A Temporal Approach for Testing Distributed Systems-A MapReduce Case Study

Abstract

To observe the behavior of a distributed implementation under test (IUT), we commonly coordinate parallel testers in order to check not only if the output events have been observed, but also the dates when these events have been occurred. However, the use of multiple testers can produce several coordination problems amongst remote testers known as Controllability and Observability issues. To cope with such problems, we suggest-in previous works- the inclusion of some timing constraints in the test architecture. In this paper, we choose as a case study the use of a MapReduce architecture with several worker components running essentially two functionalities: Map and Reduce tasks. These systems often face failures caused by various conditions, e.g., network connection delays, power outages or hardware problems. Moreover, we define the temporal properties in the specification of such systems. Then, we explain how to apply these properties in the case of MapReduce testing to identify faulty workers, in order to reschedule their tasks to a healthy worker.