PDIS and The Information Highway

Advanced networking technologies offer opportunities for new and better PDIS. 1.0 The Information Hypeway and Reality ‘‘For a successful technology, reality must take precedence over public relations, for Nature cannot be fooled.’’ [4] The so-called ‘‘Information Highway’’ attracts a great deal of attention, and yet a precise definition eludes us. This tends to be the result of definitions of convenience and overuse of the roadway analogy. There are two trendlines that are real. The first, and the one that will have the most impact, is the expansion of the Internet and commercial analogues. The impact stems directly from the scale and services of these networks, and the fraction of the population which can access them. The other trendline is the technological changes in higher performance networking elements. This trendline presents problems and opportunities for PDIS researchers, unlike the largely operational difficulties of expanding access. Since the backbone technologies will extend their reach to the home, the problems of scale are only postponed. We now focus on Asynchronous Transfer Mode (ATM). 2.0 ATM as Information Asphalt In the past half-decade three major developments in the high-speed networking [11] occurred: (1) deployment of fiber optic technology; (2) availability of fast packet switches; and (3) network architectures which support paced cell traffic. Fiber-optic technologies such as Synchronous Optical Network (SONET) [7], which has commonly deployed rates of OC-3 (155 Mbps), OC-12 (622 Mbps), and OC-48 (2.4 Gbps) are now available, providing a high-bandwidth transmission infrastructure. Such infrastructures have been deployed and tested, as shown in the AURORA Gigabit Testbed’s ATM network topology (partial), Figure 2, after an illustration of its geography in Figure 1. The highest-performance networks are based on the idea of packet or cell-switching to share links between topologies and to share bandwidth between applications. The concept of Asynchronous TimeDivision Multiplexing (ATDM), used in modern cellswitched architectures, makes link-sharing programmable. Switched communications networks, as exemplified by the Public-Switched Telephone Network [10] use a digital Hierarchy based on Synchronous Time Division Multiplexing (STDM) internally. Higher-rate channels are aggregated from multiple lower-speed channels in a process called ‘‘multiplexing’’; the inverse process disintegrates the higher-speed channel into its lower-speed components. In a synchronous hierarchy, a clock is used for channel identification. Synchronous multiplexing can thus be implemented entirely in hardware, yielding a multiplexing scheme which is simple, fast, and relatively inflexible with respect to bandwidth dynamics. The computer communications community has attempted to address the dynamics of typical computer communications traffic through the idea of packetswitching [2] which is well-suited to the burstiness inherent in computer communications. Most data communications networks are optimized for this bursty behavior, especially local-area technologies [9]. With the Asynchronous Time Division Multiplexing (ATDM) [5] technique a hardware clock ‘‘tick’’ is replaced with a virtual ‘‘channel’’ identifier (VCI). Thus, logical sub-channels are assigned with a tag (’circuit ID’). This tagging allows burstiness to be accommodated by allocating more of the tags to a busy queue. In addition, this tagging can clearly be software controlled. The fact that this multiplexing behavior can be software controlled means that the problem has been transformed from a simple hardware system to a much more complex, but much more flexible, combination of hardware and software. The control of tagging is the key to resource allocation, and brings with it all the problems and opportunities of distributed control, resource allocation versus dynamics, real-time software, reservations, and policy versus mechanism. ATM is a specific implementation of the ATDM technique, using 48 byte ‘‘cells’’ of data with 5 byte headers containing the VCI. As of about five years ago [3], the main cell-switching challenges were believed to be: (1) switching at high speed (per-cell routing); (2) delivering of data to applications in their own data units at high speed; and (3) balancing ATM flexibility with distributed resource