giraph-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Eli Reisman (JIRA)" <>
Subject [jira] [Commented] (GIRAPH-297) Checkpointing on master is done one superstep later
Date Mon, 13 Aug 2012 18:02:40 GMT


Eli Reisman commented on GIRAPH-297:

Is there anything master.compute() needs to do during INPUT_SUPERSTEP? what if we skipped
master.compute in INPUT_SUPERSTEP and we could get it running before each calculation superstep,
before the vertex.compute() cycle begins? On superstep 0 it would still be the first thing
to happen, and could do any init/setup stuff it needs then. This way it would run in the same
superstep (count-wise) as the vertex.compute() cycle does. Getting all this stuff to line
up might justify another quick ZK barrier at the beginning of each calc superstep to ensure
the order of events?

You mentioned you ran into problems trying to get this to execute before vertex.compute()
cycle, what sort of issues did you run into?

> Checkpointing on master is done one superstep later
> ---------------------------------------------------
>                 Key: GIRAPH-297
>                 URL:
>             Project: Giraph
>          Issue Type: Bug
>            Reporter: Maja Kabiljo
>            Assignee: Maja Kabiljo
> On workers we store checkpoint X before compute() for superstep X are executed. On master
we do it after those compute() are executed and after master.compute() for superstep X+1.

This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:!default.jspa
For more information on JIRA, see:


View raw message