Optimizing systems by work schedules: (a stochastic approach)

Many systems have very predictable points in time where the usage of a network changes. These systems are usually characterized by shift changes where the manning and functions performed change from shift to shift. We propose a pro-active optimization approach that uses predictable indicators like manning schedules, season, mission, and other foreseeable periodic events to configure distributed object servers. Object-Oriented computing is fast becoming the de-facto standard for software development and distributed object servers are becoming more common as transaction rates increase.Optimal deployment strategies for object servers change due to variations in object servers, client applications, operational missions, hardware modifications, and various other changes to the environment.As distributed object servers become more prevalent, there is more need to optimize the deployment of object servers to best serve the end user's changing needs. A system that automatically generates object server deployment strategies would allow users to take full advantage of their network of computers.The proposed method profiles object servers, client applications, user inputs and network resources. These profiles determine an optimization model that is solved to produce an optimal deployment strategy for the predicted upcoming usage by the users of the system of computers and servers.The validity of the model was tested by experimental measurement. A test bed was created and different manning schedules were simulated. The results of the experimentation showed that the average response time for a user could be improved by altering the deployment of the servers according to the scheduled manning of the system. The model was robust in the sense that the deployments that produced optimal response times in the model also produced optimal or near-optimal response times in the actual implementation of the test-bed.

[1]  Ian T. Foster,et al.  The Anatomy of the Grid: Enabling Scalable Virtual Organizations , 2001, Int. J. High Perform. Comput. Appl..

[2]  William J. Ray Optimization of distributed, object-oriented systems , 2000, OOPSLA '00.

[3]  William J. Ray,et al.  Object model driven code generation for the enterprise , 2001, Proceedings 12th International Workshop on Rapid System Prototyping. RSP 2001.

[4]  Benjamin W. Wah,et al.  Synthetic workload generation for load-balancing experiments , 1995, IEEE Parallel & Distributed Technology: Systems & Applications.

[5]  Heejo Lee,et al.  Replicated Process Allocation for Load Distribution in Fault-Tolerant Multicomputers , 1997, IEEE Trans. Computers.

[6]  Luqi,et al.  Software engineering with abstractions , 1991 .

[7]  David C. Luckham,et al.  An Event-Based Architecture Definition Language , 1995, IEEE Trans. Software Eng..

[8]  Ian T. Foster,et al.  End-to-end quality of service for high-end applications , 2004, Comput. Commun..

[9]  Chung-Ta King,et al.  The thread-based protocol engines for CC-NUMA multiprocessors , 2000, Proceedings 2000 International Conference on Parallel Processing.

[10]  Richard M. Adler,et al.  Distributed Coordination Models for Client/Server Computing , 1995, Computer.

[11]  David C. Luckham,et al.  Complex Event Processing in Distributed Systems , 1998 .

[12]  Wentong Cai,et al.  How network topology affects dynamic loading balancing , 1996, IEEE Parallel Distributed Technol. Syst. Appl..

[13]  Walter Mann,et al.  Event Mining with Event Processing Networks , 1999, PAKDD.

[14]  Donald F. Towsley,et al.  Bounding the Mean Response Time of the Minimum Expected Delay Routing Policy: An Algorithmic Approach , 1995, IEEE Trans. Computers.