You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@couchdb.apache.org by Chris Anderson <jc...@apache.org> on 2009/10/16 23:51:16 UTC

bench/ directory

Devs,

I wrote some benchmarks today and added them to trunk. They are still
rudimentary, but I wanted to open the door for other people to follow.
I've written a blog post with some numbers here:

http://jchrisa.net/drl/_design/sofa/_show/post/Benchmarking-CouchDB

To use the JavaScript benchmarks:

cd bench/
./runner.sh

The scripts are written in benches.js so it should be easy to add new
ones. They use the same client as the test suite so you already know
how to hack it. There's no reason these shouldn't be able to run in
the browser with a little glue code. Keep that in mind as you are
writing new ones, but it should't be a big deal to get them to run in
either place eventually.

There are also plain curl + bash benchmarks you can run from:

./benchbulk.sh

I'd like to work in a way to collect and share results from these as
well. One step at a time.

Chris

-- 
Chris Anderson
http://jchrisa.net
http://couch.io

Re: bench/ directory

Posted by Joe Williams <jo...@joetify.com>.

On Fri, 16 Oct 2009 15:35:28 -0700
Chris Anderson <jc...@apache.org> wrote:

> On Fri, Oct 16, 2009 at 3:21 PM, Joe Williams <jo...@joetify.com> wrote:
> >
> > I have also been working on benchmarking Couch and wrote Baracus
> > (http://github.com/joewilliams/baracus) the last few days. It's a
> > wrapper for httperf that generates session work logs that httperf
> > reads through and generates (reads and writes) requests from. Doc
> > size, session count, requests per session, request rate and etc are
> > all configurable. It even dumps the results into Couch so you can
> > write views to get more from the data.
> >
> > I have a few new features in the works such as work logs as
> > attachments so you can replay them at a later date (thanks davisp)
> > and etc. Any feedback is appreciated.
> >
> 
> Cool! We should come up with a simple JSON document format for saving
> benchmark results. That way they can be aggregated, visualized and
> mixed into the Relaxville data corpus, etc.


Thats probably not a bad idea, make it easy to compare results and etc.
FWIW the format I have been using looks like:
http://friendpaste.com/9yzxPOI5TYsl79gOEtvfG


> 
> Chris
> 
> > -Joe
> >
> >
> >
> > On Fri, 16 Oct 2009 14:51:16 -0700
> > Chris Anderson <jc...@apache.org> wrote:
> >
> >> Devs,
> >>
> >> I wrote some benchmarks today and added them to trunk. They are
> >> still rudimentary, but I wanted to open the door for other people
> >> to follow. I've written a blog post with some numbers here:
> >>
> >> http://jchrisa.net/drl/_design/sofa/_show/post/Benchmarking-CouchDB
> >>
> >> To use the JavaScript benchmarks:
> >>
> >> cd bench/
> >> ./runner.sh
> >>
> >> The scripts are written in benches.js so it should be easy to add
> >> new ones. They use the same client as the test suite so you
> >> already know how to hack it. There's no reason these shouldn't be
> >> able to run in the browser with a little glue code. Keep that in
> >> mind as you are writing new ones, but it should't be a big deal to
> >> get them to run in either place eventually.
> >>
> >> There are also plain curl + bash benchmarks you can run from:
> >>
> >> ./benchbulk.sh
> >>
> >> I'd like to work in a way to collect and share results from these
> >> as well. One step at a time.
> >>
> >> Chris
> >>
> >
> >
> > --
> > Name: Joseph A. Williams
> > Email: joe@joetify.com
> > Blog: http://www.joeandmotorboat.com/
> >
> 
> 
> 


-- 
Name: Joseph A. Williams
Email: joe@joetify.com
Blog: http://www.joeandmotorboat.com/

Re: bench/ directory

Posted by Chris Anderson <jc...@apache.org>.
On Fri, Oct 16, 2009 at 3:21 PM, Joe Williams <jo...@joetify.com> wrote:
>
> I have also been working on benchmarking Couch and wrote Baracus
> (http://github.com/joewilliams/baracus) the last few days. It's a
> wrapper for httperf that generates session work logs that httperf
> reads through and generates (reads and writes) requests from. Doc size,
> session count, requests per session, request rate and etc are all
> configurable. It even dumps the results into Couch so you can write
> views to get more from the data.
>
> I have a few new features in the works such as work logs as attachments
> so you can replay them at a later date (thanks davisp) and etc. Any
> feedback is appreciated.
>

Cool! We should come up with a simple JSON document format for saving
benchmark results. That way they can be aggregated, visualized and
mixed into the Relaxville data corpus, etc.

Chris

> -Joe
>
>
>
> On Fri, 16 Oct 2009 14:51:16 -0700
> Chris Anderson <jc...@apache.org> wrote:
>
>> Devs,
>>
>> I wrote some benchmarks today and added them to trunk. They are still
>> rudimentary, but I wanted to open the door for other people to follow.
>> I've written a blog post with some numbers here:
>>
>> http://jchrisa.net/drl/_design/sofa/_show/post/Benchmarking-CouchDB
>>
>> To use the JavaScript benchmarks:
>>
>> cd bench/
>> ./runner.sh
>>
>> The scripts are written in benches.js so it should be easy to add new
>> ones. They use the same client as the test suite so you already know
>> how to hack it. There's no reason these shouldn't be able to run in
>> the browser with a little glue code. Keep that in mind as you are
>> writing new ones, but it should't be a big deal to get them to run in
>> either place eventually.
>>
>> There are also plain curl + bash benchmarks you can run from:
>>
>> ./benchbulk.sh
>>
>> I'd like to work in a way to collect and share results from these as
>> well. One step at a time.
>>
>> Chris
>>
>
>
> --
> Name: Joseph A. Williams
> Email: joe@joetify.com
> Blog: http://www.joeandmotorboat.com/
>



-- 
Chris Anderson
http://jchrisa.net
http://couch.io

Re: bench/ directory

Posted by Joe Williams <jo...@joetify.com>.
I have also been working on benchmarking Couch and wrote Baracus
(http://github.com/joewilliams/baracus) the last few days. It's a
wrapper for httperf that generates session work logs that httperf
reads through and generates (reads and writes) requests from. Doc size,
session count, requests per session, request rate and etc are all
configurable. It even dumps the results into Couch so you can write
views to get more from the data.

I have a few new features in the works such as work logs as attachments
so you can replay them at a later date (thanks davisp) and etc. Any
feedback is appreciated.

-Joe



On Fri, 16 Oct 2009 14:51:16 -0700
Chris Anderson <jc...@apache.org> wrote:

> Devs,
> 
> I wrote some benchmarks today and added them to trunk. They are still
> rudimentary, but I wanted to open the door for other people to follow.
> I've written a blog post with some numbers here:
> 
> http://jchrisa.net/drl/_design/sofa/_show/post/Benchmarking-CouchDB
> 
> To use the JavaScript benchmarks:
> 
> cd bench/
> ./runner.sh
> 
> The scripts are written in benches.js so it should be easy to add new
> ones. They use the same client as the test suite so you already know
> how to hack it. There's no reason these shouldn't be able to run in
> the browser with a little glue code. Keep that in mind as you are
> writing new ones, but it should't be a big deal to get them to run in
> either place eventually.
> 
> There are also plain curl + bash benchmarks you can run from:
> 
> ./benchbulk.sh
> 
> I'd like to work in a way to collect and share results from these as
> well. One step at a time.
> 
> Chris
> 


-- 
Name: Joseph A. Williams
Email: joe@joetify.com
Blog: http://www.joeandmotorboat.com/