Abstract: Consider the problem of providing a logical channel for message exchange between two user processes in a network environment. When is protocol conversion needed? To answer this question, we first define a model of layered architectures. Specifically, three stepwise refinement rules are given. Any architecture that can be obtained by a sequence of applications of the stepwise refinement rules is said to be well-structured. We show that this class of well-structured architectures has several correctness properties. It is also very general and includes many well-known networking and internetworking architectures in the literature. Logical connectivity in such an architecture is defined recursively. As a result, to determine if a logical channel can be provided between two user processes, it is sufficient to examine peer protocols specified for each level of the architecture's hierarchy of processes one at a time. Thus the original problem reduces to the problem of determining if a set of processes will interoperate.When protocol conversion is needed to achieve interoperability between processes that implement different protocols, how should it be done? How does one prove that a conversion is correct? What is meant by a correct conversion? We propose the use of projections and image protocols (previously developed by Lam and Shankar for protocol verification [10]) for specifying conversions and for reasoning about the correctness of conversions. Given two processes implementing different protocols P and Q, our objective is to find the largest protocol that is an image protocol of P as well as Q. The correctness of the conversion is a consequence of the correctness properties of image protocols.There are several open problems. Most importantly, heuristics are used for finding the necessary image protocol for conversion. Although, an image protocol common to both P and Q can always be found, it may not be easy to find one with useful functionality. There are also some implementation and design issues to be addressed, such as: the construction of converters that are transparent and converters that add functionality to an image protocol common to P and Q.

Abstract: Protocols can be viewed as predefined sequences of message exchanges between machines for performing network control functions and for providing network services. One way of modeling protocols is by using communicating finite state machines. The interaction between finite state machines can be very involved, even for machines with few states. To counteract the complexity for protocol analysis and synthesis, we propose a partitioning method based on protocol subgraphs. We found that if there are ‘cross interactions’ involving the entire protocol graph, then the protocol is not decomposable by our technique and must be analyzed or synthesized as a whole. However, if there are subunits of message exchanges within the protocol graph that are self-contained, in other words, if there are no ‘cross interactions’ between protocol subgraphs, then the protocol is decomposable. For protocols that are decomposable, we show that it is only necessary to examine a subspace of the entire reachability space to understand the behavior of the protocol and to guarantee its progress properties. This allows us to analyze and synthesize protocols based on these subgraphs.

Abstract: We present a resilient distributed protocol that enables a synchronous algorithm to run on an asynchronous network. The protocol is resilient in the sense that it can continue providing network synchronization in the presence of topological changes in the underlying communication network of a distributed system. These changes are caused by link/node failures and recoveries that occur while running the protocol. In general, the protocol is a useful tool in the design of resilient distributed algorithms as it isolates the algorithm from the characteristics of the communication network.

Abstract: The designer of communication protocols has to formulate rules to govern the communications between processes that are distributed; share common resources concurrently and asynchronously; communicate through unreliable channels that incur random delays; and behave in a time-dependent fashion. The first step is to formally specify the behavior of each of the communicating processes in the protocol. The protocol designer then has to analyze their concurrent behavior to ensure that it satisfies given functional requirements. He also has to analyze their timing behavior to ensure that is meets given timing requirements, and to evaluate key measures that indicate the efficiency of their performance. In this thesis we address the specification and analysis of timing requirements and performance measures of protocols. Contributions of the thesis fall into three major categories. The first major contribution is the development of a methodology for automated performance analysis of protocols that is based on their formal specifications. Rules that map an algebraic specification of a protocol, and the exponential rates of its events times, to probability and time attributes of its timing behavior are devised. The methodology constitutes two main steps. First, the functional behavior of a protocol is specified algebraically and analyzed to detect any progress errors. Second, protocol timing requirements and performance measures are formally specified in terms of the timing attributes and automatically analyzed. The second major contribution is the design and development of ANALYST: a software environment that integrates functional and performance specification and analysis of protocols. This integration allows a protocol designer to analyze protocol performance without requiring much expertise in the field. It also facilitates and enhances the design process of protocols by supporting experimental design, and providing an interactive and uniform user interface to the supported functions. The third major contribution is the automated derivation of performance analysis and optimum timing of a connection establishment protocol, the Alternating Bit protocol, and a two phase locking protocol. Some results obtained are novel while others are shown to agree remarkably well with those obtained manually by other researchers. Finally, limitations of the proposed methodology, and directions for future research are outlined.

Abstract: The tier automation is presented as a model for communication protocols. Several advantages of the model are cited, among which are universality of representation and manipulability. A scheme for using the tier automation to model specific distributed architectures and their protocols is described. The scheme is then used on a sample protocol and a transmission session with the sample protocol as exhibited.