Title :
A low-cost processing element recovery mechanism for fault tolerant Networks-on-Chip
Author :
Latif, Khalid ; Rahmani, Amir-Mohammad ; Seceleanu, Tiberiu ; Tenhunen, Hannu
Author_Institution :
Univ. of Turku, Turku, Finland
Abstract :
A fault in one component of Networks-on-Chip (NoC) based system makes the fault-free connected units out of use and this in turn leads to considerable performance degradation. Many fault tolerant architectures and routing algorithms have already been proposed for NoC but the utilization of resources, affected indirectly by faults is yet to be addressed. It is indispensable step needed to be taken in order to implement the reliable on-chip systems especially with nano-scale technologies. In this paper, we present a technique to recover healthy processing elements for NoC architectures in case of associated routers failure by using the Partial Virtual-Channel Sharing (PVS) approach. The proposed architecture divides the network into cluster regions, where each cluster comprises of two nodes. Each node in a cluster provides a backup data-path for other node in the cluster. Each processing element can use the backup data-path to transmit and receive the packets in case of corresponding router failure. The simulation results show that the proposed architecture has low hardware overheads.
Keywords :
fault tolerance; integrated circuit reliability; network routing; network-on-chip; NoC-based system; PVS approach; cluster regions; fault tolerant network-on-chip; fault-free connected units; hardware overheads; low-cost processing element recovery mechanism; nanoscale technologies; on-chip system reliability; partial virtual-channel sharing approach; router failure; routing algorithm; Detectors; Reliability; Switches; Very large scale integration;
Conference_Titel :
NORCHIP, 2011
Conference_Location :
Lund
Print_ISBN :
978-1-4577-0514-4
Electronic_ISBN :
978-1-4577-0515-1
DOI :
10.1109/NORCHP.2011.6126734