Integrative monitoring and control framework based on software distributed shared memory non-locking model.

Khalil, MA, 2004. Integrative monitoring and control framework based on software distributed shared memory non-locking model. PhD, Nottingham Trent University.

[thumbnail of 10183440.pdf]
Preview
Text
10183440.pdf - Published version

Download (22MB) | Preview

Abstract

Distributed shared memory (DSM) paradigm provides an illusion of one physical shared memory in network of workstations where in reality shared data physically reside on different machines and different address spaces. DSM algorithms facilitate accessing the shared memory and exchanging data via normal read and write operations, concealing the inter-process communication and remote memory accesses. Such algorithms, and the environments they belong to, are used as processing platforms for distributed and parallel applications. This thesis investigates new approaches and algorithms for improving the performance of distributed shared memory systems. It assumes that the reduction of data retrieval time from the point of view of distributed applications is a major factor for measuring the performance of DSM systems.

The investigation introduces a framework that uses a non-locking approach and a purposely designed memory consistency model in order to achieve the above mentioned goal. This approach allows an application in a distributed environment to access the shared memory in a nearby location in a relatively short time thus saving valuable time for performing its native tasks. The framework is presented as a computing environment for building hierarchical traffic telematics distributed systems. It develops further the successful features of DIME DSM system (developed and designed at DOCM, NTU) and at the same time avoids its shortcomings. The main feature of the architectural design of the new framework is its flexibility, which allows the reconfiguration of the communication paths or routes of the system at run-time, thus improving the overall performance of the whole system.

To maintain consistent view of the distributed shared memory in the framework, a variant definition of sequential consistency (SC) model has been developed. This model has been designed specifically to support certain features in the urban traffic control (UTC) system. It also incorporates the flavour of SC definition that is intuitively favoured by distributed applications programmers. Also, to reliably manage the dissemination of messages and data across the distributed system, the thesis presents a novel proprietary communication protocol for the framework. This protocol sends messages only to the applications that are involved in the operation rather than broadcasting the messages to every application in the distributed system. The algorithm of the protocol reduces the number of messages exchanged in the system, and therefore saves the resources of the network. Furthermore, a new novel heuristic algorithm is presented in this thesis, which allows system re-configurability at run-time and optimization of the performance of DSM systems.

The research work discusses some important issues for designing and building distributed systems. It also discusses new techniques that have been introduced in recent research papers for improving the performance of DSM systems. The presented implementation of the proposed framework demonstrates the applicability of the non-locking approach and the consistency model for building DSM systems.

Item Type: Thesis
Creators: Khalil, M.A.
Date: 2004
ISBN: 9781369316414
Identifiers:
Number
Type
PQ10183440
Other
Rights: The copy of this report has been supplied on the understanding that it is copyright material, and that no quotation from the report may be published without proper acknowledgment. The work described in this report is the Author’s own, unless otherwise stated, and it is, as far as he is aware, original.
Divisions: Schools > School of Science and Technology
Record created by: Jeremy Silvester
Date Added: 25 Sep 2020 13:49
Last Modified: 07 Sep 2023 10:40
URI: https://irep.ntu.ac.uk/id/eprint/40946

Actions (login required)

Edit View Edit View

Statistics

Views

Views per month over past year

Downloads

Downloads per month over past year