Thank you for your reply. How do you think about this scene, there is a big
graph while not all vertices are useful and is necessary for computing. We want
to prune some verteices and then to compute accurately. (Pruning and
Verifying). The pruning job finished when there is no messages and inactivated
vertices. Then we must output the result of pruning job and then read it in
verifying job. It takes many time for pointless reading and writing.
Now I use the code as follow to solve this problem. I set the stepsBeforeNext
and then run next job.
if(getSuperstepCount() < stepsBeforeNext){
????
} else {
????
}
But I don't think it's a good idea.
------------------ Original ------------------
From: "Edward J. Yoon";<[email protected]>;
Send time: Sunday, Feb 14, 2016 3:52 AM
To: "[email protected]"<[email protected]>;
Subject: Re: Could hama start next computation after all vertexes vote to halt?
Sorry for late reply. I'm not sure but it's impossible with current
APIs. Basically no messages and inactivated vertices are job
termination condition by default.
If there's a clear example that shows how combining two jobs into one
is useful, we can consider to add it as a new feature.
Thanks!
On Sun, Feb 7, 2016 at 5:54 PM, ?????? <[email protected]> wrote:
> Hi, I want to use hama to run a job, which is consist of two phase.
> The first phase try to find proper neighbors. The vertex execute
> voteToHalt() method when it doesn't receive any message. When all vertexes
> vote to halt, the first phase finished.
> Then the second phase starts based on the result of the first phase. The
> output of the first phase is the input of the first phase.
> Now I use two hama tasks to finish these two phase. While I think it's
> unnecessary to output the result and then read it. I want to finish these two
> phase in one hama task. How could I do this?
> Best Regards, Ping Liu.
--
Best Regards, Edward J. Yoon
