Other Free Encyclopedias » Online Encyclopedia » Encyclopedia - Featured Articles » Contributed Topics from A-E

Data Management Techniques for Continuous Media in AD-HOC Networks of Wireless Devices - Introduction, Overview of CHaMeLeoN, Application Requirements, Design Objectives

communication display time content

Shahram Ghandeharizadeh, Ahmed Helmy, Bhaskar Krishnamachari,
Francois Bar, and Todd Richmond
University of Southern California, Los Angeles, USA

Definition: This article introduces CHaMeLeoN as a large scale software effort to facilitate exchange of both traditional (text, image) and continuous media (audio and video clips).

CHaMeLeoN is designed to support both delayed and immediate modes of communication. It is targeted toward mobile devices configured with a heterogeneous mix of wired and wireless components. Our primary focus is on the limited radio-range of wireless connections (such as 802.11) and exchange of data among those devices that are in close geographical vicinity of one another to participate in an ad-hoc network. We outline the objectives of CHaMeLeoN, and in short articles summarize our research findings to date.

Introduction

Rapid advances in wireless technology, computational, mass storage, and sensor devices are realizing new paradigms of communication in wireless mobile environments. They offer a compelling vision of the future Internet as a seamlessly integrated heterogeneous system. Continuous media (video and audio) services are of particular interest because they are necessary for many classes of important and interesting applications, ranging from entertainment to scientific collaboration. These applications can be broadly categorized into those that manage content pertaining to either immediate, delayed, or both modes of communication. Delayed communication is when a user records and stores data in anticipation of its future retrieval, e.g., display of a recordings from a recent meeting or seminar. Immediate communication is when a user generates data for display at a client in real-time, e.g., scientists communicating and collaborating across different laboratories. The boundary between these two modes of communication is defined by the delay from when data is produced to the time it is displayed. As this delay increases, immediate mode of communication becomes delayed. A good example is when the collaborating scientists record their real-time communication for future display and analysis.

A number of visionary systems such as Memex and MyLifeBits depend on devices that capture and retrieve data pertaining to both communication modes. The computer and communication industry has started to develop such devices. For example, today’s cellular phones facilitate phone conversations between multiple parties, capture photos and record videos of their surroundings, and transmit the recorded information to other cellular phones and devices with Internet Protocol (IP) connectivity. Similarly, cable set-top boxes, termed Digital Video Recorders (DVRs), support telephone conversations (Voice over IP, VoIP), provide time-shifted viewing of broadcasted programs, and store personal audio and still image libraries. (TiVo terms these additional features “home media” in its Series2 offerings.) It is not far fetched to envision DVR extensions to store personal video libraries along with meta-data associated with the different data items. It might be extended with either one or multiple wireless networking cards to facilitate (1) an in-home network for uploading of content from devices that capture data such as a camcorder, and (2) an ad-hoc network consisting of other DVRs in the neighboring households. Such a device can be implemented using today’s wireless technology such as 802.11a/g which offers limited radio-range with bandwidths in the order of tens of Mbps. We term such a device a Home-to-Home Online (H2O) device.

Current technological trends point towards multi-purpose Personal Digital Assistants (PDAs) that are small, inexpensive, weigh less than a pound, and have long lasting battery lives. To facilitate efficient retrieval of data, these PDAs might be equipped with sensors that capture a variety of metadata about their surroundings. An example sensor might be a global positioning system to record the geographical location of different recordings and communications. This would enable a user to retrieve all those photos captured from a specific location, e.g., India. This PDA will communicate with diverse networks. For example, it might interface with a vehicle’s multimedia network to enable a passenger to share his or her personal recordings with other passengers using the vehicle’s fold-down screen. The same PDA will communicate with any H2O device for sharing of photos and video clips. Mobility of PDAs and their potential intermittent network connectivity differentiates them from H2O devices.

This article provides an overview of the ChaMeLeoN framework and its design decisions. CHaMeLeoN is designed to support both delayed and immediate mode of communication in a wireless, heterogeneous network of multihop, mobile devices. This framework touches upon a wide range of issues such as security and privacy, economic fairness, user interfaces, strategies to satisfy the requirements of diverse applications sharing the same infrastructure, and many others. A study of all topics in depth is beyond the scope of this paper. Instead, we focus on the core infrastructure of the network, namely, techniques to deliver continuous media in a manner that meets the application requirements and challenges of a mobile wireless environment.

Overview of CHaMeLeoN

A unique feature of CHaMeLeoN is its awareness of (and adaptation to) the application requirements and environmental constraints. These are shown as the top and bottom layers of Figure 1. CHaMeLeoN, the mid layer, consumes input from these two layers: First, communication mode of the application (immediate, delayed, or both) and its requirements (QoS, efficiency, and availability of data; see below for details). Second, environment and network characterization in the form of: (a) mobility models -including synthetic or trace based mobility models, (b) traffic patterns and information association (e.g., due to users’ social behavior or correlation of sensed data), © constraints of limited power, computation or communication capabilities, and (d) model of the wireless channel. The core of the framework consists of parameterized adaptive algorithms and protocols for data placement, scheduling and merging, admission control, data and resource discovery, data delivery and routing protocols. In addition, protocol design for mobility is a component that may span multiple other core components to include mobility prediction, mobility-resilient and mobility-assisted protocols. For example, data placement may be based on mobility prediction, and resource discovery may be mobility-assisted.

