Automatic detection of errors in distributed systems

Debugging in a distributed environment is very complex when compared to debugging in a uniprocessor or sequential environment. The order of events in a distributed environment is not deterministic, and this order of events may some times produce some unexpected errors. These errors which depend on the timing of events are typical of distributed systems. In this paper we proposed an approach based on the timing graphs concept to debug these timing errors. Here we give some conditions which should be satisfied by a timing graph, if the execution of that instance of the program, from which the graph is constructed, doesn’t have any errors. The graphs are constructed using the information gathered at the run time using tracing mechanism. This approach uses topological sorting of directed acyclic graphs in analyzing the timing graphs. In this paper, we have developed the theory behind this approach and presented an algorithm to automate the procedure. Since the manual approach to this problem is cumbersome and error prone, this automatic error detection procedure simplifies the detection of the cause of the error for some standard error patterns.