.