A message controller for distributed processing systems

• Main Author: Wong, Kar Leong
• Format: Dissertation
• Language: English
• Published: Nottingham Trent University 2000
• Subjects: Interface; Network

This thesis presents a novel message-passing hardware network interface controller which has been designed and developed for integration into a router interconnection network for distributed parallel processing systems. It describes how the network interface controller can improve the efficiency of message transfers in packet switching based communications and reduce the overheads incurred to microprocessor tasks. Initially, a review focusing on the architectures and techniques of interfacing inter-processor interconnection networks to parallel processor computing nodes was carried out. Various parallel processing system packet routing devices and network interface controllers have been investigated. Following the review, a novel network interface controller was designed, to link a microprocessor node to a parallel processor system, interconnection network. The network interface controller design was captured in a Hardware Description Language (VHDL) following a top-down design methodology. A series of comprehensive tests were written to verify the functionality of the design model. The design was synthesised into a target programmable logic device, tested via a working prototype processor node incorporating the StrongARM SA-110 microprocessor. This was followed by the construction of a distributed parallel processing system using an ICR C416 packet routing interconnection network. The successful implementation has demonstrated how an efficient interprocessor communication can be achieved using the network controllers to link to an ICR C4I6 packet routing network to StrongARM microprocessor nodes. This offers the processing power of high performance microprocessors in an embedded distributed parallel processing system. Key features of the system incorporating the network controller are discussed and the system is compared and contrasted with the state-of-the-art in parallel processing communication networks.