Flink-ER: An Elastic Resource-Scheduling Strategy for Processing Fluctuating Mobile Stream Data on Flink

Algorithm 3

Data redistribution algorithm.

	Input:
	Operator to be rescaled: operator[]
	Resource pool: pool
	Output:
	Rescaled flow network: G′
	Begin
	foreach O ∈ operator[] do
	foreach  ∈ O do
	Send the stateful data of vertex to the HDFS and the data handler to Zookeeper.
	end foreach
	← pool.getNode()
	O.add()/^∗Enlarge the parallelism of the Operator^∗/
	bucket[]←JobManager.remapping(bucket[],|O|)
	/^∗Remapping stateful data as shown in Figure 5^∗/
	foreach  ∈ O do
	Require stateful data from Zookeeper and restore the corresponding stateful data from the HDFS
	end foreach
	end foreach
	flowNetwork_selfLearning(G′)
	return G′
	End