ATM (asynchronous transfer mode) is an international telecommunications standard designed for broadband integrated services; it is also well-suited for use within local-area networks. ATM LANs can provide the networking support needed for multimedia communication at rates of 100 Mbits/sec and higher. The authors first analyze the protocol processing required to handle ATM communication. Based on this analysis, they then discuss the architectural issues in the design of host interfaces for ATM local-area networks. In particular, they conclude that a simple host interface, which leaves most of the ATM protocol processing to be done by the host computer, supports good performance for data communication (around 100 Mbits/sec). However, to support real-time video and audio communication, the ATM interface should include an embedded processor.>

Multicomputersbuiltaround a general network are now a viable alternative to multicomputersbased on a system-speci~c interconnect because of architectural improvements in two areas.First, the host-network interface overhead can be minimized by reducing copy operations and host interrupts.Second, the network can provide high bandwidth and low latency by using high-speed crossbar switches and efficient protocol implementations.While still enjoying thejexibility of general networks, the resulting network-based multicomputers achieve high performance for typical multicomputer applications that use system-specijic interconnects.We have developed a network-based mtdticomputer called Nectar that supports these claims.

A spectrum of abstractions for multicast communications is introduced in increasing order of both desirability and semantic level: functional mapping, iterators, and streams. Examples of distributed algorithms from the literature are used to illustrate the expressive power of each mechanism. Streams, in particular, provide first-class status for multicast communication in progress and can be implemented efficiently in typical multicast communication architectures. Related work and the criteria for multicast support are discussed. The layers of a typical distributed system that are relevant to language support for multicast communication are discussed.>

We have built a high-speed local-area network called Nectar that uses programmable communication processors as host interfaces. In contrast to most protocol engines, our communication processors have a flexible runtime system that supports multiple transport protocols as well as application-specific activities. In particular, we have implemented the TCP/IP protocol suite and Nectar-specific communication protocols on the communication processor. The Nectar network currently has 25 hosts and has been in use for over a year.

Nectar is a “network backplane” for use in heterogeneous multicomputers. The initial system consists of a star-shaped fiber-optic network with an aggregate bandwidth of 1.6 gigabits/second and a switching latency of 700 nanoseconds. The system can be scaled up by connecting hundreds of these networks together.

A troupe is a set of replicas of a module, executing on machines that have independent failure modes.Troupes are the building blocks of replicated distributed programs and the key to achieving high availability.Individual members of a troupe do not communicate among themselves, and axe unaware of one another's existence; this property is what distinguishes troupes from other software architectures for fault tolerance.Replicated procedure call is introduced to handle the many-to-many pattern of conmmnication between troupes.The semantics of replicated procedure call can be summarized as exactly-once execution at all replicas.An implementation of troupes and replicated procedure call is described, and its performance is measured.The problem of concurrency control for troupes is examined, and a commit protocol for replicated atomic transactions is presented.Binding and reconfiguration mechanisms for replicated distributed programs are described.

A new mechanism for constructing highly available distributed programs is described. It combines remote procedure call with replication of program modules for fault tolerance. The set of replicas of a module is called a troupe. In a program constructed from troupes, what appears to the programmer as a single inter-module procedure call results in a replicated procedure call. A distributed program constructed in this way will continue to function as long as at least one member of each troupe survives. The semantics of replicated procedure calls and troupes are defined and algorithms are presented that support these semantics

The query languages used in relational database systems are a special class of programming languages. The majority, based on first-order logic, lend themselves to analysis using formal methods. First, we provide a definition of relational query languages and their expressive power. We prove some general results and show that only a proper subset of first-order logic formulas may be used as a practical query language. We characterize this subset in both semantic and syntactic terms. We then analyze the expressive power of several real query languages, including languages based on the relational calculus, languages with set operators and aggregate functions, and procedural query languages.Since the partial ordering "is more expressive than" determines a lattice among relational query languages, the results of the paper may be viewed as determining some of the structure of this lattice. We conclude with some applications of the results to the optimization problem for query processing.

The reliability of atomic commit protocols for distributed systems is investigated. Recent research has proved that blocking is unavoidable after certain site or network failures. The results of this paper enable one to quantify the expected amount of such blocking.A probabilistic model of a distributed system is developed, consisting of two components: a model of the communication network and a model of distributed transaction processing. We introduce the notion of the window of uncertainty for a site participating in a distributed transaction, which characterizes when that site is vulnerable to blocking in the event of a network partition. The model is then used to evaluate and compare the expected number of blocked sites for several atomic commit protocols.