A testbed for experiments with concurrency control primitives in persistent systems

In current research operating systems, support for objects and the persistence of these objects has been recognised as an important consideration. Research systems provide support for these concepts in the kernel. In these systems a great deal of effort has gone into integrating persistence and object structures with the other low level functions of the operating system. The low level nature of the support ensures that these concepts are in every respect a seamless part of the operating system which interacts sensibly with other components of the system. There has been a tendency in such systems to ignore the issues of concurrency control and resilience, with the attitude that these concepts are well understood by the database community, and as such, can be easily added on top of the system at a later date. A research project is described which aims to investigate the implementation of concurrency control and resilience support at the lowest level of the operating system. Concurrency control and resilience are considered to be sufficiently important and complex issues that they merit support at a low layer of the operating system in the same sense that support for persistence belongs at this level. It is a motivating belief for this project that support for fully fledged concurrency control systems in a truly seamless fashion requires appropriate low level functionality provided by the system. The goal of this project is the identification a set of low level operating system primitives through which a variety of both traditional and more advanced concurrency control methods can be implemented on a persistent store. This goal is being approached through the development of a prototype persistent store which can be used as both an experimental environment and a testbed for the demonstration of the viability of the projects results.<
>