2018/03/31 by Ashish Tapdiya, Daniel Fabbri, Tapdiya, Ashish +1 · 1 voice
Computer Science · #Cloud Computing and Resource Management #Data Management and Algorithms #Advanced Database Systems and Queries
paper · pdf · doi:10.48550/arxiv.1804.00224
Hadoop is emerging as the primary data hub in enterprises, and SQL represents\nthe de facto language for data analysis. This combination has led to the\ndevelopment of a variety of SQL-on-Hadoop systems in use today. While the\nvarious SQL-on-Hadoop systems target the same class of analytical workloads,\ntheir different architectures, design decisions and implementations impact\nquery performance. In this work, we perform a comparative analysis of four\nstate-of-the-art SQL-on-Hadoop systems (Impala, Drill, Spark SQL and Phoenix)\nusing the Web Data Analytics micro benchmark and the TPC-H benchmark on the\nAmazon EC2 cloud platform. The TPC-H experiment results show that, although\nImpala outperforms other systems (4.41x - 6.65x) in the text format, trade-offs\nexists in the parquet format, with each system performing best on subsets of\nqueries. A comprehensive analysis of execution profiles expands upon the\nperformance results to provide insights into performance variations,\nperformance bottlenecks and query execution characteristics.\n