+1 creating SIS processes for Hadoop would be what I would expect, Martin too. I had a student in my Search Engines course that I teach at USC do some Hadoop and Geospatial stuff a few years back (in 2010). The goal was to build a Point/Radius evaluation algorithm for tiling over Hadoop. It was mildly successful, and we may have something similar here as an option for SIS.
Cheers, Chris ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Chris Mattmann, Ph.D. Senior Computer Scientist NASA Jet Propulsion Laboratory Pasadena, CA 91109 USA Office: 171-266B, Mailstop: 171-246 Email: [email protected] WWW: http://sunset.usc.edu/~mattmann/ ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Adjunct Assistant Professor, Computer Science Department University of Southern California, Los Angeles, CA 90089 USA ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ -----Original Message----- From: Martin Desruisseaux <[email protected]> Organization: Geomatys Reply-To: "[email protected]" <[email protected]> Date: Friday, May 3, 2013 1:45 PM To: "[email protected]" <[email protected]> Subject: Re: Implement Hadoop base back-end for SIS >Hello 张亚 > >Le 03/05/13 17:42, 张亚 a écrit : >> I want to implement Hadoop base back-end for SIS during the GSoC 2013 >> period. And have submitted a proposal. >> Any suggestion will be welcome. > >I would like to help, but it is not clear to me which part of SIS could >be the subject of a Hadoop work yet... Hadoop is a framework that allows >for the distributed processing of large data sets across clusters of >computers. Problem is that there is not yet (to my knowledge) any >process in SIS which are numerically intensive enough for experimenting >Hadoop. Those processes will exist later, but are not yet there. > >Actually we experimented Hadoop on our side about 2 or 3 years ago. We >had a student who worked on that subject. His work was to rotate a tiled >image, where each tiles were processed by a different node on a cluster. >Of course doing an image rotation is a relatively simple process, but >the goal was to experiment distributed processing rather than doing >"real" work. The experiment was not very conclusive in part because we >tried to perform the rotation with Java Advanced Imaging (JAI) and it >was pretty hard to use JAI with Hadoop (JAI was not designed for that), >and in part because the time needed for transferring large tiles between >the nodes (even with ultra fast transfers) overcome the gain of using >many nodes for such a "simple" task as image rotation. So we gained just >enough experience for concluding that this is a challenging topic. > >But on the "SIS + Hadoop" topic, it seems to me that before to try >Hadoop with needs to have some SIS processes? Would creating those >processes be part of the Google Summer of Code? If so, it seems to me >that this work alone could keep someone busy for the whole summer... > >Regards, > >Martin >
