Autonomous decentralized resource allocation for tracking dynamic load change

Abstract

System architecture is proposed to adapt computing resource to load changes of the service requests from the Internet. The architecture employs an autonomous decentralized computing scheme for resource allocation. A subsystem has a production server to process service requests and a coordination server to adapt computing resource to load changes. Coordination is executed among subsystems to calculate how many resources are needed. Three system architectures, centralized resource allocation, autonomous decentralized resource allocation with independent adaptation and autonomous decentralized resource allocation with coordinated adaptation, are compared for tracking ability through simulation. Autonomous decentralized resource allocation with coordinated adaptation shows a good tracking ability comparable to the centralized resource allocation that should be nearly optimal. Since autonomous decentralized architecture shows better performance for availability than centralized architecture, autonomous decentralized allocation achieves the best performance for overall criteria.