Online Event - Based Program Visualization for Distributed Systems

Concurrent systems maintain a distributed state and thus require coordination and synchronization between components to ensure consistency. To provide a coherent design approach to concurrent systems, recent work has employed an object-based methodology which emphasizes interactions through well-de ned interfaces. The Actor model has provided formal reasoning about distributed object systems. Nonetheless, due to the complex interactions among components and the high volume of observable information produced, understanding and reasoning about concurrent algorithms in terms of simple interactions is a di cult task. Coordination patterns, which abstract over simple interactions, are not biased by low-level event orderings and are the appropriate mechanism for reasoning about concurrent algorithms. In this thesis, we present a methodology for visualizing coordination patterns in concurrent algorithms which emphasizes observable interactions and causal connections between objects. We introduce visualization groups as an intuitive notion for mapping coordination patterns to visualization. Visualization groups are speci ed linguistically using visualizers. Visualizers are speci ed separately from algorithm code and thus respect code integrity. Moreover, visualizers may be implemented strictly in terms of object interfaces and thus preserve object encapsulation. We describe the implementation of stagehand, a prototype environment which supports visualizers for the purpose of specifying visualization over actor computations implemented on the actor platform broadway. iii Acknowledgements First and foremost, I would like to thank my advisor, Gul Agha, for providing the foundational ideas behind much of my research. In addition, I would like to thank Helena Mitasova, Bill Brown, and the other ne folks at the U.S. Army Construction Engineering Research Laboratory for providing a source of funding and tolerating my moonlighting as a graduate student. I would also like to thank the other members of the Open System Laboratory, both past and present, for their input and contributions. In particular, I would like to thank Dan (Anybody want some co ee?) Sturman, Brian (How much co ee should I make?) Nielsen, Shangping (I don't want any co ee, honest) Ren, and Rajendra Panwar for their invaluable contributions to this work. I would like to thank Suzanne (you know who you are) for keeping me well fed and stress free for the duration. Lastly, I would like to thank my brother Scott and my parents for being there through good times and bad. iv Table of