I use Apache Flink to process the data stream, and I need to split the index between all nodes processing the input data. The index is often updated by nodes.
I would like to know if it is good from the point of view of efficiency to distribute a data set through broadcast variables?
Will the broadcast variable be updated in all nodes after each update or not?
Is Apache Flink wise to update broadcast variables only for recent changes or not?
source
share