This book shows how to program a system so that the software continues to work in the face of a variety of failures in parts of the system.
The revolution in the use of distributed computing systems makes their reliability an important concern. While it is generally well understood how to connect hardware so that most components can continue to work when others fail, connecting reliable software remains a problem. This book shows how to program a system so that the software continues to work in the face of a variety of failures in parts of the system. It introduces the concept of nested transactions along with techniques for implementation, algorithms for concurrency control, recovery, distributed commitment, and deadlock detection and avoidance. In addition, it provides a convenient and clear exposition of a number of techniques used in reliability and concurrency control from current literature. The transactions the book presents are collections of primitive actions that are indivisible, thus insuring that consistent results are obtained even when requests are processed concurrently or failures occur during a request. The design permits these transactions to be nested, providing universes of synchronization and recovery from failures. The advantages of nested over single-level transactions are that they provide concurrency control by serializing subtransactions appropriately, and permit parts of a transaction to fail without necessarily aborting the entire transaction.The nested transaction approach to reliable distributed computing will be of particular value to students and researchers in information systems and databases, and professionals designing or implementing databases or information systems.
J. Eliot B. Moss received his doctorate in computer science from MIT. This book inaugurates The MIT Press Series in Information Systems (Research Reports and Notes), edited by Michael Lesk.
