Self organization on a swarm computing fabric: a new way to look at fault tolerance

  • Authors:
  • Danilo Pani;Simone Secchi;Luigi Raffo

  • Affiliations:
  • University of Cagliari, Cagliari, Italy;University of Cagliari, Cagliari, Italy;University of Cagliari, Cagliari, Italy

  • Venue:
  • Proceedings of the 7th ACM international conference on Computing frontiers
  • Year:
  • 2010

Quantified Score

Hi-index 0.00

Visualization

Abstract

Recent studies have demonstrated the possibility to exploit Swarm Intelligence (SI) as an inspiration for the design of scalable VLSI tiled architectures exhibiting multitasking, adaptability, absence of centralized low-level control and fault-tolerance. SI approach to fault-tolerance, in principle, can be regarded as a reconfiguration-free cell-exclusion mechanism. The key elements at the basis of a reconfiguration free solution are: loose structure of the system, homogeneity, cooperative behaviors and self organization. In this paper, these self organization aspects, introduced in a recently developed multi-agent VLSI tiled architecture for array processing, expressly developed resorting to the SI inspiration, are presented along with some theoretical and experimental results. The architecture presents two forms of cell-exclusion (bypass and block of faulty elements), implementing self-adaptive behaviors rather than reconfiguration to face faults preserving system functionality. The proposed approach, exploiting indirect communications to provide workload spreading into the computing fabric, is also successful in reducing the effects of the presence of faulty elements without spare resources and with limited performance degradation.