riak-search Benchmar.k

Pablo Borges pablort at gmail.com
Fri Nov 5 08:51:52 EDT 2010


Due to disk space constraints I had to move to 5 node cluster anyway. I'm
uploading a compressed 6GB json file. I'm not sure the actual datasize, but
is should be about 10 ~ 12 times that (By the time I've reached 1.3mi docs,
the disk usage was around 23GB)

I restarted the test and now I'm with 2mi docs.

The upload speed is a constant 75 docs/s.

Query response doesn't look too good now, but then again I didn't do any
tuning. I wanted some numbers how it would perform out-of-the-box.

<response>
-
<lst name="responseHeader">
<int name="status">0</int>
<int name="QTime">37021</int>
-
<lst name="params">
<str name="indent">on</str>
<str name="start">0</str>
<str name="q">publicado:SIM</str>
<str name="q.op">or</str>
<str name="df">value</str>
<str name="wt">standard</str>
<str name="version">1.1</str>
<str name="rows">2017736</str>
</lst>
</lst>

On Fri, Nov 5, 2010 at 5:51 AM, Prometheus WillSurvive <
prometheus.willsurvive at gmail.com> wrote:

> Hi Pablo,
>
> Thanks for your reply. I am looking forward to your results.  Please also
> add the data size details. We have customers that dealing with web scale
> data size ( billion docs) so I am looking riaksearch whether it is fast
> enough and feature ready to this job.  We want all queries should be below
> 900 ms ..
>
> I will also make other tests and will be publis in this wiki..
>
> Regards
>
> Prometheus
>
>
>
> On Nov 5, 2010, at 3:06 AM, Pablo Borges wrote:
>
> I haven't used your data, but I'm trying to benchmark riak search against
> solr.
>
> For my test, I'm using each paragraph in the english version of wikipedia
> as a document as well as some sequential and random data in a couple of
> fields to reflect our current usage (which is a CMS). That's about 22
> million documents. My initial test is using just one machine to load it all,
> but in the next couple of days, I'll be using a 5 node Dell R610  with 24GB
> RAM and SSD drives to see how well it will perform against solr, which will
> be using the same machines.
>
> Right now, I'm looking into stats information to plot graphs (we use
> cacti). So far, I've been using a few numbers from /stats page (and trying
> to figure out what they mean. lol).
>
> The only numbers I've got so far is that I was able to index about 1
> million documents at a rate of 65 documents/s while the solr interface works
> pretty fast and the load on the machine seems to be steady (around 2.5).
>
> I'd really like some real world input to build the graphs, and when I'm
> done, I'm gonna post them on the wiki. :D
>
> Cheers,
>
> On Thu, Nov 4, 2010 at 2:13 PM, Prometheus WillSurvive <
> prometheus.willsurvive at gmail.com> wrote:
>
>> Hi Guys,
>>
>> Week ago I put data ready to index riaksearch via solr interface in the
>> rapidshare to make it available to community.
>>
>> I would love to get some benchmark resuts from you guys.  Is there anybody
>> test it ?
>>
>>
>> Prometheus..
>>
>>
>>
>> _______________________________________________
>> riak-users mailing list
>> riak-users at lists.basho.com
>> http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com
>>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.basho.com/pipermail/riak-users_lists.basho.com/attachments/20101105/a486bc78/attachment.html>


More information about the riak-users mailing list