Roman Khachatryan created FLINK-18990:
-----------------------------------------
Summary: Optimize reading ResultSubpartition state
Key: FLINK-18990
URL: https://issues.apache.org/jira/browse/FLINK-18990
Project: Flink
Issue Type: Improvement
Components: Runtime / Task
Reporter: Roman Khachatryan
Assignee: Roman Khachatryan
Fix For: 1.12.0
Curently, channel state is read not sequentially.
Inverting control would make it more efficient.
Current call chain:
{code:java}
StreamTask.readRecoveredChannelState
ResultPartition.readRecoveredState - loop through subpartitions
PipelinedSubpartition.readRecoveredState - loop while have data;
bufferBuilder =
parent.getBufferPool().requestBufferBuilderBlocking(subpartitionInfo.getSubPartitionIdx());
ChannelStateReader.readOutputData {code}
Proposed call chain:
{code:java}
StreamTask.readRecoveredChannelState
ChannelStateReader.readOutputData loop through state handles ordererd by
handle, offset
request buffer in the same way: BufferBuilder bufferBuilder =
resPart.getBufferPool().requestBufferBuilderBlocking(subpartitionInfo.getSubPartitionIdx());
pass to resPart.getSubpartition(idx).add(BufferConsumer, boolean, boolean)
{code}
Depends on FLINK-18989
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
