Package org.apache.flink.runtime.scheduler.adaptivebatch.util

Class PointwiseVertexInputInfoComputer

java.lang.Object

org.apache.flink.runtime.scheduler.adaptivebatch.util.PointwiseVertexInputInfoComputer

public class PointwiseVertexInputInfoComputer
extends Object

Helper class that computes VertexInputInfo for pointwise input.

Constructor Summary

Constructors

Constructor	Description
PointwiseVertexInputInfoComputer()

Method Summary

Modifier and Type	Method	Description
Map<IntermediateDataSetID,JobVertexInputInfo>	compute(List<BlockingInputInfo> inputInfos, int parallelism, int minParallelism, int maxParallelism, long dataVolumePerTask)	Decide parallelism and input infos, which will make the data be evenly distributed to downstream subtasks for POINTWISE, such that different downstream subtasks consume roughly the same amount of data.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Details

PointwiseVertexInputInfoComputer

public PointwiseVertexInputInfoComputer()

Method Details

compute

public Map<IntermediateDataSetID,JobVertexInputInfo> compute(List<BlockingInputInfo> inputInfos,
 int parallelism,
 int minParallelism,
 int maxParallelism,
 long dataVolumePerTask)

Decide parallelism and input infos, which will make the data be evenly distributed to downstream subtasks for POINTWISE, such that different downstream subtasks consume roughly the same amount of data.

Assume that `inputInfo` has two partitions, each partition has three subpartitions, their data bytes are: {0->[1,2,1], 1->[2,1,2]}, and the expected parallelism is 3. The calculation process is as follows:
1. Create subpartition slices for input which is composed of several subpartitions. The created slice list and its data bytes are: [1,2,1,2,1,2]
2. Distribute the subpartition slices array into n balanced parts (described by `IndexRange`, named SubpartitionSliceRanges) based on data volume: [0,1],[2,3],[4,5]
3. Reorganize the distributed results into a mapping of partition range to subpartition range: {0 -> [0,1]}, {0->[2,2],1->[0,0]}, {1->[1,2]}.
The final result is the `SubpartitionGroup` that each of the three parallel tasks need to subscribe.

Parameters:

inputInfos - The information of consumed blocking results

parallelism - The parallelism of the job vertex

minParallelism - the min parallelism

maxParallelism - the max parallelism

dataVolumePerTask - proposed data volume per task for this set of inputInfo

Returns:

the parallelism and vertex input infos