You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@tez.apache.org by "Adrian Nicoara (JIRA)" <ji...@apache.org> on 2018/12/25 03:42:00 UTC
[jira] [Created] (TEZ-4029) tez-api: InputDataInformationEvent:
decouple payload storage from access
Adrian Nicoara created TEZ-4029:
-----------------------------------
Summary: tez-api: InputDataInformationEvent: decouple payload storage from access
Key: TEZ-4029
URL: https://issues.apache.org/jira/browse/TEZ-4029
Project: Apache Tez
Issue Type: Improvement
Reporter: Adrian Nicoara
Attachments: 0001-tez-api-InputDataInformationEvent-decouple-payload-s.patch
The current InputDataInformationEvent has different fields, constructors and
methods of access for a payload, depending on whether it is stored in Object
form, or if it is stored as a serialized payload within a ByteBuffer.
This can be simplified to having a single Object reference to the payload.
Access, at runtime, checks if the payload has to be (de)serialized, depending on
which method is used, and what is stored.
In addition to the one reference that this saves in the
InputDataInformationEvent, user code can leverage the deserialized storage
format, to reduce the footprint associated with the same payload being used
across different events within the AM.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)