Abstract

Traditional Network-on-Chip (NoC) systems comprised of many cores suffer from debilitating bottlenecks of latency and significant power dissipation due to the overhead inherent in multi-hop communication. In addition, these systems remain vulnerable to malicious circuitry incorporated into the design by untrustworthy vendors in a world where complex multi-stage design and manufacturing processes require the collective specialized services of a variety of contractors. This thesis proposes a novel small-world tree-based network-on-chip (SWTNoC) structure designed for high throughput, acceptable energy consumption, and resiliency to attacks and node failures resulting from the insertion of hardware Trojans. This tree-based implementation was devised as a means of reducing average network hop count, providing a large degree of local connectivity, and effective long-range connectivity by means of a novel wireless link approach based on carbon nanotube (CNT) antenna design. Network resiliency is achieved by means of a devised adaptive routing algorithm implemented to work with TRAIN (Tree-based Routing Architecture for Irregular Networks). Comparisons are drawn with benchmark architectures with optimized wireless link placement by means of the simulated annealing (SA) metaheuristic. Experimental results demonstrate a 21% throughput improvement and a 23% reduction in dissipated energy per packet over the closest competing architecture. Similar trends are observed at increasing system sizes. In addition, the SWTNoC maintains this throughput and energy advantage in the presence of a fault introduced into the system. By designing a hierarchical topology and designating a higher level of importance on a subset of the nodes, much higher network throughput can be attained while simultaneously guaranteeing deadlock freedom as well as a high degree of resiliency and fault-tolerance.

Library of Congress Subject Headings

Networks on a chip--Design and construction; Networks on a chip--Security measures; Computer systems--Energy conservation

Publication Date

2-1-2013

Document Type

Thesis

Department, Program, or Center

Computer Engineering (KGCOE)

Advisor

Yang, Jay

Advisor/Committee Member

Kwasinski, Andres

Comments

Note: imported from RIT’s Digital Media Library running on DSpace to RIT Scholar Works. Physical copy available through RIT's The Wallace Library at: TK5105.546 .B46 2013

Campus

RIT – Main Campus

Plan Codes

CMPE-MS

Download

Share

COinS