QA Testing for a Big Data Management System: A Case Study.

摘要

This thesis reports on the outcome of a six-month case study involving software testing of a large, open source code base for Big Data Management. The focus is two-fold, covering issues related to both testing a declarative query language and to the special challenges related to ensuring its correctness when data is partitioned and the execution of queries is parallelized. The system under study, AsterixDB, is a Big Data Management System (BDMS) that uses a shared-nothing parallel architecture. The parallel query processing abilities of the system, like different kinds of parallel joins (Hybrid Hash Join, Index Nested Loop Join and Nested Loop Join), aggregate functions, group by, order by, and limits, are extensively tested in a parallel shared-nothing environment.