Benchmarking rails app

I’ve been poking around at some benchmarking to see how my app is
performing
(mongrel cluster/pound). Currently, I’m using ab and httperf and
collecting
requests/second. I’m not convinced that I’m using realistic loads.

What are others using for the options to these benchmark programs?

Thanks

View this message in context:
http://www.nabble.com/benchmarking-rails-app-tf2448012.html#a6823524
Sent from the RubyOnRails Users mailing list archive at Nabble.com.

On Oct 15, 2006, at 11:40 AM, s.ross wrote:

I’ve been poking around at some benchmarking to see how my app is
performing
(mongrel cluster/pound). Currently, I’m using ab and httperf and
collecting
requests/second. I’m not convinced that I’m using realistic loads.

What are others using for the options to these benchmark programs?

http://dev.robotcoop.com/Tools/production_log_analyzer/index.html

I attack the problem from the other end. Take actual usage and make
that faster. Getting a report of what actions need optimization is a
better way to focus your energies.

Creating realistic loads can be hard, so I don’t bother trying to
simulate it.


Eric H. - [email protected] - http://blog.segment7.net
This implementation is HODEL-HASH-9600 compliant

http://trackmap.robotcoop.com

Thanks, Eric. I’ve read your Tools section before and your analysis is
excellent. As an optimization technique, I agree with what you say.

What I’m after right now is a way to understand how much headroom my app
has. I.e., what would happen if we never touched a line of code and then
experienced an order of magnitude more traffic.

I think ab and httperf are somewhat blunt instruments but they can be
useful
in understanding some of the response characteristics of different
configurations, no?

Eric H. wrote:

What are others using for the options to these benchmark programs?

Eric H. - [email protected] - http://blog.segment7.net
This implementation is HODEL-HASH-9600 compliant

http://trackmap.robotcoop.com


View this message in context:
http://www.nabble.com/benchmarking-rails-app-tf2448012.html#a6825024
Sent from the RubyOnRails Users mailing list archive at Nabble.com.

Take a lot at railsbench (http://railsbench.rubyforge.org/)

I tend to use that to have a general idea of where I’m at (and to
evaluate the impact my changes have), together with action_profiler to
pin down where those cpu cycles are going

Fred

On Oct 15, 2006, at 2:14 PM, s.ross wrote:

http://dev.robotcoop.com/Tools/production_log_analyzer/index.html
has. I.e., what would happen if we never touched a line of code and
then
experienced an order of magnitude more traffic.

I think ab and httperf are somewhat blunt instruments but they can
be useful
in understanding some of the response characteristics of different
configurations, no?

Yes, and to get the right traffic mix you can use your logs to get an
idea of where that load is likely to go.

I bet you could combine pl_analyze’s report (for actions and
frequencies) with railsbench’s benchmarks.yml to get a realistic
answer to your question. You might need to throw in some extra log
file analysis to look at params, since some parameters may have
wildly different performance characteristics.


Eric H. - [email protected] - http://blog.segment7.net
This implementation is HODEL-HASH-9600 compliant

http://trackmap.robotcoop.com