Multimedia Networking

Slide Presentation

Overview

In May of 1992, we successfully submitted a proposal for a communications infrastructure for our Science and Technology Center. In our proposal we specified a telecommunications system which would allow video, audio, graphics, and data to be transmitted bidirectionally between Center sites. We installed this system in the Fall of 1992 and have been using it continuously since January 1993.

The system network is based on dedicated T1 lines, leased from MCI which transmit information at a rate of 1.5 Mbits/second. The network is hubbed at the University of North Carolina Chapel Hill. The video and audio conferencing functions of the network are supported by CLI codec units. The data and graphics transmission functions are supported by Cisco packet routers.

We have developed several software applications to facilitate using the network and to enhance its functionality. The VCC is a distributed client- server application with a graphical user interface that allows users at all sites to dynamically control and configure the videoconferencing network. We have also developed a set of video widgets which provide graphical interfaces to video cameras, tape transports, and signal routing equipment.

One of the unique features of our multimedia network is its ability to simultaneously transmit compressed video and digital data over the same low latency, fixed bandwidth link. The links have sufficient bandwidth to remotely control software programs and demonstrations in conjunction with the live video. The ability to transmit data, graphics, and video over our dedicated network will enable us to profile the network traffic; to study methods for appropriately allocating bandwidth; and to develop tools to allow interactive steering of scientific simulations.

This communications environment has proved extremely useful thus far, and has enabled research collaborations and the sharing of unique, site-specific resources such as the NC machining laboratory at Utah and the Pixel Planes parallel display system at North Carolina. This easy to manage environment is also used for our weekly administrative meetings, and has allowed the interactive remote teaching of our multi-site advanced graphics seminar.

Software Projects

Videoconference Control Software (VCC): A suite of software tools has been developed to provide simple but powerful control of videoconferencing functions on our digital multimedia network. The Videoconference Control Software (VCC) gives workstation users both graphical and command-line control of videoconference configuration and scheduling from anywhere within the Center. One application supports on-the-fly control of the network for spontaneous interactions, another provides comprehensive calendar-based conference scheduling, a third monitors and controls the videoconferencing coder-decoders. The system supports up to three simultaneous conferences on the network. Any of the conferences may be encrypted for greater security. Motif-based graphical interfaces provide Center-wide information on the status of the network.

Video Widgets: We’ve developed software tools to support end-user control of video devices on our digital multimedia network. These tools provide close integration of graphics workstation platforms with video production and videoconferencing environments. Standard hardware provides for direct serial control of video signal switching, video tape transports, robotic cameras, and conferencing equipment. The software system provides command-line and graphical user interfaces to the video devices, and incorporates an extensible scripting system for coordinated control of multiple devices. The software is UNIX-based, following a client-server model and uses Motif as the graphical interface toolkit. This standards-compliant implementation allows remote control of any of our video devices from graphics workstations anywhere on the Internet. It also allows groups with similar commercially available hardware to immediately make use of the software tools and applications.

Low-latency Data Network Control: A separate software system is being developed to monitor and control data communications on our digital multimedia network. The system provides Center-wide access to specialized instrumentation, computational resources, and visualization systems at individual Center sites. The system ensures the security and reliability of data connections by allocating network bandwidth on an application-specific basis. The system’s support for the combination of low-latency, guaranteed-bandwidth data transmission and high quality video feedback provides a unique resource for studying a number of important research problems including: time-critical visualization; bandwidth-latency tradeoffs in remote interactive systems; image compression technologies for synthetic images, and collaborative research tools.

Future Research Plans

Our long-range goal is to develop high-quality multimedia networking systems to support interactive visualization and collaboration at-a-distance. This kind of infrastructure must be developed if scientific collaboratories are to become a reality. Multimedia networking is a nascent field, so little is known about what the requirements of such systems will be either at the network or the applications level. We propose to develop prototype systems and study their behavior to develop hardware and software strategies to support the needs of the research collaboratory.

At the network level much work needs to be done on profiling the transmission characteristics of different media (video/audio/graphics/data), to develop appropriate protocols for allocating network bandwidth. Both packet-switched and circuit-switched approaches to multimedia networking need to be evaluated. The different character of local-area and wide-area networks needs to be taken into account in the design of multimedia network systems. Appropriate strategies and protocols also need to be developed for networks of varying bandwidth, from ISDN through Gigabit speeds, to accommodate the different properties of the network substrate, and provide the best service possible at any level and in heterogeneous networks.

At the applications level we need to better understand the computational, storage, and transmission requirements of visualization and collaboration applications. Given the "information explosion" we see in the traditional graphics pipeline, careful thought needs to be given to the issue of where to place the network in the pipeline. The answer to this question is likely to vary for different classes of applications. Support for time-critical multimedia applications is another natural focus for research in this area.

The following is a list of specific research investigations for the next five years.

Multimedia networking on the desktop: We are extending the functionality of our STC network to the workstation desktop using a/v digitizing hardware, high-speed local area networks, and software codecs. We plan to investigate the performance differences between local area and wide area multimedia networks and its impact on the design of interactive visualization at-a-distance applications.

Heterogeneous multimedia networks: We are investigating ways to extend the reach of our STC network to include a larger segment of the graphics community. We have already successfully experimented with interfacing our system to the Internet Multicast Backbone (MBONE). The performance of this hybrid system is limited by the restricted bandwidth of the Internet. We would like to collaborate with developers of the gigabit testbed networks to investigate how to extend the functionality of our system to high-speed networks in anticipation of their widespread deployment. The high-quality visualization applications we are developing in our Center would provide a unique and important test of the capabilities of the testbeds, and one that will be particularly relevant to the requirements of scientific collaboratories.

Profiling multimedia networks: The amount of information transmitted across a network in the course of remotely steering a scientific simulation can be vast. Intelligent use of compression, and appropriate allocation of the modeling, rendering, and interaction components of a simulation to local and remote machines can greatly reduce this network load, but the load statistics of these kinds of visualization applications are not known. We plan to conduct tests to profile the characteristics of different media on local and wide-area networks (video/audio/graphics/data), to develop strategies for allocating network bandwidth to support interactive visualization at-a-distance. We will also look at the relationships between network bandwidth, latency, and variability to determine how to best support the demands of visualization and collaboration applications.

Time-critical vizualization at-a-distance: Regardless of improvements in the technology, visualization applications will continue to outstrip the capacity of processing, storage, and transmission systems. We plan to develop methods to support both guaranteed quality-of-service as well as graceful degradation in time-critical visualization applications. We will develop methods for predicting the moment-to-moment computational demands of a simulation to allow adaptive allocation of resources, and will conduct experiments to develop a perceptual model which will allow us to maximize the visual quality and interactive responsiveness of a simulation at different levels of service.

Rendering to compressed image streams: There is currently a great deal of wasted overhead in rendering synthetic images sequences for dissemination via compressed digital media (MPEG, H.261, etc.). The images are first rendered to bitmaps and then the bitmaps are compressed for transmission. We propose to exploit symmetries in the rendering and compression processes to greatly reduce the computational demands of creating such compressed image streams by completely bypassing the bitmap generation stage. This approach should greatly facilitate generating real-time synthetic image sequences for transmission over multimedia networks. We also plan to exploit the perceptually-based economizing used in compression algorithms to develop a new generation of efficient rendering algorithms.

Technical Reports

 Video Widgets (PDF 185Kb)

 Videoconference Control Software (VCC) (PDF 125Kb)

 Digital Media Network (PDF 95Kb)

Slide Presentation