flink-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From nltran <...@git.apache.org>
Subject [GitHub] flink pull request: Stale Synchronous Parallel Iterations
Date Mon, 03 Aug 2015 14:15:34 GMT
Github user nltran commented on the pull request:

    Thank you for your feedback!
    I think this is a fine way to go. I will then provide the parameter server related code
in a different module and include an example of how to run a SSP job with calls to the parameter
    >You seem to re-initialize the caches in every superstep. In general, initialization
in open() in iterations needs to be done carefully, because open() is called at the beginning
of each superstep. You can always check in which superstep you are using getIterationRuntimeContext().getCurrentSuperstep()
    We initialize the caches at the first superstep by checking if the reference is null,
as the assumption is that since the same instantiated task is reused through the iterations,
the cache is initialized only in the first superstep and the reference is not null in the
subsequent supersteps. Would you recommend to do that check rather on the number of superstep?

If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastructure@apache.org or file a JIRA ticket
with INFRA.

View raw message