This paper describes a new methodology for protocol design, using incremental construction of the protocol from elements called "protocol boosters" on an as-needed basis. Protocol boosters allow: (1) dynamic protocol customization to heterogeneous environments and (2) rapid protocol evolution. Unlike alternative adaptation approaches, such as link layer, conversion, and termination protocols, protocol boosters are both robust (end-to-end protocol messages are not modified) and efficient (do not replicate the functionality of the end-to-end protocol). We give examples of error and congestion control boosters, and give initial results from booster implementations.

Software implementations of error detection codes are considered to be slow compared to other parts of the communication system. This is especially true for powerful error detection codes such as CRC. However, we have found that powerful error detection codes can run surprisingly fast in software. We discuss techniques for, and measure the performance of, fast software implementation of the cyclic redundancy check (CRC), weighted sum codes (WSC), one's-complement checksum, Fletcher (1982) checksum, CXOR checksum, and block parity code. Instruction count alone does not determine the fastest error detection code. Our results show the computer memory hierarchy also affects performance. Although our experiments were performed on a Sun SPARCstation LX, many of the techniques and conclusions will apply to other processors and error detection codes. Given the performance of various error detection codes, a protocol designer can choose a code with the desired speed and error detection power that is appropriate for his network and application.

Reordering and reassembly of data before processing can reduce communication system performance as seen by the application. We examine a method of explicitly labelling blocks of data with sufficient information to allow processing of misordered data. Our labelling syntax for data blocks, which we call chunks, is cleaner and more general than that of other protocols. We show how chunks can be used for efficient fragmentation/reassembly and compare chunks with other fragmentation systems. End-to-end error detection is complex for chunks or other systems that allow both fragmentation and processing of misordered data. We show that it is possible to design an end-to-end error detection system that does not compromise chunk processing performance. Chunks can take advantage of processing techniques such as Integrated Layer Processing and can be used to implement concepts such as Application Layer Framing [CLAR 90].

A framework for understanding high-speed communication architectures and protocols is presented. The framework is validated by showing how several concepts fit into it. Two approaches to reduce the number of processing operations required for a single datum, the reduction of control processing and the elimination of unneeded or replicated operations, are discussed. Two other approaches for increasing communication system speed by performing more processing operations per unit time, parallel processing and increased processor utilization, are also discussed.>

This paper considers some of the multiplexing issues in communication system design by examining overall system issues. In particular, we distinguish physical multiplexing of resources from logical multiplexing of streams. Both physical-resource multiplexing and logical multiplexing determine the service that can be provided by a communication system. We also discuss two issues affected by logical multiplexing - flow control and the relationship between control and data streams of a connection.

High-speed packet-switched communications to the home can be provided using an existing community antenna television (CATV) system. The author focuses on an efficient access scheme for a network that operates on a CATV system. The scheme takes advantage of packet address locality to minimize the expected network access overhead. Its efficiency is compared with that of other access schemes that might be used on a CATV-based network.>

A way to increase gateway throughput is to reduce the routing-table lookup time per packet. A routing-table cache can be used to reduce the average lookup time per packet and the purpose of this study is to determine the best management policies for this cache as well as its measured performance. The performance results of simulated caches for a gateway at MIT are presented. These results include the probability of reference versus previous access time, cache hit ratios, and the number of packets between cache misses. A simple, conservative analysis using the presented measurements shows that current gateway routing-table lookup time could be reduced by up to 65%.>