Declarative Scheduling in Highly Scalable Systems

Abstract

In modern architectures based on Web Services or Cloud Computing, a very large number of user requests arrive concurrently and has to be scheduled for execution constrained by correctness criteria, service-level agreements etc. The state of the art is to develop hand-coded schedulers, though this tails great costs, long development times, reduced developer productivity and inflexibility of mapping frequently changing requirements. In this paper, we present our approach for a scheduler component that can be programmed using declarative rules. Instead of handling one request at a time, we propose to treat sets of requests as data collections and to employ database query processing techniques to produce high-quality schedules in an efficient manner. Our declarative scheduler will allow for a more flexible and productive way to define existing scheduling protocols, service level agreements and novel application specific consistency protocols. First results presented here are encouraging and motivate for further investigation.