[Skinnyarms]

I recently stumbled upon this quote on the ColdFusion livedocs:

Quote

It takes relatively more processing time to instantiate a CFC than to process a custom tag. In turn, it takes substantially more time to process a custom tag than to execute a user-defined function (UDF).

This got me thinking and I wanted to run some benchmark tests. I'm was able to find a link to a tool that works on Windows. However I have my ColdFusion server running on Linux so I'd prefer to use a linux or OS independent tool. Anyone have any suggestions?



*PS: Theoretically I could use emulate the Windows tool, but I was worried about skewing my results. Cftimer is a cool tag, but it doesn't return anything. Also, I thought about ColdFusion Debugging and a spreadsheet, but I'd like to run more tests than I care to keep track of.

This post has been edited by Skinnyarms: 04 December 2006 - 06:50 AM

[psykoprogrammer]

Give a look to Fusion Reactor. We use it here where I work. It allows you a great bit of monitoring, including memory, query performance, etc. Even if it isn't exactly what you are looking for, it is something worth looking at.

Cheers.

[Skinnyarms]

That looks like an awesome tool, I'll be trying to sell the boss on it later today.

However, I'd really like to benchmark so that I can compare the processing times for small chunks of code, whereas this software looks like it's better for monitoring/debugging in a production enviornment.

I especially think the "Request Detail" looks really promising, I think it'd be great to gloss over and see which pages are taking the longest!

[psykoprogrammer]

Yes, you will find you like this tool. As for your specific issue, I know of know tools aside from test automation software, maybe. Thinking about it a bit makes me wonder if you couldn't make a handful of CFCs and custom tags and simply invoke each one wrapped in CFTIMER tags. The reason I suggest multiple of each is due to ColdFusion's caching.

I also have personally seen speed effects using CFCs vs. custom tags. Though I've only noticed the difference when using them in hundreds of thousands of calls in loop itereations in something like a data import.

[Skinnyarms]

I believe I'll be attempting this project pretty soon. It shouldn't be so difficult, I'm just really suprised nobodies done it already. The only thing that irks me is that CFTIMER doesn't actually return any values so I'll have to parse my return data. It's not a huge deal, but considering the volume of tests I'd like to run (as you said, refreshing once or twice would be horribly inaccurate) I'd prefer not to add any frivolous steps!

Thanks!

[psykoprogrammer]

Nice site btw. Bookmarked.

[Skinnyarms]

Thanks! I still have a lot of work to do. =/

[psykoprogrammer]

How's your test coming along?

[Skinnyarms]

It's going ok, all the results I've gotten back are pretty much what I expected. Inline code is faster than UDFs, UDFs are faster than components whi.

As the processing time for the code increases the difference between the three becomes less noticeable. I want to try and convert what I made into a component and add it to the "Dream Kudos", but it won't be today. =P

PS: The boss picked up SeeFusion!