Hi Nico,

I've duplicated your results, so you're right, Yaws 1.96 is slower on your benchmarks than Yaws 1.95. But the good news is that by the same tests, the current HEAD of Yaws master is slightly faster than Yaws 1.95. Thanks for identifying this issue.


On Sat, Jul 20, 2013 at 9:13 AM, Steve Vinoski <vinoski@ieee.org> wrote:
Hi Nico, sorry for the delayed reply, I was on a short holiday. I'll start looking into this.


On Wed, Jul 17, 2013 at 8:53 AM, Nico Kruber <nico.kruber@gmail.com> wrote:
we've been using embedded yaws for a while now in our Scalaris(1) project for
offering a JSON-API via JSON-RPC requests and provide example implementations
for python and ruby.

Ever since the upgrade from 1.95 to 1.96 we experienced a performance drop of
around 7% in our benchmark tests on one of our test servers (2x 4-core Opteron
2376) in Erlang R14B04 as well as R16B.
Locally on R16B01 (Core i7-2620M CPU @ 2.70GHz) I get even higher differences
between using yaws 1.95 and 1.96: 15%-35%.

Has anybody else observed this in his application?
What change could have introduced this drop?

Nico Kruber

(1) https://code.google.com/p/scalaris/

PS: You can test yourself using Scalaris from latest trunk (includes yaws
1.96) and benchmark with
SCALARIS_BENCHMARKS="500 10 6 9 18" make python-bench
reverting r4753 (and fixing the small conflict in
contrib/yaws/src/mime_types.erl) gets back yaws 1.95

svn co http://scalaris.googlecode.com/svn/trunk/ scalaris

See everything from the browser to the database with AppDynamics
Get end-to-end visibility with application monitoring from AppDynamics
Isolate bottlenecks and diagnose root cause in seconds.
Start your free trial of AppDynamics Pro today!
Erlyaws-list mailing list