The Design and Implementation of Job Management System Based on Feedback Control

Abstract

We design and implement a Job Management System for improving automation level of job execution process in network-based digital multimedia content production. Jobs are automatically parallelized and executed on a large number of execution machines. Our implementation of the Job Management System has two novel aspects. One aspect is using a feedback mechanism in execution time estimation base on Extreme Learning Machine (ELM) for self-optimizing. Based on the feedback mechanism, we can continually use the newly obtained data to improve the system. The other one is having the capability of autonomic fault recovery. The system has been integrated to a network-based digital multimedia content production platform.