Distributed systems have received increased interest in recent years. Distributed systems offer numerous advantages over centralized systems such as increased reliability and availability. Even with increasingly reliable hardware it is unreasonable to expect distributed systems to run without failures. Given that failures occur, how can a distributed system maintain consistency in the face of failures and concurrency. The Clouds project is one of many groups studying techniques to exploit the potentials of distributed systems.;Unfortunately, simplicity of coding is not one of the advantages of distributed systems. More often than not distributed programs are more difficult to construct than their centralized counterparts. This difficulty is often due to lack of tools and methodologies to support the programmer in the debugging process. In addition, characteristics of the object-action paradigm of programming supported by Clouds require the design of new representations of program execution.;This dissertation explores debugging methodologies and techniques appropriate for object-action programs running in a distributed environment. Problems with constructing a distributed debugger are identified and analyzed. In this dissertation, the design of a prototype distributed debugger is presented. Included is an algorithm which uses the semantics of object-action computations to allow interactive debugging of distributed programs. The debugger allows a user to debug a distributed program from multiple viewpoints, at various abstraction levels, and with various degrees of control over program execution. The design of a tool capable of detecting potential time-dependent errors is also included.
展开▼
