TOD is a tool that helps understand a program's behavior by recording a trace of the events that occur during execution (method calls, object creation, field access…) and letting the user navigate in predefined and custom views of this trace.
The official website of TOD is here: http://pleiad.dcc.uchile.cl/tod.
It is often harder to find the cause of a bug than fix it. This is due to the fact that traditional debugging tools provide a very limited view of the computations, and usually force the programmer to repeatedly run the program, narrowing down the focus area until the cause of the problem is found. This methodology, which is tedious when applied to single-machine programs, becomes completely impractical in a distributed context: traditional debugging tools can't provide a global view of the computations on the whole grid; they instead provide a per-host view.
Check the videos here.