Grid Computing

Final Project of DCS course

Basic Features

• Connection between GS&RM, RM&Node, JobSender&RM

GS communication

• GS assign job to RM who are less busy (RM has to accept, unless its offline)
• Each GS has checker for other GSs, regularly check if they are still online
• If one GS is crashed, anounce all the GS, transfer his RM to others
• GS has checker for each RM under it, regularly check if they are still online
• Once RM is crashed, GS transfer its joblist to another RM under it (hasn't thought about what all the RM under GS are crashed)

RM communication

• RM is able to receive jobs from JobSender
• RM assign jobs to free nodes (check free nodes first)
• When RM gives job to a Node, it starts a Thread
• RM hand in jobs to GS when all the nodes under it are busy

Nodes communication

• When the RM that the nodes belong to is not online, ignore the creation procedure
• When a job is done and the RM is not online (thus cannot receive complete message), ignore

JS communication

• When JS try to give jobs to a RM and it is not online, automatically search for other RMs to give jobs

TODO

• Run on DAS or other server

Useful Things

• Concept of Technichal Debt, we spent long time to fix bugs, writing catch blocks
• Possible flaw in the system