Indiana University is continuing its development of the VARIATIONS project, addressing specifically client software, networking topologies appropriate to multimedia information delivery systems, and server technologies that will support this information system. The Indiana University Libraries recognizes the Music Library's VARIATIONS project as consistent with its strategic direction and as having implications for libraries on the Bloomington campus. Indiana University Computing Services recognizes that the VARIATIONS project is a testbed for the manner in which the campus network and distributed computing will evolve on the Bloomington campus.
The project will deliver digital audio and full-motion video across building, campus (Indiana University--Bloomington) and inter-campus (including Indiana University Purdue Universty at Indianapolis) networks using emerging ATM technology and using switched ethernet technology. As a library-oriented project, it will result in the digitization of information objects associated with music. Music as a discipline represents an ideal setting to understand these issues and to deploy these services. Music's primary information format is sound, and to a lesser extent, video. Music also uses traditional textual and graphic data. Furthermore, the sound quality involved, particularly associated with an internationally known School of Music, is of primary importance. The VARIATIONS project also has several years of experience in working with these issues. For the reasons briefly outlined in this paragraph, the VARIATIONS project represents an ideal setting in which to explore one of the central issues confronting technology and libraries: the distribution of continuous data as an information service across wide-area networks.
The current focus is the delivery of continuous, compact-disc quality digital audio over a network without audio break-up. Future goals include the distribution of score notation and the development of score databases. Digitizing multimedia information allows each student to control, independently and interactively, multimedia sources. The state of the analog source at the time of conversion will be preserved in digital form. Analog sources, such as long-playing records, cassette tapes, and videocassettes once converted into digital form will not deteriorate with use. The VARIATIONS project's technology will be deployed in the Music Library and Recital Hall Building now under construction with projected occupancy of the Music Library wing by December 1995. Once deployed, it will be used by students enrolled in music courses (and others) on the Bloomington and Indianapolis campuses.
During the last quarter of 1993, we successfully demonstrated the project at COMDEX and at IBM's Leadership Conference in Chantilly, Virginia. We also gave two sessions on the project at IBM's ACIS conference in February, 1994. During the second half of 1994, the VARIATIONS project investigated competing technologies, valuated the continuing evolution of Mosaic clients, completed testing Netscape clients, and identified the server and networking technologies required for the projects. During the first part of 1995, we refined design decisions, continued many of the preceding activities, and began constructing an IBM ATM testbed. Initially, this ATM LAN was successfully tested using a Novell Netware server. Next steps include testing the ATM LAN with IBM's OS/2 Ultimedia Server and then with RS6000-based software. In the second half of 1995, the VARIATIONS project will make final decisions about server, network and client technologies. The Music Library will move into these facilities in December 1995.
In January 1996, the VARIATIONS project will distribute digital sound and some full-motion video for all musical works on course reserve (about 300 titles) and provide real-time digitization of analog sources responding to individual requests (see the description of the digitization center). The new Music Library has a Media Center consisting of a computer classroom cluster, a machine room for servers, a studio for the creation of multimedia information, and editing and support facilities. In addition, within the first year, 100 of its 375 patron spaces will contain computers to use the VARIATIONS project's database. Infrastructure is being installed so that the remaining 275 patron spaces may be added to the network as needed. Data from the VARIATIONS project will also be distributed to faculty offices and classrooms in the Music Library and Recital Hall Building. Following the January 1996 deployment VARIATIONS will be distributed to the other buildings in the School of Music complex, to the rest of the campus and to the music program on the IUPUI campus as quickly as ATM network deployment will allow.
Most of the challenges facing the VARIATIONS project involve integration. Hardware integration issues, especially with emerging server and network technologies, form one set. Creating an interface that integrates these technologies for the user is another challenge. The proof of concept applications and technology VARIATIONS has demonstrated to date have successfully delivered digital sound and full-motion video from a server over a network to as many as eight clients. While this progress is significant within a national context, it does not begin to assure us that these accomplishments will necessarily scale up to the level of performance that is the central focus of this project: 130 workstations in the Music Library plus those in Indianapolis accessing data and applications from the Music Library's machine room with many of them using continuous, multimedia data streams..
