org.apache.spark.storage

Class BasicBlockReplicationPolicy

Object

org.apache.spark.storage.BasicBlockReplicationPolicy

All Implemented Interfaces:

Logging, BlockReplicationPolicy

public class BasicBlockReplicationPolicy
extends Object
implements BlockReplicationPolicy, Logging

Constructor Summary

Constructors

Constructor and Description
BasicBlockReplicationPolicy()

Method Summary

Modifier and Type	Method and Description
scala.collection.immutable.List<BlockManagerId>	prioritize(BlockManagerId blockManagerId, scala.collection.Seq<BlockManagerId> peers, scala.collection.mutable.HashSet<BlockManagerId> peersReplicatedTo, BlockId blockId, int numReplicas) Method to prioritize a bunch of candidate peers of a block manager.

Methods inherited from class Object

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

Methods inherited from interface org.apache.spark.internal.Logging

initializeLogging, initializeLogIfNecessary, initializeLogIfNecessary, isTraceEnabled, log_, log, logDebug, logDebug, logError, logError, logInfo, logInfo, logName, logTrace, logTrace, logWarning, logWarning

Constructor Detail

BasicBlockReplicationPolicy

public BasicBlockReplicationPolicy()

Method Detail

prioritize

public scala.collection.immutable.List<BlockManagerId> prioritize(BlockManagerId blockManagerId,
                                                                  scala.collection.Seq<BlockManagerId> peers,
                                                                  scala.collection.mutable.HashSet<BlockManagerId> peersReplicatedTo,
                                                                  BlockId blockId,
                                                                  int numReplicas)

Method to prioritize a bunch of candidate peers of a block manager. This implementation replicates the behavior of block replication in HDFS. For a given number of replicas needed, we choose a peer within the rack, one outside and remaining blockmanagers are chosen at random, in that order till we meet the number of replicas needed. This works best with a total replication factor of 3, like HDFS.

Specified by:

prioritize in interface BlockReplicationPolicy

Parameters:

blockManagerId - Id of the current BlockManager for self identification

peers - A list of peers of a BlockManager

peersReplicatedTo - Set of peers already replicated to

blockId - BlockId of the block being replicated. This can be used as a source of randomness if needed.

numReplicas - Number of peers we need to replicate to

Returns:

A prioritized list of peers. Lower the index of a peer, higher its priority