Considering Data Skew in Multi-way Joins for MapReduce

Data analyzing and processing are important tasks in cloud computing. The MapReduce can provide a cost-effective, flexible, fault-tolerant and scalable distributed programming model over large clusters. However, how to implement join operation using MapReduce efficiently is an attractive point. Data skew problem has a strong impact on the performance of join operation. In this paper, we implement the range partition method based on the way of sampling, and apply it to multi-way joins to avoid the influence of data skew. The results of the experiments we have conducted show that our approach is more efficient than current algorithms.