[
https://issues.apache.org/jira/browse/TEZ-1231?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14101678#comment-14101678
]
Siddharth Seth commented on TEZ-1231:
-------------------------------------
Looks good. DATA_OVER_EVENTS should be marked private - some of those config
fields will be removed soon.
Couple more renames required.
TEZ_RUNTIME_SHUFFLE_INPUT_BUFFER_PERCENT ->
TEZ_RUNTIME_SHUFFLE_FETCH_BUFFER_PERCENT (percentage of heap to use while
fetching data)
TEZ_RUNTIME_INPUT_BUFFER_PERCENT -> TEZ_RUNTIME_INPUT_POST_MERGE_BUFFER_PERCENT
(percentage of heap to use after merge is complete (while the iterator is being
served to the processor)
> Clean up TezRuntimeConfiguration
> --------------------------------
>
> Key: TEZ-1231
> URL: https://issues.apache.org/jira/browse/TEZ-1231
> Project: Apache Tez
> Issue Type: Improvement
> Reporter: Siddharth Seth
> Assignee: Bikas Saha
> Priority: Blocker
> Fix For: 0.5.0
>
> Attachments: TEZ-1231.1.patch
>
>
> Separate configs used by runtime Inputs/Outputs and framework specific
> configs As an example - task memory configs are not related to any Input /
> Output. Similarly for Credentials.
> Other possible config changes
> - Rename TEZ_RUNTIME_INPUT_BUFFER_PERCENT, which is a fairly confusing name
> - Get rid of the separation between INPUT/OUTPU in config names (e.g.
> INPUT_KEY, OUTPUT_KEY replaced with just KEY)
> - Move TEZ_RUNTIME_INTERMEDIATE_INPUT_KEY_SECONDARY_COMPARATOR_CLASS to
> MapReduce
--
This message was sent by Atlassian JIRA
(v6.2#6252)
