Package org.apache.flink.runtime.io.network.partition.hybrid

Class HsSpillingStrategyUtils

java.lang.Object

org.apache.flink.runtime.io.network.partition.hybrid.HsSpillingStrategyUtils

public class HsSpillingStrategyUtils
extends Object

Utilities for HsSpillingStrategy.

Constructor Summary

Constructors

Constructor	Description
HsSpillingStrategyUtils()

Method Summary

Modifier and Type	Method	Description
static TreeMap<Integer,List<BufferIndexAndChannel>>	getBuffersByConsumptionPriorityInOrder(List<Integer> nextBufferIndexToConsume, TreeMap<Integer,Deque<BufferIndexAndChannel>> subpartitionToAllBuffers, int expectedSize)	Calculate and get expected number of buffers with the highest consumption priority.

Methods inherited from class java.lang.Object

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

Constructor Detail

HsSpillingStrategyUtils

public HsSpillingStrategyUtils()

Method Detail

getBuffersByConsumptionPriorityInOrder

public static TreeMap<Integer,List<BufferIndexAndChannel>> getBuffersByConsumptionPriorityInOrder(List<Integer> nextBufferIndexToConsume,
                                                                                                        TreeMap<Integer,Deque<BufferIndexAndChannel>> subpartitionToAllBuffers,
                                                                                                        int expectedSize)

Calculate and get expected number of buffers with the highest consumption priority. For each buffer, The greater the difference between next buffer index to consume of subpartition it belongs to and buffer index, the higher the priority.

Parameters:

nextBufferIndexToConsume - downstream next buffer index to consume.

subpartitionToAllBuffers - the buffers want to compute priority, are grouped by subpartitionId.

expectedSize - number of result buffers.

Returns:

mapping for subpartitionId to buffers, the value of map entry must be order by bufferIndex ascending.