Distributed locking: a mechanism for constructing highly available objects

A description is given of the results of a study of methods of achieving fault tolerance in the Clouds system and, in particular, of achieving increased availability of objects. The problems explored in this work, the model of distributed computation in which the problems posed by the research were examined (the Clouds system), the tools that were used to address these problems (the Aeolus programming language), and some related research are briefly described. The authors present a methodology for achieving available services by conversion of resilient single-site implementations into replicated implementations. A mechanism with which they propose to support this methodology, called distributed locking (DL), is presented. A description is also given of a linguistic feature for the specification of the availability properties of an object replicated via DL. The language runtime support features (primitives) required for DL and the operating system support needed for these features are presented