Autonomous Decentralized Load Tracking Systems And Evaluation Criteria For Response and Stability

This paper proposes autonomous decentralized load tracking system to adapt computing resource to unpredictably fluctuating load. Autonomous measurement technology is proposed to achieve better response time, by communicating load difference among subsystems and normalized integral to estimate total load using limited information gathered within a measurement period. Autonomous controlled decision technology is proposed. Characteristics for the system to maintain are clarified and a decision mechanism by binary functions with uniformly distributed thresholds is proposed. Evaluation criteria for response and stability are proposed to measure response satisfaction for end users and stability satisfaction for system administrators. A trade-off relationship is shown between the two criteria. Comparisons with existent technologies are done using the criteria