Dynamic tuning of feature set in highly variant interactive applications

Abstract

For important classes of interactive consumer applications, such as gaming and video, the Quality-of-Service requirement is to create a maximally immersive experience for the interactive user. This necessitates a trade-off between maximizing the computational complexity of application features versus the need to maintain a smooth and sufficiently high frame-rate. The implementation of these applications using conventional C/C++/Java development flows, their highly data-dependent time-varying nature, and the lack of analytical models for their execution time behavior pose unique challenges in obtaining significant QoS improvements. In this paper, we propose an adaptive feedback controller that dynamically tunes the application feature set in the face of the challenges outlined above. We use a system-identification strategy where the controller estimates an application's execution characteristics based on i) a limited amount of domain knowledge common to video and gaming, and ii) the observed response of the application to control inputs. Therefore, the proposed controller is suitable for a range of interactive applications without needing application-specific knowledge. We use a commercial game engine and the MPEG2 encoder as representative real-world applications to show that our strategy offers a simple practical solution to achieve substantial improvements in QoS across a wide range of operating conditions