The trinity system seem to be very much still in the works. Would their be some trade off to online query processing?
On Mar 24, 2011 1:00 PM, "Thomas Jungblut (JIRA)" <[email protected]> wrote: [ https://issues.apache.org/jira/browse/HAMA-359?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13010803#comment-13010803] Thomas Jungblut commented on HAMA-359: -------------------------------------- I read the papers about microsoft trinity Alois posted on Twitter. It was quite interesting. Maybe we should need something equal to the graph database, HBase would be an option. But since it is no more a requirement for Hama, I tend to write a SequenceFile Database. Storing byte offsets to a special lookup key or something.. Had a bit time and worked through Hama and BSP! Never the less I have another question: What is the advantage to use a "master groom" like in the PiEstimator example? I mean this could turn into a large bottleneck. BSP uses the communication phase to communicate between the nodes and not the nodes to a master node. Or did I missed something? Thanks! > Development of Shortest Path Finding Algorithm > ---------------------------------------------- >...
