Running RecommenderJob remotely

[Paul Scott <ps...@gmail.com>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi,

I may be approaching this all wrong, but I have a shiny hadoop cluster
(with HBase, HDFS etc running) and would like to leverage that to do
Taste based recomendations (user based).

I have been looking through the docs and Javadoc and see that there is
a RecommenderJob in mahout.cf.taste.hadoop that can be used to kick
off Hadoop M/R jobs etc.

I would like to use this, as well as my existing data from a remote
app. Is that possible. I.e., I have an application that parses user
ratings etc and stores them to MongoDB and/or HBase. That application
should then be able to ask the Hadoop cluster to do the data analysis
and return a set o recommendations, leveraging the power of the cluster.

The code, using a variety of the recommenders on a 4 node MongoDB
sharded replicaSet takes up to ~850s and I would like to get that down
by distributing the work a little better (mongoDB only works with 2
threads)

If anyone has any insight into this, I would greatly appreciate it!

- -- Paul
- -- 
http://paulscott.co.za/blog/
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.12 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/

iQEcBAEBAgAGBQJSOZfaAAoJEP7GEwhwShZxyIQIAKl6DNtkYxXGUpONM29SfKTK
mFuyzFamXd3SLXRPdVSJeCcvoAsC8zuEKUEk5WTj4P9J0W6X17jqrF/xhYnfUSAZ
RDM072ucw4gmJUdeflYbUk2Nhs0nb16B6fwQTBGiV/VuvOjP+0NWpN/xA4e277hc
1gmeRBW6DPSR06jfgsKwOAbgGRrQ2cY1X1nsunQTdlaU/K5sYgSSEWDypf8CzwWl
NoLSEKoNKoiVsv6YGTPBBwZyANUSUg+Z9XbpoNWWtawpvklW6dJA/RRl2Df4h93U
gFuVf3SIfdPn7sPI01fz12pF9QDeijeYhtkyFwvrzdd3Vvea3MGAYDnUxtIv1Ko=
=eGiS
-----END PGP SIGNATURE-----