|
Application Level Adaptation and Control for Retrieval and Delivery of Continuous Media over the Internet |
||
|
Richard R. Muntz (PI) Computer Science Department, UCLA.
Don Towsley and Jim Kurose (Co-PIs)
Leana Golubchik (Co-PI)
Brazilian Collaborators (this is a joint program between NSF and CNPq):
Other international collaborators: |
||
| Contact Information | ||
|
Leana Golubchik Computer Science Department University of Southern California 941 West 37th Place, SAL 312 Los Angeles, CA 90089-0781 Voice: (213) 740-4524 FAX: (213) 740-7285 Email: leana@cs.usc.edu URL: http://cs.usc.edu/~leana/ |
||
| WWW Page | ||
| http://bourbon.usc.edu/iml/projects/comit/ | ||
| Project Award Information | ||
|
||
| Keywords | ||
| Internet-based applications, application-layer adaptation and control, multimedia applications. | ||
| Project Summary | ||
|
Future networked multimedia information systems will carry a wide variety of
applications including digital libraries, video, audio and image services,
distance learning and collaboration, networked virtual environments,
and entertainment.
The main characteristics of multimedia applications that lead to
difficulties in end-to-end systems design
are that they have very large bandwidth and storage requirements,
with vastly different performance and reliability requirements,
often coupled with real-time constraints.
Along with these characteristics, the highly interactive nature of a
Along with these characteristics, the highly interactive nature of a
variety of multimedia applications, resulting
in fairly unpredictable workloads, makes the design and evaluation
of networked multimedia information systems an exceptionally challenging
problem.
The work in this project is focused on three aspects of this problem:
Another important feature of this project is that the participants have available a number of prototypes of state of the art multimedia systems. These applications are used to motivate the development of new algorithms for retrieving information and for maintaining the desired quality of service after sending the data over a wide area network. They also form the basis of the many experimental and analytical studies that will be performed to evaluate these new algorithms. To aid in this evaluation, our group also developed performance evaluation and modeing tools. Thus, the environment of our labs as well the long distance among them will provide a unique testbed for this type of an evaluation due to the drastically different connectivities available to our applications, from gigabit low utilized links to intercontinental congested links. This research represents an important step in the design and performance evaluation of next generation information servers and the networked applications that will operate on top of them. As a result of our research we expect to have a better understanding of how storage server resource management policies, channel allocation policies, and network adaptation policies interact to satisfy the required QoS of CM applications, despite the fairly unpredictable network delays. |
||
| Publications and Products | ||
The following are project articles.
|
||
| Project Impact | ||
|
||
| Goals, Objectives, and Targeted Activities | ||
|
The research in this project addresses a variety of
inter-related issues that arise in the design of next generation
information systems, with a focus on providing
the specified quality of service to the end users.
We investigate a broad range of problems from issues on how to
efficiently store and retrieve CM information in large systems,
to issues on how to efficiently transmit the retrieved information
via the Internet.
Although broad in scope, the problems we investigate are tightly related. Our research is divided into three main thrust areas: (1) efficient storage and retrieval, (2) efficient transmission, and (3) performance evaluation Efficient storage and retrieval. Data placement and scheduling techniques are key to sustaining the necessary data rates of CM applications. We investigate data placement strategies, which are studied not only in the context of dedicated servers for a particular multimedia application but also under the assumption that different applications can run on a single server as well as under the assumption of distribution of servers across wide-area networks. Efficient transmission. We are developing and evaluating application-level adaptation mechanisms to enhance the quality of services delivered by the network. This includes mechanisms for coping with packet losses and high delay variability imposed by the network which have significant effects on the QoS experienced by clients viewing CM applications. Currently, our primary interest is in delivery of live voice (with an experimental setup between UMD and a couple of sites in Brazil, UFRJ and UFMG) and pre-recorded video (with an experimental setup between USC, UMD, UCLA, Purdue, a site in Hong Kong, CUHK, and a site in Italy, Univeristy of Torino) over wide-area networks (and specifically, over long-haul highly lossy international links). We focus on the development of application layer techniques for improving the perceived QoS when real time continuous media is delivered over wide-area networks, through the use of multiple paths. The goal is to improve performance metrics such as transfer delay, throughput and loss probability, for instance by enhancing the efficiency of FEC schemes and reducing packet loss correlations, and by dynamically adapting to network congestion characteristics, for long lasting transfers. Performance Analysis. Performance analysis plays an important role in evaluating the design alternatives outlined above. We intend not only to use the tools we develop to aid in the performance evaluation tasks but also to develop new analysis techniques, as needed, in order to evaluate the performance characteristics of algorithms we will investigate. Our goal here is to be able to solve large-scale complex models that will undoubtedly arise in the course of this project. A key issue in building accurate models for our investigation is to perform a careful characterization of the workload generated by the applications in our laboratories, and then develop workload models (based on these characterizations) which can be used in the performance evaluation of server architectures. Furthermore, developing models for understanding the end-to-end network behavior is essential for evaluating the application adaptation mechanisms. The environment of our labs as well the long distance among them will provide a unique testbed for this type of an evaluation due to the drastically different connectivities available to our applications, from gigabit low utilized links to intercontinental congested links. |
||
| Area Background | ||
|
A common thread
among the applications considered in this project is the so-called
``continuous'' nature of their data. In continuous media (CM),
strict timing relationships exist
which define the schedule by which CM data must be rendered
(e.g., a video displayed, 3D graphics rendered, or audio played out).
These timing relationships coupled with the high aggregate bandwidth needs,
high individual application bandwidth needs, and high storage requirements
(terabytes) pose significant challenges to the design of such systems.
These systems will consist of large-scale servers servicing large numbers of clients over public wide area networks. They pose yet another challenge because, although the applications require a minimum quality of service (QoS), they will run over networks which for many years to come will be unable to guarantee this. Thus, it is important not only to design servers that can provide minimal QoS to applications in a scalable manner, but also to design application-level mechanisms that will permit the application to adapt to variations in the quality of the intermediate network connection. |
||
| Area References | ||
|
||