This section summarizes the testbed investigations and findings. The emphasis here is on bringing together the work by topic area -- detailed individual testbed perspectives are provided in the final reports written by each of the testbed research organizations. The topics are:
· transmission
· switching
· interworking
· host I/O
· network management
· applications and support tools
Each subsection begins with a brief summary of the investigations and findings for that area. Figures have been included to assist in conveying some of the information, but readers are encouraged to consult the individual reports for more complete expositions of the material. Specific references are given where a particular report or publication is especially relevant -- in other cases, the final report(s) for the relevant testbed or organization should be assumed as the implicit reference.
One of the stated project goals was to "investigate alternative architectures for gigabit networking". Because of the relatively primitive stage of gigabit technology existing at the beginning of the project and the experimental nature of the testbeds, this of necessity translated into investigations focused on the elements required to realize such a network, rather than on a set of "complete" network architectures. In fact, the rapid pace of today's technology changes effectively dictates an incremental approach to successful networking systems -- a top-down system design is almost assured of being overtaken by technology advances before it can be fully implemented.
Nevertheless, at least four distinctive end-to-end network layer architectures were used as research platforms in the testbeds. This was a result both of architecture component choices made by researchers after the work was underway and of the a priori testbed selection process discussed in an earlier section. Figure 4-1 gives a high-level view of these end-to-end architectures.
Figure 4-1. Testbed Architectures
Figure 4-1A, Seamless Wide-Local Area ATM, reflects one of the resulting architectures which was investigated in the Aurora testbed. In this case all networks use ATM cell switching and are interconnected without use of an IP layer. SONET is used as the underlying transmission technology for the wide area network, and in some cases also for local distribution. In other cases other transmission technologies are used under ATM in the local area. In addition to the usual notion of LANs, Aurora introduced ATM Desk Area Networks, or DANs, as part of the overall architecture.
Figure 4-1B, Heterogeneous Wide Area ATM and Local Area technologies, shows the architecture used in the Blanca, Nectar, and Vistanet testbeds. ATM is again the switching technology used in the wide area network, but a non-ATM technology (in these instances, HIPPI) is used for local area connectivity. Both bridging and gateway approaches were investigated, and wide area transmission included SONET and all-optical infrastructures.
Figure 4-1C, Seamless Wide-Local Area PTM, is the Packet Transfer Mode analogue to the seamless ATM architecture and, like the latter, was part of the Aurora testbed work. In this case variable-length packets are forwarded across both the local and wide area networks, which were designed to operate as an integrated system.
Figure 4-1D, Wide Area HIPPI via Local Switching, reflects the resulting Casa testbed architecture. In this case HIPPI was used for both local and wide area switching, with SONET providing the wide area transmission infrastructure. The combination of one or more specially designed gateways at each site, in conjunction with a site's local HIPPI switch, provided the wide area routing/switching structure for variable-length packet forwarding through intermediate sites.
The following subsections address, by topic, the technology investigations associated with these architectures, along with the work on applications and the metalevel aspects of the testbeds. It should be noted that, while some work was specific to a particular architecture, in many cases the results can be applied to one or more of the other architectures of Figure 4-1 and to other architectures not considered here.