Hello, Currently, we're working on BSP (Bulk Synchronous Parallel) computing engine for large-scale matrix and graph problems, since we decided to implement more generalized BSP computing solution instead of the Angrapa, a Pregel-like graph-specific computing solution.
> framework to process social network data, what is the recommended arch > (e.g. hama + neo4j/ hdfs)? The hama-0.2 version doesn't provides input/output API. So, you may want to connect to a remote database. I would recommend to use HBase, at this time. On Mon, Nov 22, 2010 at 3:07 PM, Lin Chia-Hung <[email protected]> wrote: > I came across to read the following discussion > > > http://markmail.org/message/ev24gg5khsgskxmx#query:+page:1+mid:jzhyxus53gpwe7sv+state:results > > regarding to the graphic library. Does this indicate that `Angrapa' > may be similar to graph database? > > From what I have discovered so far, hama resembles to map/reduce and > data may store on e.g. hdfs or hbase > (http://www.mail-archive.com/[email protected]/msg00211.html). > So we can use graph database like neo4j to store graph data if > necessary. > > However, in Angrapa (http://wiki.apache.org/hama/GraphPackage), it > refers that the lib consists of master and workers to compute the > data; this somehow seems to me the usage of hama in essence. So what > is the role of Angrapa? Or if a user/ developer decide to use hama as > framework to process social network data, what is the recommended arch > (e.g. hama + neo4j/ hdfs)? > > Thanks. > Chiahung > -- Best Regards, Edward J. Yoon [email protected] http://blog.udanax.org
