You are viewing a plain text version of this content. The canonical link for it is here.
Posted to solr-dev@lucene.apache.org by Tim Archambault <ti...@gmail.com> on 2007/11/02 16:50:21 UTC

Solr production live implementation

If this is the wrong email forum, I apologize in advance.

Looking to use Solr as the PRIMARY search engine for our newspaper website.
The index will initially hold between 200,000 - 500,000 documents.

I'm not sure what analytic data you'd need to help me with my question, but
I can tell you our website incurs roughly 4 million page views monthly and
about 30,000 absolute unique visitors per month. Our website traffic is
concentrated between 8am - 12noon so we have a lot of off-peak time on our
server.

I am currently trying SOLR out off of my dedicated Windows server (IIS 5)
with Jetty.  My server has 2GB Ram and tons of space. What is the likelyhood
that this environment is "good enough" for my production environment?

Any feedback is greatly appreciated.

Tim

Re: Solr production live implementation

Posted by Tim Archambault <ti...@gmail.com>.
Thank you kindly.

On 11/3/07, Bertrand Delacretaz <bd...@apache.org> wrote:
>
> On 11/2/07, Tim Archambault <ti...@gmail.com> wrote:
>
> > ...I am currently trying SOLR out off of my dedicated Windows server
> (IIS 5)
> > with Jetty.  My server has 2GB Ram and tons of space. What is the
> likelyhood
> > that this environment is "good enough" for my production environment?...
>
> FWIW, http://tsrvideo.ch/ (a video browser backed by a Solr index)
> runs on a mid-range 2006 Intel server (don't have the exact setup
> here, probably a dual Xeon 2GHz, 4GB RAM), with Linux + Apache httpd,
> and is coping very well with about 1.5 million Solr queries a month on
> average.
>
> BTW, I just had a look and the Solr instance there has been up since
> July 11th, not bad!
>
> I'd suggest estimating your peak load and running stress tests with
> something like ab (the httpd utility), JMeter or httpstone [1] to find
> out what happens to your index under peak load. You might use your
> current bandwidth stats to estimate how big your peaks are compared to
> the average load.
>
> -Bertrand
>
> [1] http://code.google.com/p/httpstone/source (some assembly required,
> that's a very simple java-based HTTP stress-test framework that I
> wrote)
>



-- 
True innovation is not just about changing a product, a service or even a
marketplace; its also about recognizing and relishing the need to change
yourself.

Re: Solr production live implementation

Posted by Bertrand Delacretaz <bd...@apache.org>.
On 11/2/07, Tim Archambault <ti...@gmail.com> wrote:

> ...I am currently trying SOLR out off of my dedicated Windows server (IIS 5)
> with Jetty.  My server has 2GB Ram and tons of space. What is the likelyhood
> that this environment is "good enough" for my production environment?...

FWIW, http://tsrvideo.ch/ (a video browser backed by a Solr index)
runs on a mid-range 2006 Intel server (don't have the exact setup
here, probably a dual Xeon 2GHz, 4GB RAM), with Linux + Apache httpd,
and is coping very well with about 1.5 million Solr queries a month on
average.

BTW, I just had a look and the Solr instance there has been up since
July 11th, not bad!

I'd suggest estimating your peak load and running stress tests with
something like ab (the httpd utility), JMeter or httpstone [1] to find
out what happens to your index under peak load. You might use your
current bandwidth stats to estimate how big your peaks are compared to
the average load.

-Bertrand

[1] http://code.google.com/p/httpstone/source (some assembly required,
that's a very simple java-based HTTP stress-test framework that I
wrote)