[cogen] Comparison of python web servers, cogen does poorly

[TK]

I see Cogen is pretty much a loser of the following test:
http://nichol.as/benchmark-of-python-web-servers

I really like Cogen from a developer's perspective. To the Cogen
devs; what is being done to make Cogen faster? And also he mentioned
something about memory leaks that sounds scary.

I would like to use the Cogen in a production system as soon as
possible, but this article worries me a bit.

Again, I really like Cogen, it's great work, keep it up. Let me
know what you think about my questions above.

Thanks

--
You received this message because you are subscribed to the Google Groups "cogen" group.
To post to this group, send email to co...@googlegroups.com.
To unsubscribe from this group, send email to cogen+un...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/cogen?hl=en.

[Ionel Maries Cristian]

When cogen had been designed it had three goals: leverage generators, be portable, be at least as fast as twisted. So far I think it achieved these goals.

About Nicholas's benchmark ... well, the raw throughput numbers do not always translate into a decision on choosing a webserver since there are a lot of other factors that have more importance and the application is usually your real bottleneck. The raw http performance isn't great since cogen creates a lot of objects (an "Operation" for every I/O call at least).

It didn't have any leak when I benchmarked it and the memory usage in Nicholas's benchmark doesn't look like a leak: it's high but it has a reasonable cap. It remains high due the way the python's memory allocator works (it doesn't release free memory to the system due to a number of reasons like memory fragmentation and quirks of the libraries python was compiled with). If you want to read more on that here's a starting point: effbot.org/pyfaq/why-doesnt-python-release-the-memory-when-i-delete-a-large-object.htm

About what is being done right now, well ... it's a shame really: cogen is in maintenance state right now as I don't have enough time/motivation to develop new features.

-- ionel