.

ISSN 2063-5346
For urgent queries please contact : +918130348310

COMPARISON ANALYSIS OF DEADLOCK-FREE WORMHOLE ROUTING FOR INTERCONNECTED MULTIPROCESSOR NETWORK

Main Article Content

Kamalesh Chandra Maurya1 and Dr. Pankaj Kumar2
» doi: 10.48047/ecb/2023.12.10.752

Abstract

Interconnected multiprocessor networks (IMNs) play a pivotal role in modern parallel computing systems, facilitating efficient communication among processors and minimizing latency. Wormhole routing has emerged as a prominent mechanism to manage data transfer in these networks, offering lower latency and higher bandwidth utilization compared to other routing techniques. However, wormhole routing is susceptible to deadlocks, a critical concern that can disrupt the overall system performance. This research presents a comprehensive comparison analysis of various deadlock-free wormhole routing algorithms used in interconnected multiprocessor networks. The objective is to evaluate their respective strengths and weaknesses, aiding system designers in making informed decisions when selecting the most suitable routing strategy for their specific application. The study commences with an indepth review of deadlock scenarios and the challenges associated with wormhole routing. It delves into the theoretical foundations of deadlock avoidance and resolution algorithms, including virtual channels, path-based, and adaptive routing techniques. Each algorithm's mechanisms and overheads are analyzed to assess their suitability in diverse IMN topologies. To gauge the performance of these deadlock-free wormhole routing algorithms, a simulation framework is developed, encompassing various traffic patterns and network scales. Key metrics, such as, and network congestion, are employed to measure the algorithms' efficacy under both uniform and non-uniform traffic loads. The experimental results reveal the distinct advantages and limitations of each algorithm under specific scenarios. Some algorithms excel in uniform traffic environments, while others demonstrate superior performance in dealing with varying traffic patterns or large-scale network configurations.

Article Details