Mobile Computing - Optimization For Multimedia Workload

ADAPTIVE ALGORITHMIC POWER

OPTIMIZATION FOR MULTIMEDIA WORKLOAD IN MOBILE ENVIRONMENTS

Pragnesh Goyani

Software Engineer

Motorola Inc., Plantation, Florida

Hitesh Joshi

Software Engineer

eFunds Inc., Sunrise, Florida

ABSTRACT

This paper addresses the problem of power consumption in mobile devices with multimedia and presents algorithmic optimization techniques to achieve reduction of power usage. We also present researched approaches for adaptive minimization of the total energy consumption in multimedia wireless communications subject to achieving a given quality of service. We discuss the energy optimization techniques, and collaborative and non-collaborative techniques used for power management, after which we discuss in detail, the adaptive encoding and decoding algorithms as well as RTOS based and Server assisted techniques.

This paper discusses the following five algorithms and their overheads:

Algorithm #1: Window Follower Algorithm for Encoding (Motion Estimation)

Algorithm #2: Transmission Adaptation Encoding Algorithm

Algorithm #3: MPEG4 Adaptive Parameters Encoding Algorithm

Algorithm #4: Dynamic Voltage Scaling вЂ" Adaptive Clock/Voltage Setting

Algorithm #5: Buffer Insertion Decoding Algorithm

Algorithm #6: Adaptive Multimedia Streaming Algorithm

This paper also presents several simulation results achieved by other authors, which show the effectiveness of the discussed algorithms in reduction of energy or power consumption.

KEYWORDS

Low-power, multimedia streaming, adaptive encoding, adaptive decoding, voltage scaling, frequency scaling

MOTIVATION

Unlike other enabling technologies for mobile information systems, the specific energy of commercially available rechargeable batteries has improved at only about 2 percent per year over the past half century [9]. Considering this track record of available solutions on energy optimization, wireless multimedia systems has to be optimized for low energy consumption subject to a desired quality of service. These solutions have to be adaptive in nature, considering the complexity of the multimedia workload and modest resources the mobile systems posses.

INTRODUCTION

A key challenge for mobile devices today is the management of power considering the variable nature of workload, the heterogeneity of multimedia content and the inconsistent wireless network quality. Energy consumption of the mobile devices is not only dominated by wireless communication, but data processing or computation takes heavy toll on power as well. Since the conception of multimedia in mobile phones, high computational power as well as communicational power has become an important requirement. Videos (MPEG), Audio (MP3) and Photos (JPEG) on cell phones are becoming more common now, and new technologies like MPEG4/H263 for multimedia are being used for multimedia processing. Current high-end mobile devices integrate wireless wideband data modems, video cameras, net browsers, and phones into small packages. Being of modest sizes and weights, these devices have inadequate resources - lower processing power and limited battery life. Multimedia applications used on these devices have distinctive Quality of Service (QoS) and processing requirements which make them very resource hungry [5].

Energy usage of a 3G phone in Video Streaming mode is provided in Table 1 below.

System Component

Energy consumption (mW)

RF Receiver and Cellular modem 1200 mW

Application Processor and Memories 600 mW

Memories 200 mW

User Interface (audio, display, keyboard, with backlights) 1000 mW

Total 3000 mW

Table 1: Energy consumption for different system components

This usage can only increase with the increasing demand of higher resolution videos. Among its limitations, although battery power has significant importance, it is probably the least researched and has experienced the least improvement.

Research has been done in both hardware and software areas where power consumption can be reduced. As far as the physical parameters are concerned, there are many limitations for mobile wireless communications, including high and variable error rate, available bandwidth variation and limitation and limited battery power [1]. [1] proposes a scheduling algorithm that decides which flow should be dealt with at what time, enabling the wireless network interface to sleep for longer periods, hence saving battery power. Lan et al [9] investigates low power video transmission techniques based on H263 video coding standards, which are majorly based on software knobs.

Several interesting solutions have been proposed at various computational levels вЂ" system cache and external memory access optimization, dynamic voltage scaling (DVS) [10, 11], dynamic power management of disks and network interfaces, efficient compilers and application or middleware based adaptations [12, 13]. Adaptive algorithmic power optimization is another solution that can maintain the QoS in spite of reducing power usage.

Encoding algorithms perform data compression before transmission and reduce information’s bandwidth requirement by reducing the bit rate. On the other hand, decoding algorithms help to playback multimedia information (saved on the mobile unit or streaming from server) in a compressed format. Power efficient encoding/decoding algorithms can help overcome the battery power limitations of the mobile unit, and when these algorithms are made adaptive to workload and physical parameters, better battery performance is achieved.

NON-COLLABORATIVE