Date of Submission


Date of Award


Institute Name (Publisher)

Indian Statistical Institute

Document Type

Doctoral Thesis

Degree Name

Doctor of Philosophy

Subject Name

Computer Science


Advance Computing and Microelectronics Unit (ACMU-Kolkata)


Sinha, Bhabani Prasad (ACMU-Kolkata; ISI)

Abstract (Summary of the Work)

A recent trend in computing is to distribute the computations among a set of processing elements. There are two basic appronches to this - one is to build a loosely-coupled system and the other is to form a tightly-coupled system [PS85].In a loosely-coupled system, the processors do not share common memory or a common clock; but sharing of important resources like data files, softwares, special hardware components etc., is possible without duplicating the resources themselves. The processing nodes may even be geographicully separated from each other and are connected through databuses, telephone/radio links, satellite, etc. Such loosely- coupled systems are also commonly referred to as distributed systems.In a tightly-coupled system, the processors share a common clock and/or common memory resulting in a parallel processing environment. Such a system enables one to meet the requirement for enormous amount of fast real-time computations in many practical applications, e.g., weather forecasting, image processing, etc. Examples of parallel processing systems include pipelined computers, array processors and multiprocessors. A distributed system or a parallel processing system involving array processors requires an interconnection network for communication among the processing elements.Interconnection Networks may be classified into two broad categories [Fe81]. They are :i) static interconnection networkii) dynamic interconnection network.and A static interconnection network is a collection of processors and links among them. Such an interconnection network is usually represented by a graph. The nodes of the graph represent the processing elements and the links of the graph stand for the communication links. This graph is usually called the network graph or the network topology. A dynamic interconnection network has a set of sources and a set of destinations (may be the same as the set of sources) which are connected through switches and links. Different input-output connection patterns may be achieved by changing switch-settings.The desirable properties of a good network topology are : i) Low degree of nodes : The upper bound on the degree is, in practice, restricted by the number of 1/0 ports that can be provided by a node.ii) Low number of links : This helps in reducing the cost of the interconnection.iil) Low diameter : This is to reduce the inter-node communication time. iv) High degree of fault-tolerance.v) Regular structure : This is for ensier implementation in Incremental extensibility andvii) Simple routing algorithms in both fnult-free and faulty conditions.Some of these properties are mutually conflicting. For example, the third property is in conflict with the first two. The fourth one in the above list is impeded by the previous three. Some of the problems of recent research interests in the area of static networks are as follows :1. Design of a network topology : An integrated design approach, simultaneously optimizing all of these aspects, is very difficult. The usual practice is to consider one or some of them, but not all at a time, to arrive at an optimal or near-optimal design. Many such topologies, e.g., tree, mesh (Le93], de Bruijn [Br46], Möbius [LS82], hypercube [Já92], cube-connected-cycle [IV81], double-loop [DHL90] etc.


ProQuest Collection ID:

Control Number


Creative Commons License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.


Included in

Mathematics Commons