Guardians for Concurrent Systems

In this paper we survey the current state of the art on fundamental aspects of concurrent systems. We discuss the notion of concurrency and discuss a model of computation which unifies the lambda calculus model and the sequential stored program model. We develop the notion of a guardian as a module that regulates the use of shared resources by scheduling their access, providing protection, and implementing recovery from hardware failures. A shared checking account is an example of the kind of resource that needs a guardian. We introduce the notions of a customer and a transaction manager for a request and illustrate how to use them to implement arbitrary scheduling policies for a guardian. A proof methodology is presented for proving properties of guardians, such as a guarantee of service for all requests received.MIT Artificial Intelligence Laborator