[
https://issues.apache.org/jira/browse/HUDI-316?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Venkatesh Rudraraju updated HUDI-316:
-------------------------------------
Description:
* Repartition WriteStatus before index writes, in a way that each WriteStatus
with new records are not clubbed together.
* This repartition will improve parallelism for this hbase index operation.
* In HBaseIndex puts call, there is a sleep of 100 millis for each batch of
puts. This implementation assumes negligible time for puts, but for large
batches of puts it is inefficient.
* Using rate limiter will be efficient compared to sleep as it accounts for
the time taken for puts as well.
was:
* In HBaseIndex puts call, there is a sleep of 100 millis for each batch of
puts. This implementation assumes negligible time for puts, but for large
batches of puts it is inefficient.
* Using rate limiter will be efficient compared to sleep as it accounts for
the time taken for puts as well.
> Improve performance of HbaseIndex puts by repartitioning WriteStatus and
> using rate limiter instead of sleep()
> --------------------------------------------------------------------------------------------------------------
>
> Key: HUDI-316
> URL: https://issues.apache.org/jira/browse/HUDI-316
> Project: Apache Hudi (incubating)
> Issue Type: Task
> Components: Index
> Reporter: Venkatesh Rudraraju
> Assignee: Venkatesh Rudraraju
> Priority: Major
>
> * Repartition WriteStatus before index writes, in a way that each WriteStatus
> with new records are not clubbed together.
> * This repartition will improve parallelism for this hbase index operation.
> * In HBaseIndex puts call, there is a sleep of 100 millis for each batch of
> puts. This implementation assumes negligible time for puts, but for large
> batches of puts it is inefficient.
> * Using rate limiter will be efficient compared to sleep as it accounts for
> the time taken for puts as well.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
