OPTIMAL POLICIES FOR MULTI-LEVEL ADAPTIVE DISTRIBUTED COMPUTER SYSTEMS

Modern distributed applications, such as distributed multi-media and mobile applications, face unpredictable operating conditions and load variations. Performance cannot be designed into such applications in advance; they have to be able to tune themselves into unexpected environments and to adapt to changes over time. We see many examples of single adaptations in applications and middleware, but the opportunities are even greater if many features of the system, at all levels, are adaptive. This paper proposes an architecture to support coordinated adaptive changes in all levels (application, middleware and operating system), with an optimal controller at its core. The controller uses optimal policies based on Markov Decision Processes (MDP) which seek to s atisfy a set of system quality-of-service and resource-usage goals.

[1]  Keith Cheverst,et al.  Developing a context-aware electronic tourist guide: some issues and experiences , 2000, CHI.

[2]  Jonathan M. Smith,et al.  Protocol boosters , 1998, IEEE J. Sel. Areas Commun..

[3]  Satoshi Matsuoka,et al.  Implementing Parallel Language Constructs Using a Re ective Object-Oriented Language , 1998 .

[4]  Jonathan M. Smith,et al.  Operating System Support for Protocol Boosters , 1996 .

[5]  Gordon S. Blair,et al.  Configuring and Reconfiguring Resources in Middleware , 2006 .

[6]  Ramana Rao,et al.  Implementational Reflection in Silica , 1991, ECOOP.

[7]  Henk Tijms,et al.  Stochastic modelling and analysis: a computational approach , 1986 .

[8]  Jim Dowling,et al.  Using Reflection to Support Dynamic Adaptation of System Software: A Case Study Driven Evaluation , 1999, Reflection and Software Engineering.

[9]  Mario Tokoro,et al.  SCONE: using concurrent objects for low-level operating system programming , 1995, OOPSLA.

[10]  Rodger Lea,et al.  DART: A Reflective Middleware for Adaptive Applications , 1998 .

[11]  Gordon S. Blair,et al.  Developing Adaptive Applications: The MOST Experience , 1999, Integr. Comput. Aided Eng..

[12]  Gordon S. Blair,et al.  L2imbo: A distributed systems platform for mobile computing , 1998, Mob. Networks Appl..

[13]  Aruna Seneviratne,et al.  DARTS-a dynamically adaptable transport service suitable for high speed networks , 1993, [1993] Proceedings The 2nd International Symposium on High Performance Distributed Computing.

[14]  Pattie Maes.  Concepts and experiments in computational reflection , 1987, OOPSLA 1987.

[15]  Luca Benini,et al.  An adaptive algorithm for low-power streaming multimedia processing , 2001, Proceedings Design, Automation and Test in Europe. Conference and Exhibition 2001.

[16]  Krzysztof Czarnecki,et al.  Generative Programming , 2001, ECOOP Workshops.

[17]  Klara Nahrstedt,et al.  Adaptive middleware architecture for a distributed omnidirectional visual tracking system , 1999, Electronic Imaging.

[18]  Wouter Joosen,et al.  Customization of Object Request Brokers by Application Specific Policies , 2000, Middleware.

[19]  Riccardo Bettati,et al.  Imprecise computations , 1994, Proc. IEEE.

[20]  Fabio Kon,et al.  Monitoring, Security, and Dynamic Configuration with the dynamicTAO Reflective ORB , 2000, Middleware.

[21]  Ray Jain,et al.  The art of computer systems performance analysis - techniques for experimental design, measurement, simulation, and modeling , 1991, Wiley professional computing.

[22]  Gregor von Laszewski,et al.  A fault detection service for wide area distributed computations , 2004, Cluster Computing.

[23]  MaesPattie.  Concepts and experiments in computational reflection , 1987 .