We have closely tracked the development of Z39.50 clients. Two years ago, we created one under NeXTStep as a proof-of-concept project. Since early Fall, we have been concerned more with the integration of Z39.50 into WWW environments. The gateway can be found elsewhere on this server.
The VARIATIONS network is being designed to support the delivery of CD quality stereo audio to 130 local workstations and to a smaller number of workstations located on remote campuses. We have tested various audio compression techniques and have found that acceptable audio qualities can be delivered within the range of 384 Kb/s to 1.5 Mb/s of bandwidth. In addition, digital video consisting of streams measuring from 1.5Mb/s (MPEG-I) to 6Mb/s (MPEG-II), and traditional, bursty, LAN traffic with bursts measuring up to 6Mb/s will be supported. Audio will be stored on several file servers in a compressed format (MPEG) and will be decompressed by clients in real time.
Network Technology Overview
In order to guarantee adequate bandwidth to any given host involved with this project, three areas of the network topology must be considered. First, each host's network connection must be capable of supplying the aggregate bandwidth needs of each of the applications running on that host. Second, the project's backbone network must be able to support the sum bandwidth requirements of all hosts that communicate over that portion of the network. Third, the connections to the project's servers must support the sum bandwidth requirements of all clients served. Bandwidth allotment in each of these areas must be contention free to avoid frame/cell loss and subsequent deterioration of audio quality.
Local Network Environment
VARIATIONS is a demonstration project. Thus, while we have installed Category 5 twisted-pair wiring in the building that houses the VARIATIONS project, we plan to use some components that would operate using Category 3 twisted-pair wiring for client connectivity, rather than focusing on higher-speed options that would be limited to Category 5 wiring or fiber. In particular, low-speed ATM for client connectivity and limit higher speed ATM (OC-3) connections to server connections and for connections between ATM switches. Please refer to the Maps and Diagrams section for graphic depictions of the planned network.
Wide Area Network Environment
As part of a larger network expansion effort, Indiana University plans to lease a clear channel DS3 circuit between its Indianapolis and Bloomington campuses and T1 circuits to each of its regional campuses. T3 access devices allow for efficient utilization of the T3 bandwidth by converting various access media (e.g. data, circuit based video codecs, and voice) to ATM for transport over the intercampus trunk (the T3 connection will be a "clear channel" service). These access devices will provide infrastructural support for workstations used by students and faculty in Indianapolis to access materials, including digital audio, from the Music Library in Bloomington.
Network Issues
While much theoretical research has focused on the design of ATM networks, there has been very little effort devoted to building large-scale demonstration ATM networks comprised of relatively inexpensive components and utilizing existing infrastructure. Indiana University was one of the first educational institutions to implement 10baseT on a large scale (i.e. 10,000 nodes). It has also been very successful in integrating the research work of others into well-managed and easily implementable network services. We plan to utilize our extensive expertise in systems integration to provide an affordable working network model that other providers and consumers of continuous media materials will be able to emulate.
There are many open issues surrounding the implementation of ATM and its integration with existing networks. Some of these issues are vendor interoperability (especially switched virtual circuit support); LAN Emulation implementations for support of multiple protocols; network layer address to SVC mapping; route and ARP server implementation; network management schemes; quality of service negotiation; and congestion control. Many of these issues will not be solved in the short term (1-2 years). VARIATIONS hopes to provide the community with insight into how to implement functional, low cost, ATM networks in the short term, given the host of unsolved issues mentioned above. We feel this insight will be very helpful in furthering the goals of the NII by providing many potential producers of information with a model for developing low cost solutions to continuous media production and delivery.
The following are quantifiable or objective goals by which progress can be measured:
