LORE - local reconfiguration for fault management in irregular interconnects

Summary form only given. Dynamic faults in irregular networks are difficult to handle when discarding uncorrupted packets is undesirable, deadlocks must be avoided, and in-order delivery of packets must be maintained. Current reconfiguration approaches involve all or many nodes to handle a fault. We introduce the LORE approach, a dynamic local reconfiguration which allows most of the network to remain oblivious of a one-fault. Only a small subnetwork, consisting of fault adjacent nodes and some nodes and links to connect these, is affected. LORE can be adapted to several topologies and several routing algorithms without introducing deadlocks, and can maintain in-order delivery in deterministic routing.

[1]  Olav Lysne,et al.  FROOTS - Fault Handling in Up*/Down* Routed Networks with Multiple Roots , 2003, HiPC.

[2]  Michael Burrows,et al.  Autonet: A High-Speed, Self-Configuring Local Area Network Using Point-to-Point Links , 1991, IEEE J. Sel. Areas Commun..

[3]  Antonio Robles,et al.  Improving the Up*/Down* Routing Scheme for Networks of Workstations , 2000, Euro-Par.

[4]  José Duato,et al.  Performance evaluation of dynamic reconfiguration in high-speed local area networks , 2000, Proceedings Sixth International Symposium on High-Performance Computer Architecture. HPCA-6 (Cat. No.PR00550).

[5]  Olav Lysne,et al.  Deadlock avoidance for switches based on wormhole networks , 1999, Proceedings of the 1999 International Conference on Parallel Processing.

[6]  William J. Dally,et al.  Deadlock-Free Message Routing in Multiprocessor Interconnection Networks , 1987, IEEE Transactions on Computers.

[7]  Charles L. Seitz,et al.  Myrinet: A Gigabit-per-Second Local Area Network , 1995, IEEE Micro.

[8]  Sudhakar Yalamanchili,et al.  Interconnection Networks: An Engineering Approach , 2002 .

[9]  José Duato,et al.  The double scheme: deadlock-free dynamic reconfiguration of cut-through networks , 2000, Proceedings 2000 International Conference on Parallel Processing.

[10]  Suresh Chalasani,et al.  Fault-tolerant routing with non-adaptive wormhole algorithms in mesh networks , 1994, Proceedings of Supercomputing '94.

[11]  Antonio Robles,et al.  Efficient Adaptive Routing in Networks of Workstations with Irregular Topology , 1997, CANPC.

[12]  José Duato,et al.  Fast dynamic reconfiguration in irregular networks , 2000, Proceedings 2000 International Conference on Parallel Processing.

[13]  Olav Lysne,et al.  Load Balancing of Irregular System Area Networks through Multiple Roots , 2001 .