Modular Specification of Interaction Policiesin

Software executing on distributed systems consists of many asynchronous, autonomous components which interact in order to coordinate local activity. The need for such coordination, as well as requirements such as heterogeneity, scalability, security and availability, considerably increase the complexity of code in distributed applications. Moreover, changing requirements, as well as changes in hardware platforms, lead to software that is constantly evolving and complicates reuse. To support development and evolution of distributed applications requires techniques which allow coordination code to be speci ed, customized, and maintained independently of application components; goals which cannot be realized solely through object-oriented techniques. This thesis demonstrates that meta-level speci cation of interaction policies enables modular description of component interaction policies, as well as customization of policy implementations. We present the high-level language Dil for speci cation of protocols. The language identi es the abstractions and concepts required for modular speci cation of interaction policies. The abstractions developed in Dil are then realized through a meta-level architecture. We provide several Dil protocols implementing policies for reliable communication, atomicity, and dependability in a multi-media application. The implementation techniques necessary to realize the meta-architecture are discussed in the context of Broadway, a run-time system for distributed C++ objects. This meta-architecture provides access to the system-level features necessary to support the modularity mechanisms de ned in Dil. We provided examples of protocol translation from Dil to Broadway C++ classes that utilize the meta-architecture. Measurements show that these techniques may be implemented with minimal performance overhead. A formal semantics for reasoning about the approach is also developed. iii To my parents, Joel and Diane iv Acknowledgements I wish thank my parents, Joel and Diane, whose continual love and encouragement have been an essential ingredient in the development of this thesis. Their continuous queries as to whether or not \my thesis was done yet" and, in general, their never-ending interest in my work and accomplishments, has always kept me oriented and on track. I thank my brother and oldest friend, Michael, who I will not ask to read this thesis, hoping he returns the favor. I would also like to thank Lily Chu, who has been my best friend, psychotherapist, navigator, fellow gardener and chef, providing love and support, not to mention delicious deserts, when I need them most. My doctoral committee | Gul Agha, Roy Campbell, and Prithviraj Banerjee | have been invaluable through their advice and comments. I would especially like to thank my advisor, Gul Agha, for his mentorship and support, for always treating me as a peer, and for providing an open and exible research environment. I would also like to thank Carolyn Talcott for her feed-back and comments and, in particular, for her advice in the development of my operational semantics. I have bene ted greatly from countless discussions with members of the Open Systems Lab. In particular I would like to thank Svend Fr lund, Rajendra Panwar, Mark Astley, Brian Nielsen, Shangping Ren, and WooYoung Kim for their input and contributions. In addition, I would like to thank Theron Tock, Amitabh Dave, and Nayeem Islam1 for always providing a fresh outlook outside that of the OSL. Finally, I would like to thank the members of the Illini Ju-Juitsu Club who have provided an essential outlet for all my stress and frustration throughout my tenure in Champaign-Urbana. I am particularly thankful to Doug Musser and Greg Schomburg for their personal instruction and for serving as examples through their dedication. 1Nayeem Islam chose the name Broadway for my run-time Actor system. v Table of