Package org.apache.flink.runtime.io.network.api.writer

Enum SubtaskStateMapper

java.lang.Object

java.lang.Enum<SubtaskStateMapper>

org.apache.flink.runtime.io.network.api.writer.SubtaskStateMapper

All Implemented Interfaces:

Serializable, Comparable<SubtaskStateMapper>

@Internal
public enum SubtaskStateMapper
extends Enum<SubtaskStateMapper>

The SubtaskStateMapper narrows down the subtasks that need to be read during rescaling to recover from a particular subtask when in-flight data has been stored in the checkpoint.

Mappings of old subtasks to new subtasks may be unique or non-unique. A unique assignment means that a particular old subtask is only assigned to exactly one new subtask. Non-unique assignments require filtering downstream. That means that the receiver side has to cross-verify for a deserialized record if it truly belongs to the new subtask or not. Most SubtaskStateMapper will only produce unique assignments and are thus optimal. Some rescaler, such as RANGE, create a mixture of unique and non-unique mappings, where downstream tasks need to filter on some mapped subtasks.

Enum Constant Summary

Enum Constants

Enum Constant	Description
ARBITRARY	Extra state is redistributed to other subtasks without any specific guarantee (only that up- and downstream are matched).
FIRST	Restores extra subtasks to the first subtask.
FULL	Replicates the state to all subtasks.
RANGE	Remaps old ranges to new ranges.
ROUND_ROBIN	Redistributes subtask state in a round robin fashion.
UNSUPPORTED

Method Summary

Modifier and Type	Method	Description
RescaleMappings	getNewToOldSubtasksMapping(int oldParallelism, int newParallelism)	Returns a mapping new subtask index to all old subtask indexes.
abstract int[]	getOldSubtasks(int newSubtaskIndex, int oldNumberOfSubtasks, int newNumberOfSubtasks)	Returns all old subtask indexes that need to be read to restore all buffers for the given new subtask index on rescale.
boolean	isAmbiguous()	Returns true iff this mapper can potentially lead to ambiguous mappings where the different new subtasks map to the same old subtask.
static SubtaskStateMapper	valueOf(String name)	Returns the enum constant of this type with the specified name.
static SubtaskStateMapper[]	values()	Returns an array containing the constants of this enum type, in the order they are declared.

Methods inherited from class java.lang.Enum

clone, compareTo, equals, finalize, getDeclaringClass, hashCode, name, ordinal, toString, valueOf

Methods inherited from class java.lang.Object

getClass, notify, notifyAll, wait, wait, wait

Enum Constant Details

ARBITRARY

public static final SubtaskStateMapper ARBITRARY

Extra state is redistributed to other subtasks without any specific guarantee (only that up- and downstream are matched).

FIRST

public static final SubtaskStateMapper FIRST

Restores extra subtasks to the first subtask.

FULL

public static final SubtaskStateMapper FULL

Replicates the state to all subtasks. This rescaling causes a huge overhead and completely relies on filtering the data downstream.

This strategy should only be used as a fallback.

RANGE

public static final SubtaskStateMapper RANGE

Remaps old ranges to new ranges. For minor rescaling that means that new subtasks are mostly assigned 2 old subtasks.

Example:
old assignment: 0 -> [0;43); 1 -> [43;87); 2 -> [87;128)
new assignment: 0 -> [0;64]; 1 -> [64;128)
subtask 0 recovers data from old subtask 0 + 1 and subtask 1 recovers data from old subtask 1 + 2

For all downscale from n to [n-1 .. n/2], each new subtasks get exactly two old subtasks assigned.

For all upscale from n to [n+1 .. 2*n-1], most subtasks get two old subtasks assigned, except the two outermost.

Larger scale factors (<n/2, >2*n), will increase the number of old subtasks accordingly. However, they will also create more unique assignment, where an old subtask is exclusively assigned to a new subtask. Thus, the number of non-unique mappings is upper bound by 2*n.

ROUND_ROBIN

public static final SubtaskStateMapper ROUND_ROBIN

Redistributes subtask state in a round robin fashion. Returns a mapping of newIndex -> oldIndexes. The mapping is accessed by using Bitset oldIndexes = mapping.get(newIndex).

For oldParallelism < newParallelism, that mapping is trivial. For example if oldParallelism = 6 and newParallelism = 10.

New index	Old indexes
0	0
1	1
...
5	5
6
...
9

For oldParallelism > newParallelism, new indexes get multiple assignments by wrapping around assignments in a round-robin fashion. For example if oldParallelism = 10 and newParallelism = 4.

New index	Old indexes
0	0, 4, 8
1	1, 5, 9
2	2, 6
3	3, 7

UNSUPPORTED

public static final SubtaskStateMapper UNSUPPORTED

Method Details

values

public static SubtaskStateMapper[] values()

Returns an array containing the constants of this enum type, in the order they are declared.

Returns:

an array containing the constants of this enum type, in the order they are declared

valueOf

public static SubtaskStateMapper valueOf(String name)

Returns the enum constant of this type with the specified name. The string must match exactly an identifier used to declare an enum constant in this type. (Extraneous whitespace characters are not permitted.)

Parameters:

name - the name of the enum constant to be returned.

Returns:

the enum constant with the specified name

Throws:

IllegalArgumentException - if this enum type has no constant with the specified name

NullPointerException - if the argument is null

getOldSubtasks

public abstract int[] getOldSubtasks(int newSubtaskIndex,
 int oldNumberOfSubtasks,
 int newNumberOfSubtasks)

Returns all old subtask indexes that need to be read to restore all buffers for the given new subtask index on rescale.

getNewToOldSubtasksMapping

public RescaleMappings getNewToOldSubtasksMapping(int oldParallelism,
 int newParallelism)

Returns a mapping new subtask index to all old subtask indexes.

isAmbiguous

public boolean isAmbiguous()

Returns true iff this mapper can potentially lead to ambiguous mappings where the different new subtasks map to the same old subtask. The assumption is that such replicated data needs to be filtered.