Mode-based resource-usage prediction for runtime resource reservations of video components

摘要

Scalability and reusability of software components enable consumer electronics products to become more open and flexible. Some important features in many of those products are provided via video components. Since these components have highly fluctuating resource usage, dynamic re-allocation of resources is more cost-effective than static allocation. Dynamic resource allocation, however, requires strategies and mechanisms for predicting the requested resources of a component. In this paper, a novel mechanism to detect resource-usage modes for video components is proposed. First, these resource-usage modes are defined and experimentally identified for two example video components. Second, a modedetection algorithm and heuristics for choosing proper parameters for mode detection are introduced. Finally, the run-time properties of a mode-prediction mechanism are evaluated.