Just to add some opinion here. I've followed each round of the techempower benchmarks religiously. The OpenResty tests are actually pretty typical on terms of the frameworks represented there. Many of the framework tests were implemented once and then not touched, which is why so many have partial test results. Some have been tuned a little, but very few have been tuned for each rest run. So what do the results mean ?
They are trivial tests, but you can see that some frameworks are 100x slower than others. When I look at the results I blur my eyes and imagine that I have three or four buckets (slow, medium, fast) and I think these benchmarks do a great job of indicating where framework X roughly sits.
I'm a big fan of Openresty and of Ruby on Rails. One is fast, the other slow. There's nothing wrong with that. Unfortunately most technologists don't have access to concrete performance data about the frameworks they use, and this sometimes leads to them making bad choices.
The tech empower benchmarks are open independent and easy to run on your own kit.
Sent from my iPhone
> On Mar 21, 2014, at 7:24 PM, "Yichun Zhang (agentzh)" <age...@gmail.com> wrote:
>
> Hello!
>
>> On Fri, Mar 21, 2014 at 6:26 AM, Tor Hveem wrote:
>> Framework benchmark round 9 will (according to the source
>> toolset/setup/linux/installer.py) be
>> http://openresty.org/download/ngx_openresty-1.5.8.1.tar.gz
>
> This version of openresty does not have the "trace stitching" feature
> in LuaJIT v2.1 yet. This feature should speed up lua-resty-mysql a lot
> according to my local benchmark :)
>
> And yeah, there're a lot more that we can do to get a better score in
> those trivial benchmark programs :)
>
> Regards,
> -agentzh.
