You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@beam.apache.org by "Ian Zhou (JIRA)" <ji...@apache.org> on 2016/06/15 21:08:09 UTC
[jira] [Created] (BEAM-347) Progress updates inaccurate for
non-uniform keys in Bigtable
Ian Zhou created BEAM-347:
-----------------------------
Summary: Progress updates inaccurate for non-uniform keys in Bigtable
Key: BEAM-347
URL: https://issues.apache.org/jira/browse/BEAM-347
Project: Beam
Issue Type: Improvement
Components: sdk-java-gcp
Reporter: Ian Zhou
Assignee: Daniel Halperin
Priority: Minor
When reading from a Bigtable source with clustered keys, fraction consumed progress updates are inaccurate. For example, for a range spanning ['a', 'z'], a cluster of keys starting with the letter 'm' (e.g. 'me100,' ..., 'me999') will be recorded as ~50% complete upon reading the first key, and will remain at this percentage until the final key has been read. Instead, the start of the range should be changed to the first key read (e.g. new range ['me100', 'z']). The end of the range can be changed over time through dynamic work rebalancing.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)