Application Requirements

CHaMeLeoN characterizes an application’s requirement using Quality of Service (QoS), efficiency, and data availability. We describe these in turn.

QoS is a qualitative requirement, characterizing the perception of end users. It can be quantified using different metrics. One is the delay observed from when a user references content to the onset of its display, termed startup latency. A second might be the frequency and duration of disruptions observed when displaying audio and video clips, termed hiccups. Audio and video clips are continuous media, consisting of a sequence of quanta, either audio samples or video frames, which convey meaning when presented at a pre-specified rate. Once the display is initiated, if the data is delivered below this rate (without special precautions) then the display might suffer from hiccups. Third, while the displayed data might be hiccup-free, it might have been encoded with significant loss. Scalable compression techniques control the quality of streamed data based on the available amount of network bandwidth.

Efficiency describes how intelligently resources are utilized when streaming content. A key metric is how many devices may display their referenced content simultaneously. With a network of N devices, ideally all N devices should be able to have an active display. Typically, the available wireless network bandwidth between devices must be managed intelligently to accomplish this objective.

Availability of data quantifies what fraction of accessible content is available to a client at a given instance in time. It is impacted by both the discovery and retrieval of content. The system must be able to deliver discovered content to the requesting device in order for that content to be advertised as available. As detailed below, streaming of data between a producing and a consuming device requires reservation of bandwidth. This may exhaust certain paths in the network, isolating a candidate client device from those devices that contain relevant data. This data remains unavailable until the reserved paths are released. Another important factor is the characteristics of the wireless network. Both the bandwidth and loss rate of an 802.11 connection between two devices is dependent on its deployed environment and might vary from one minute to the next due to factors such as exposed node. Finally, mobility of devices poses both challenges and opportunities. It is a challenge when it renders relevant content un-deliverable or leads to degradation in quality of service. It is an opportunity when it provides relevant content to a potential consumer directly instead of a multihop transmission.

Design Objectives

Techniques that constitute CHaMeLeoN are designed to be adaptable, facilitate physical data independence, and support streaming of continuous media. Below, we detail these in turn.

By adaptable, we mean employed techniques must support diverse applications with different and potentially conflicting requirements. We categorize these requirements along three dimensions consisting of Quality of Service (QoS), efficiency, and data availability; see below for details. One application may demand the highest QoS while another may emphasize availability of data and services. Proposed techniques are parameterized whose settings navigate the three dimensional optimization space. The challenge here is to identify the parameters of a technique, quantify how their settings navigate the optimization space, and design of algorithms to navigate this space given the physical realities of an environment.

Physical data independence means the organization of data can be modified without causing application programs to be rewritten. Physical organization of data might be manipulated either at its storage or delivery time. To illustrate, CHaMeLeoN must manage physical organization of clips when storing them. In certain environments, participating devices may collaborate by contributing some of their storage to a common pool in order to enhance QoS and efficiency metrics. With a video clip, the data placement component of CHaMeLeoN may aggressively replicate the first few blocks of a clip across many devices to enable a client to display the clip immediately while it discovers and retrieves the remaining blocks of the clip. An application may not want to use this placement strategy because its participating devices are expected to be disconnected from the network for long durations of time. This should not mean that one re-writes the data placement component. Instead, either (a) this component’s parameter settings must enable a system designer to deploy it in a manner that meets this application’s requirement or (b) metadata associated with a clip provides sufficient context for the data placement strategy to meet this application’s requirement. While both approaches realize physical data independence, the first is at system deployment time while the second is at run-time using meta-data provided either by an application or its clips. The challenge here is design, implementation, and testing of flexible software.

Streaming of continuous media facilitates overlapped display of a clip at a client with its delivery from a network of one or several servers. This is essential for immediate mode of communication because participants require real-time data exchange. Even with delayed mode of communication, streaming is useful because it minimizes the amount of time a user waits for the system to initiate display of a clip. To illustrate, a 30 minute DVD quality video is typically 900 Megabytes in size. If the system downloads the clip prior to initiating its display then the client must wait for the duration of time to download 900 Megabytes of data. Even if the network provides a bandwidth of 100 Megabits per second, the client must wait for more than a minute. With streaming, the display starts once sufficient data is prefetched at the client to hide the time required to deliver the remainder of a clip. A challenge here is how to determine the size of prefetch portion. This is a challenge because network bandwidth might fluctuate dramatically during delivery of a clip, causing one or more clients to starve for data. This causes the display to suffer from frequent disruptions and delays, termed hiccups. A related challenge is how to recover from these scenarios. The obvious solution is to free up bandwidth by terminating one or more clients. The key question is how to determine the identity of victim clients in a decentralized manner while respecting the application’s overall requirements.

Data Modeling, Multimedia [next] [back] Data Encryption Standard (DES) and Advanced Encryption Standard (AES)

User Comments

Your email address will be altered so spam harvesting bots can't read it easily.
Hide my email completely instead?

Cancel or