Abstract
MapReduce is an effective tool for processing large amounts of data in parallel using a cluster of processors or computers. One common data processing task is the join operation, which combines two or more datasets based on values common to each. In this paper, we present a network aware multi-way join for MapReduce (SmartJoin) that improves performance and considers network traffic when redistributing workload amongst reducers. SmartJoin achieves this by dynamically redistributing tuples directly between reducers with an intelligent network aware algorithm. We show that our presented technique has significant potential to minimize the time required to join multiple datasets. In our evaluation, we show that SmartJoin has up to 39 % improvement compared to the non-redistribution method, a 26.8 % improvement over random redistribution and 27.6 % improvement over worst join redistribution.
Original language | English |
---|---|
Pages (from-to) | 629-641 |
Number of pages | 13 |
Journal | Cluster Computing |
Volume | 17 |
Issue number | 3 |
DOIs | |
State | Published - Sep 2014 |
Keywords
- Hadoop
- MapReduce
- Multiway join
- Workload redistribution