Experiences with the Amoeba Distributed Operating System

ACM Computing Surveys, 1985

Distributed operating systems have many aspects in common with centralized ones, but they also differ in certain ways. This paper is intended as an introduction to distributed operating systems, and especially to current university research about them. After a discussion of what constitutes a distributed operating system and how it is distinguished from a computer network, various key design issues are discussed. Then several examples of current research projects are examined in some detail, namely, the Cambridge Distributed Computing System, Amoeba, V, and Eden.

UNIX has been around now for almost 20 years. At the time UNIX began, most departments felt themselves well-endowed indeed if they owned a single PDP-11/45 with 256K memory and a 2.5M RK05 disk. Nowadays a laptop would be embarrassed to have only that. It is our hypothesis that UNIX is no longer the appropriate kind of operating system for the 1990s. In this paper, a new system, Amoeba, will be described, that we believe meets the requirements for distributed computing in the 1990s.

Unlike many other operating systems, Amoeba is a distributed operating system that provides group communication (i.e., one-to-many communication). We will discuss design issues for group communication, Amoeba's group system calls, and the protocols to implement group communication. To demonstrate that group communication is a useful abstraction, we will describe a design and implementation of a fault-tolerant directory service. We discuss two versions of the directory service: one with Non-Volatile RAM (NVRAM) and one without NVRAM. We will give performance figures for both implementations.

1991

This paper compares two distributed operating systems, Amoeba and Sprite. Although the systems share many goals, they diverged on two philosophical grounds: whether to emphasize a distributed computing model or traditional UNIX-style applications, and whether to use a workstation-centered model of computation or a combination of terminals and a shared processor pool. Many of the most prominent features of the systems (both positive and negative) follow from the philosophical differences. For example, Amoeba provides a high-performance user-level IPC mechanism, while Sprite's RPC mechanism is only available for kernel use; Sprite's file access performance benefits from client-level caching, while Amoeba This work was supported in part by the Netherlands Organization for Scientific Research (N.W.O.) under grant NF 62-334.

Computer, 1991

udging from the enormous amount of distributed-system research carried out over the past decade, information processing experts have come t o recognize the advantages these systems possess. These research activ-J ities have led to the availability of more than 50 network and distributed systems. However, most of these systems can only partially succeed in attaining the major goals of a distributed system, which include transparency, higher performance, higher reliability and availability, and higher scalability. Of course, attaining all these goals in the first attempt is impossible. Nonetheless, gradual improvements are possible by learning from existing systems and trying to overcome their limitations. The University of Tokyo's Galaxy research project attempts to design, implement, and use a distributed computing environment based on this idea. Design goals and novel aspects Analyzing why existing In designing Galaxy, our primary goal was to build a distributed operating system suitable for use in both local and widearea network workstations. T o achieve this goal, we eliminated the use of broadcast protocols in any of our mechanisms, avoided the use of global-locking o r time-stamp-ordering mechanisms while maintaining the consistency of replicated information, and decentral-distributed systems are limited, the Galaxy research project borrows many concepts and proposed facilities ized the management of all globally useful information. novel system features that helped us achieve this goal include O u r second major goal was to ensure high performance. Design decisions and and integrates them developing the Galaxy kernel from scratch, rather than modifying a n existing using multiple-level interprocess communication mechanisms to meet the with its own novel Unix kernel; design features.

The VAMNET is a hybrid operating system environment for distributed applications in a heterogeneous environment, concerning both the hardware architectures used and operating systems already present, for example the UNIX−OS. The VAMNET consists of several parts. Some of them can operate standalone. All of them build up a hybrid distributed operating system environment with some new features never seen before. These parts are:

1979

Roscoe is an operating system implemented at the University of Wisconsin that allows a network of microcomputers to cooperate to provide a general-purpose computing facility. After presenting an overview of the structure of Roscoe, this paper reports on experience with Roscoe and presents several problems currently being investigated by the Roscoe project.

The current direction of computer systems is increasingly moving away from centralised system and moving towards distributed sets of workstations and personal computers connected by a network, known as distributed systems. This has led to the development of specialised operating systems to take full advantage of the resource sharing potential that a distributed system provides. The specialised operating systems being developed for these situations are known as Distributed Operating Systems. Research into this area is still at the primary stage with little firm evidence indicating what components of Distributed Operating systems are more important than others.

Parallel Computing, 2002

Present operating systems are not built to support parallel computing--they do not provide services to manage parallelism, i.e., to globally manage parallel processes and computational resources. The cluster operating environments that are used to assist the execution of parallel applications do not provide support for both programming paradigms, message passing (MP) or distributed shared memory (DSM)--they are mainly offered as separate components implemented at the user level as library and independent server processes. Due to poor operating systems users must deal with clusters as a set of independent computers rather than to see this cluster as a single powerful computer. A single system image (SSI) of the cluster is not offered to users. There is a need for an operating system for clusters. We claim and demonstrate in this paper that it is possible to develop a cluster operating system that is able to efficiently manage parallelism; use cluster resources efficiently; support MP in the form of standard MP and PVM, and DSM; offer SSI; and make it easy to use. We show that to achieve these aims this operating system should inherit many features of a distributed operating system and provide new services which address the needs of parallel processes, cluster's resources, and application developers. In order to substantiate the claim the first version of a cluster operating system managing parallelism and offering SSI, called GENESIS, has been developed. Ó

1986

As distributed computing becomes more widespread, both in high-energy physics and in other applications, centralized operating systems will gradually give way to distributed ones. In this paper we discuss some current research on five issues that are central to the design of distributed operating systems: communications primitives, naming and protection, resource management, fault tolerance, and system services. For each of these issues, some principles, examples, and other considerations will be given.