PET FOCUSED EFFORT PROPOSAL from NPAC/Syracuse University Focused Effort Title: Enhanced Audio-Video Conferencing Support for the TangoInteractive Collaboratory System Thematic Area(s): HPC Training and DoD User Productivity PI Name: David Bernholdt PI EMail Address: bernhold@npac.syr.edu PI Telephone: 315 443 3857 PI Fax: 315 443 1973 Project Description: Current audio/video agent (a.k.a. BuenaVista) of the TANGO Interactive client has been designed to run over today's Internet. This requirement prompted several design and implementation choices. As BuenaVista is multiplatform, all encoding processes are performed strictly in software. Since video encoding is CPU-intensive, the system has built-in limitations of audio and video quality. These limitations affect the maximal video frame rate (<10 fps), video size (QCIF only), and audio sampling rate and computational complexity of the codec. On the other hand, the system has been designed to be real-time, i.e., to have as low latency as possible, consistent with minimal audio quality requirements. As public Internet is beefed up by new trunk lines, and as the MSRCs move towards DREN and Internet2, the minimalistic design requirements of the current BuenaVista may appear too restrictive. We hence propose enhancement of the audio/video agent to support higher quality audio and video. Implementation of such software will keep MSRCs abreast of commercial technology developments and prepared to take advantage of the better quality of service available on the new generation of packet networks. In addition, the new tool, supporting a broader range of qualities of audio and video streams, will be positioned to take advantage of the emerging quality of service support and to support adaptive QoS on future integrated services networks. Currently, audio/video agent supports standard audio and video codecs, such as GSM for audio and H261/H263 for video. The benfit of standard coders is that they are also supported by a growing number of hardware encoders. In particular, H263 hardware encoders are available for both PC and UNIX platforms. Recently, the quality and robustness of the low bitrate hardware video encoders reached the point at which they can be safely deployed in mission critical information systems. Such a deployment is hampared by lack of multi-platform, multiuser, standard compliant videoconferencing software capable of supporting mixed configurations of software and hardwarei-based encoders. By designing current TANGO Interactive audio/video agent we have demonstrated capability and expertise to implement multiplatform videoconferencing solutions of industrial quality. We propose to take the next step towards high-quality audio/video support for synchronous distance learning and real-time videoconferencing by enhancing TANGO audio/video agent to support hardware-based, standard commodity audio and video encoders. Benefits: The enhancements outlined in this proposal will improve the quality of videoconferncing using Tango where appropriate resources (computer systems, network access) are available. This, in turn, will enhance the usefulness and user acceptance of electronic videoconferencing tools for both training and collaboration. Required Resources: Total: $150,000 over 12 months. NPAC will share 50% of costs through its InfoMall (State of New York funded) activity. COST TO NAVO: $75,000 over 12 months Deliverables: The enhancements implemented in this project would result in a desktop videoconferencing system with the following characteristics: - multiplatform system supporting Wintel, SGI, and Sun workstations - support for concurrent use of software and hardware encoders in session of arbitrary topology - for workstations with hardware encoders, support for the video frame rates up to 15 fps and frame size up to SIF (320x244 pixels) - 16 bit audio quality - audio and video synchronized using RTP(Real Time Protocol) mechanism - system latency not exceeding 1 second - manual control of the media streams quality to ensure acceptable performance on a variety of networks - support for both unicast and multicast