Re: Termination Conditions

Hi Nick, I guess you can use MasterCompute together with BooleanAndAggregator to solve your problem. Whenever a vertex decides to halt, it will notify the MasterCompute with "true" value for the aggregator, otherwise it will send "false". If all the vertices agree to halt at a superstep, then the

Re: Adding rb to approved email addresses?

I'm still getting the moderation requests from rb. Am I the only one? On Tue, Jul 17, 2012 at 12:52 AM, Avery Ching wrote: > I think it was the mailing list issue (see https://issues.apache.org/** > jira/browse/INFRA-5046 ). > Hopefully new re

Re: Termination Conditions

Thanks for the reply. Is there an easy modification that I can make to remove condition 2? Can you point me to the code that addresses this? The problem I am facing is the following: At every iteration a non-halted vertex needs messages from all of its neighbors. When deciding to send messa

Re: Termination Conditions

Hi Nick, You are pretty much correct, except that not all vertices need to vote to halt at the same time: some vertices might have voted to halt at a previous superstep and never received any messages after then, in which case they are never reactivated. In other words, I think you can rephras

Termination Conditions

Excuse me if this is stated somewhere obvious, but I haven't been unable to find it. What are the exact termination criteria for the global algorithm? Reading the documentation on voteToHalt, looking at the Shortest Path Example code, and looking at the results of my own application, these two