An adaptive quality of service aware middleware for replicated services

S. Krishnamurthy, W.H. Sanders, M. Cukier
2003 IEEE Transactions on Parallel and Distributed Systems  
A dependable middleware should be able to adaptively share the distributed resources it manages in order to meet diverse application requirements, even when the quality of service is degraded due to uncertain variations in load and unanticipated faults. We have addressed this issue in the context of a dependable middleware that adaptively manages replicated servers to deliver a timely and consistent response to time-sensitive client applications. These applications have specific temporal and
more » ... sistency requirements, and can tolerate a certain degree of relaxed consistency in exchange for better response time. We propose a flexible QoS model that allows clients to specify their temporal and consistency constraints. Since unanticipated faults and transient overloads cause uncertainty in most distributed environments, it is hard to provide deterministic guarantees for meeting the timeliness requirements of applications. Hence, our approach provides probabilistic timeliness guarantees. The approach dynamically selects replicas to service a client's request based on the prediction made by probabilistic models. These models use the performance history collected by monitoring the replicas at runtime to predict the ability of a replica to meet a client's QoS specification. In this thesis we describe the adaptive framework and present the probabilistic models we have developed for the cases in which the replicated content is either static or dynamic. We discuss experimental results that validate the models, and compare the timeliness/consistency tradeoffs under different scenarios. The experimental results demonstrate the role of feedback and the efficacy of simple analytical models for adaptively sharing the available replicas among the users. Given the increasing demand for different services and the diversity of the requirements of clients, such adaptive frameworks that rely on feedback-based resource allocation are likely to play an increasing role in a range of problems related to building dependable systems. iii To my parents iv
doi:10.1109/tpds.2003.1247672 fatcat:4djwvsjkifax3o7faov23lggfq