. Network parallel computing is the use of diverse computing resources interconnected by general purpose networks to run parallel applications. This paper describes NetFx, an extension of the Fx compiler system which uses the Fx model of task parallelism to distribute and manage computations across the sequential and parallel machines of a network. A central problem in network parallel computing is that the compiler is presented with a heterogeneous and dynamic target. Our approach is based on a novel run--time system that presents a simple communication interface to the compiler, yet uses compiler knowledge to customize communication between tasks executing over the network. The run-- time system is designed to support complete applications developedwith different compilers and parallel program generators. It presents a standard communication interface for point--to--point transfer of distributed data sets between tasks. This allows the compiler to be portable, and enables communicati...

Abstract-Internetworks that are global in scale, contain multiple administrative domains, and support a range of services present special requirements for routing. Multiple administrative domains introduce the need for policy-sensitive routing. Service heterogeneity intensifies the requirement for type of service (TOS) routing, as well as other protocol support for handling a range of services, from datagrams to multimedia streams. This paper summarizes the key concepts and protocols developed as part of the Interdomain Policy Routing (IDPR) architecture. We place particular emphasis on the route installation and packet forwarding mechanisms because they are critical to protocol performance and differ significantly from current practice in datagram wide area networks. I.

The growth of core Internet routing tables has been such that it is now viewed as an impediment to the continued expansion of the Internet. The main culprit is multi-homing that stems from sites’ desire for greater reliability and diversity in connectivity. These locally rational decisions have a global impact on the Internet, and there is currently no mechanism to effectively control them. A number of technical solutions are being pursued, but this paper explores the use of a “market mechanism. ” It formulates a model that accounts for sites ’ incentives and the impact their connectivity choices have on the size of routing tables, and introduces a pricing scheme that seeks to better reapportion the resulting costs. The model is solved for two configurations that capture different deployment realizations and stages. They demonstrate the scheme’s effectiveness in controlling the growth of Internet routing tables, while improving the welfare of sites and Internet Service Providers. 1.