A new adaptive fault-tolerant protocol for direct multiprocessors networks

  • Authors:
  • Omar Al-Jarrah;Hassan Tawalbeh

  • Affiliations:
  • Department of Computer Engineering, Jordan University of Science and Technology, Irbid, Jordan;Department of Computer Engineering, Jordan University of Science and Technology, Irbid, Jordan

  • Venue:
  • ICCOM'06 Proceedings of the 10th WSEAS international conference on Communications
  • Year:
  • 2006

Quantified Score

Hi-index 0.00

Visualization

Abstract

This paper investigates the fault tolerance problem in direct networks. Conservative flow control mechanisms such as Pipelined Circuit Switching (PCS) ensure the existence of a path to the destination before transmission. This ensures achieving reliable fault-tolerant system on the expense of performance. Optimistic flow control mechanisms such as Wormhole Switching (WS) realize very good performance, but they are very sensitive to the existence of deadlock of the presence of faults. This paper presents a new adaptive fault tolerant protocol in wormhole switched networks. In the fault free region, messages use WS for routing. However, when messages encounter faults, some restrictions are used with controlled buffering of the message in the node's local memory to ensure deadlock free fault-tolerance routing. The router and virtual channel models are simplified. Distributed random faulty nodes can be fully handled. A fully adaptive three-phase protocol is proposed and compared via simulation to the two phase (TP) protocol. The results show that the algorithm achieves higher throughput at different injection rates and lower latency over a wide range of injection rates.