Coordinating Power-Aware Adaptation of Applications and Hardware Resources

[Motivation] [Approach] [Publications]


With the growth of computation and communication capability, mobile devices, such as laptops and PDAs, are becoming increasingly important platforms for ubiquitous computing. Such mobile systems present two major design concerns. To support high resource demand from a dynamic application workload, they need powerful processors with a high energy consumption. For the sake of mobility, however, they require low energy consumption to maximize the battery lifetime.

Multimedia applications can usually adapt their behavior to trade off their QoS requirements for resource demands. Moreover, modern mobile devices are also able to switch among multiple operating modes with tradeoff between performance and energy consumption. For example, some modern processors, such as Mobile Intel Pentium III, Crusoe TM5600, and AMD Athlon , support multiple processor speeds with tradeoff between performance and energy consumption. The adaptbility of applications and hardware enable potential energy saving for mobile systems. Our research focuses on how to coordinate the adaptation of both multimedia applications and hardware resource under energy constraints.


1. System Architecture

The system architecture consists of multimedia applications, hardware resource, and middleware between them, as shown in Figure 1.

Figure 1. System Architecture

Hardware Resource Profile

The hardware resource profile describes the multiple operating points of resource, each with a separate performance and energy consumption. System power is essentially proportional to the processor speed. The battery lifetime is inverse proportional to system power, and can be calculated according to the current energy and system power of an operating point, as shown in Figure 2.

Figure 2. Processor Operating Space

Application QoS Profile

The application QoS profile describes the multiple QoS points of a multimedia application, each with a separate QoS requirement and resource cost. All possible QoS points together form the QoS space of the multimedia application. For example, a video on demand (VOD) application provides QoS requirements with dimensions of frame rate, frame size, and color depth, as shown in Figure 3.

Figure 3. QoS Space of a VOD Application

User QoS Profile

From the user's point of view, different QoS dimensions may have different importance, for example, frame rate is more important than color depth in a VOD application. The user QoS profile describes the order of QoS points with user preference, and the QoS point order determines the adaptation order of application.

2. Coordinating Adaptation Algorithm

We are currently investigating different adaptation policies to minimize energy consumption or maximize the application utility.

3. Resource Reservation in dynamic speed Context

The dynamic speed setting incurs some challenging to resource reservation of multimedia applications, because most proposed reservation algorithms assume a constant processor speed. The challenging in a DVS context includes, for example, how to make admission control and adjust reservations when the processor speed changes.

We are seeking mechanisms to ensure the correctness of reservation admission and enforcement, thereby delivering resource guarantees to applications, in a variable-speed context.


  • W. Yuan and K. Nahrstedt, " A Middleware Framework Coordinating Processor/Power Resource Management for Multimedia Applications," Proc. of IEEE Globecom 2001, San Antonio, Texas, November 2001, 1984-1988.
  • W. Yuan, K. Nahrstedt and X. Gu, " Coordinating Energy-Aware Adaptation of Multimedia Applications and Hardware Resource," Proc. of the 9th ACM Multimedia (Multimedia Middleware Workshop), Ottawa, Canada, October 2001, 60-63.

    Please direct comments and suggestions to Wanghong Yuan
    Last updated: Nov 30, 2001.

    Copyright © 2001, Multimedia Operating Systems and Networking Group,
    University of Illinois at Urbana-Champaign.