We believe that a congestion control algorithm should make flows finish quickly - as quickly as possible, while staying stable and fair among flows. Recently, we proposed RCP (Rate Control Protocol) which enables typical Internet-sized flows to complete one to two orders of magnitude faster than the existing (TCP Reno) and the proposed (XCP) congestion control algorithm. Like XCP, RCP uses explicit feedback from routers, but doesn't require per-packet calculations. A router maintains just one rate that it gives to all flows, making it simple and inherently fair. Flows finish quickly because RCP aggressively gives excess bandwidth to flows, making it work well in the common case. However - and this is a design tradeoff - RCP will experience short-term transient overflows when network conditions change quickly (e.g. a route change or flash crowds). In this paper we extend RCP and propose RCP-AC (Rate Control Protocol with Acceleration Control) that allows the aggressiveness of RCP to be tuned, enabling fast completion of flows over a broad set of operating conditions.

Xunet 2 is an experimental wide-area network that serves as a testbed for research on data communications techniques. The authors provide an overview of the architecture of the Xunet 2 testbed. Xunet 2 has three principal components: routers, switches and network management systems. Long-haul transmission between them will initially use DS3 facilities. Performance objectives for the backbone have been set to cover a wide range of transmission speeds. Each router attaches a local area network to the backbone network. The backbone network provides virtual circuit connectivity between the routers. Circuit setup and teardown is done by a control computer associated with each switch, and the network as a whole will be managed by a network management station in the network operations center.>

A transport layer protocol optimized for use in a byte stream network is presented in detail. A protocol engine for Datakit networks, the universal receiver, that responds to commands which are embedded in the incoming byte stream is discussed. Various ways in which a transmitter might communicate with a universal receiver and some of the performance consequences of this transport layer design are also discussed.>

It is anticipated that a ubiquitous data transport system will require integration of local area and wide area networks. The combined network will need to present a uniform appearance to the user, be effective as a transport mechanism for a great variety of traffic patterns, and be economically appropriate for a wide range of consumer products. Learning from telephony and from experiments with a local area network, we conclude that the transport system must seek a clean separation of function and protocol between the network and its users. This separation is achieved by a byte-stream architecture that carries control and data bytes over switched virtual circuits. A DATAKIT packet switch demonstrates how the byte-stream concept can integrate local area and wide area network objectives. This switch is an assembly of interface modules connected by a pair of short passive buses. Each type of interface module serves one type of remote equipment and, if need be, terminates the protocol of that equipment. There are interface modules for trunks that lead to other packet switches, for terminals, and for host computers. Other modules provide system timing, switching, network control, and maintenance support.