Journal of East China Normal University(Natural Science) ›› 2020, Vol. 2020 ›› Issue (4): 98-107.doi: 10.3969/j.issn.1000-5641.201921011

• Computer Science • Previous Articles     Next Articles

Lineage-driven distributed database system with an automated fault injection test tool

SHEN Jing, CAI Peng   

  1. School of Data Science and Engineering, East China Normal University, Shanghai 200062, China
  • Received:2019-08-22 Published:2020-07-20

Abstract: Failures are unavoidable in distributed database systems. To improve the fault tolerance of distributed database systems and verify the accuracy of fault-tolerant protocols, the system should periodically run a fault injection test to artificially trigger a fault during system operations. However, the scale and complexity of distributed database systems make it difficult to fully enumerate inputs and make it impractical to explain all the behaviors that occur in the system. One of the test methods commonly used is a random combination of faults, which is simple but not complete; the other one is guided by professional knowledge and is not universally applicable. Accordingly, we adopted and revised a research prototype, called the lineage-driven fault injection (LDFI) test, that is both complete and universally applicable. We implemented the automation fault injection tool in Cedar. Experiments showed that lineage-driven fault injection tests can successfully detect system bugs caused by complex fault combinations and improve the credibility of the database; these bugs cannot be detected by random fault injection with fewer test cases.

Key words: distributed, data lineage, automation, fault injection

CLC Number: