load-balancing, queuing parameters, system parameters, multicomputer systems
This paper presents a performance evaluation approach to compare different distributed load balancing schemes on a unified basis. This approach is an integration of simulation, statistical and analytical models, and takes into account the fundamental system parameters that can possibly affect the performance. We show that all the sender-initiated distributed load balancing strategies can be modeled by a central server open queuing network. Furthermore, these load balancing strategies can be characterized by only two queuing parameters – the average execution queue length and the probability that a newly arrived task is executed locally or migrated to another node. To capture the relation between these queuing parameters and various system parameters, a statistical analysis has been carried out on the empirical data obtained through simulation. The analytical queuing model is then used to predict the response time of a system with any combination of system parameters. Experimental results are obtained for six different load balancing strategies. The proposed model provides performance results in a straightforward manner and can be beneficial to the system designers in assessing the system under varying conditions.
Ahmed, Ishfaq; Ghafoor, Arif; and Mehrotra, Kishan, "Performance Prediction for Distributed Load Balancing on Multicomputer Systems" (1991). L.C. Smith College of Engineering and Computer Science - Former Departments, Centers, Institutes and Projects. Paper 21.
Creative Commons License
This work is licensed under a Creative Commons Attribution 3.0 License.