crunch-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Josh Wills (JIRA)" <>
Subject [jira] [Updated] (CRUNCH-294) Cost-based job planning
Date Wed, 20 Nov 2013 06:27:23 GMT


Josh Wills updated CRUNCH-294:

    Attachment: CRUNCH-294b.patch

[~gabriel.reid] took another pass at this by adding a breakpoint() method to PCollection that
allows the client to mark where on the path between two GBK operations that a split should
occur. I was hacking on the cpuFootprint() approach, but it felt a bit too abstracted from
the real thing that the client wanted to do in this case.

I also modified the split logic to first ignore any node paths that already contain materialized
SourceTargets, and then choose between either selecting a) the smallest single collection
that covered all of the node paths between two GBKs and b) the individual smallest PCollections
along the individual node paths, whichever was smaller.

> Cost-based job planning
> -----------------------
>                 Key: CRUNCH-294
>                 URL:
>             Project: Crunch
>          Issue Type: Improvement
>          Components: Core
>            Reporter: Josh Wills
>            Assignee: Josh Wills
>         Attachments: CRUNCH-294.patch, CRUNCH-294b.patch, jobplan-default-new.png, jobplan-default-old.png,
jobplan-large_s2_s3.png, jobplan-lopsided.png
> A bug report on the user list drove me to revisit some of the core planning logic, particularly
around how we decide where to split up DoFns between two dependent MapReduce jobs.
> I found an old TODO about using the scale factor from a DoFn to decide where to split
up the nodes between dependent GBKs, so I implemented a new version of the split algorithm
that takes advantage of how we've propagated support for multiple outputs on both the map
and reduce sides of a job to do finer-grained splits that use information from the scaleFactor
calculations to make smarter split decisions.
> One high-level change along with this: I changed the default scaleFactor() value in DoFn
to 0.99f to slightly prefer writes that occur later in a pipeline flow by default.

This message was sent by Atlassian JIRA

View raw message