A unified environment for distributed computing
T.B. Gendreau
Abstract
T.B. Gendreau
Abstract
A distributed system is a combination of a distributed architecture and distributed control algorithms. The potential advantages of distributed architectures include increased reliability, resource sharing, and computation speed-up. The distributed control algorithms are software which manages the distributed architecture in order to achieve the advantages listed above. The overall goal of the distributed control algorithms is to provide a unified service environment. The unified service environment provides a number of views or interfaces to the distributed system. These views may range from a simple menu driven interface in which the distributed nature of the system is completely hidden from the user to complex distributed programming environment. Three problems that occur in the context of the development of a unified service environment include the development of an appropriate interprocess communication system, scheduling of user requests, and development of distributed applications. A universal set of interprocess communication primitives is proposed. This set of primitives includes the concept of a dynamic group which allows processes of a distributed algorithm to bind together for communication purposes. Two scheduling algorithms, bidding and drafting, are compared in an Ethernet based client/server environment of five workstations. The comparison is based on an emulation of these algorithms and our results indicate that the simpler bidding algorithm outperforms the drafting algorithm in this type of environment. An environment for investigating distributed algorithms called distributed game playing is proposed. Its relationship to problems in distributed computing is shown through a number of illustrative problems.
A significance statement is not available in the OpenAlex record.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
A distributed system is a combination of a distributed architecture and distributed control algorithms. The potential advantages of distributed architectures include increased reliability, resource sharing, and computation speed-up. The distributed control algorithms are software which manages the distributed architecture in order to achieve the advantages listed above. The overall goal of the distributed control algorithms is to provide a unified service environment. The unified service environment provides a number of views or interfaces to the distributed system. These views may range from a simple menu driven interface in which the distributed nature of the system is completely hidden from the user to complex distributed programming environment. Three problems that occur in the context of the development of a unified service environment include the development of an appropriate interprocess communication system, scheduling of user requests, and development of distributed applications. A universal set of interprocess communication primitives is proposed. This set of primitives includes the concept of a dynamic group which allows processes of a distributed algorithm to bind together for communication purposes. Two scheduling algorithms, bidding and drafting, are compared in an Ethernet based client/server environment of five workstations. The comparison is based on an emulation of these algorithms and our results indicate that the simpler bidding algorithm outperforms the drafting algorithm in this type of environment. An environment for investigating distributed algorithms called distributed game playing is proposed. Its relationship to problems in distributed computing is shown through a number of illustrative problems.
Key concepts: Computer science, Distributed computing, Distributed Computing Environment, Distributed algorithm, Inter-process communication, Distributed design patterns, Distributed object, Virtual machine