John A. Morse

In the late '80s, "multimedia" was a magic word. It seduced us with glimpses of a brave new world where audio and video technology merged with computer technology. It promised us everything from instant high-impact business presentations to virtual reality. Words like "paradigm shift" and "multibillion-dollar industry" were enough to snare both the technophiles and the eager entrepreneurs into believing that the world had suddenly changed, and we were all going to get rich in the process.

Somewhere on the way to the bank, reality set in, and it wasn't virtual. The reality is that multimedia is a lot harder than it looks. Successful multimedia requires a marriage between analog TV technology and digital computer technology; it requires reconciliation between a technical /professional marketplace and a consumer marketplace. As in any marriage, a lot of hard work is required to make it succeed, and much of that work is yet to be done.

For certain segments of the computer industry, multimedia was relatively easy to implement and so caught on quickly. The first successes have been at the extremes of the cost spectrum-very low-end desktop multimedia on the one hand, and very high-end virtual reality systems on the other. This has left Digital, with its traditional focus on the middle, temporarily out of the game.

For desktop multimedia, all that is required is the ability to capture and display video and audio. Since machines like the Commodore Amiga were already based more on TV technology than on computer technology (for cost reasons), they could be quickly and cheaply adapted to handle audio and motion video. Thus desktop multimedia was born. The CD-ROM, adapted from audio CD technology, was the perfect storage medium for distribution of multimedia content; and so for this market segment, CD-ROM and multimedia became almost synonymous. There has emerged a whole industry based around the production of multimedia titles on CD-ROM.

At the high end, for purposes such as full-realism aircraft simulation or virtual reality applications, the solution was to use the highest performance hardware available, at whatever expense. Typically, high-end, three-dimensional graphics systems were coupled either to supercomputers or to massively parallel processor arrays. The result was, and still is, impressive. But the cost is still so high that such virtual reality systems are not yet commercially viable except in specialized low-volume markets.

The vast area in the middle, into which all of Digital's business falls, has developed very slowly. The problem is that our business is based on a model of enterprise-wide computation. The computer systems we design and sell not only include processors and displays but incorporate networks and servers as well. To introduce multimedia into such a model, one touches every aspect of the system, from the desktop, through the network, and back to the servers. At every turn, we have found that the technology that has

evolved over 30 to 40 years for handling numbers, text, and (more recently) two-dimensional and three-dimensional graphics is not quite right for video and audio. Every component of the system, both hardware and software, needs to change in some way. We need to evolve to a model of networked client- server multimedia computing. Change of this magnitude is a slow process.

Two challenges are so pervasive that almost every paper in this issue addresses them, each from a different perspective. First of all, multimedia involves the handling of large quantities of data. Second, for many applications, that data must be handled under very tight time constraints. The resulting stress and strain on all components of the system translates into a set of technical challenges that has occupied us for the last four years and promises to keep us busy through at least the rest of this decade.

Depending on the picture quality chosen, it may require from one million to one hundred million bytes of storage to save each second of live video in digital form. Since many applications of multimedia, such as archiving television footage for research or historic preservation purposes, will need to save many hours of video, it is easy to see that multimedia quickly builds demand for many gigabytes (1,000,000,000 bytes) of magnetic or optical disk storage. But storage is only part of the problem. Once such enormous amounts of data are stored, the challenge becomes how to retrieve a particular item of interest. Standard database techniques are oriented toward retrieval of text and numbers. Retrieval of audio and video information will require new file and database techniques that are only beginning to be understood.

An obvious application of multimedia technology, once the networks are in place, is teleconferencing. We can envision a day when we can connect to anyone any place in the world via the network and carry on a conversation with them, while each of us sees the other in full-motion video, using the audio and video capabilities of our desktop workstations and PCs. But realizing this vision has proved surprisingly hard. People expect the images they see to be synchronized with the sounds they hear, and they expect delays to be no worse than those experienced on a long-distance telephone call. Unfortunately, data networks have been designed to maximize throughput and reliability. They do this at the expense of some delay in transmission-delay that is annoying at best, and unacceptable at worst, for teleconferencing applications.

Successful infusion of multimedia technology into enterprise-wide computation is proving to require change on a scale that almost no one anticipated. We at Digital are in the midst of this process of change, and this issue of the Digital Technical Journal is a snapshot, taken at one point in time, of that process. Together, the papers describe some of the toughest technical challenges that we face and in many cases give glimpses into possible solutions.