R. Libeskind-Hadas, K. Watkins, and T. Hehre, "FaultTolerant Multicast Routing in the Mesh with No Virtual Channels," Proc. 2nd Intl. Symp. High-Perf. Comp. Arch., Feb. 3-7, 1996, pp. 180-190.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....paths (e.g. as in [12] This requires a network with connectivity (M) 2E arb max 1, which limits the degree of fault tolerance to arbitrary RP faults to E arb max , b( M) Gamma 1) 2c. If RP faults are restricted to crash RP faults, one can use adaptive fault tolerant multicast protocols [9, 10]. These will employ O( M) fewer messages than protocols using multiple node disjoint paths. In most cases, E crash max , M) Gamma 1 crash RP faults can be tolerated. In addition, adaptive protocols that handle congestion may also tolerate faults such as RPs that relay messages too slowly. ....

R. Libeskind-Hadas, K. Watkins, and T. Hehre, "FaultTolerant Multicast Routing in the Mesh with No Virtual Channels," Proc. 2nd Intl. Symp. High-Perf. Comp. Arch., Feb. 3-7, 1996, pp. 180-190.

....the recently proposed path based solutions, by enhancing routers with multi destination capability [6, 13] Examples include This research is supported by the National Science Council of the Republic of China under Grant # NSC86 2213 E 008 029 and Grant # NSC86 2213 E 216 021. 6, 13] for meshes, [12, 7, 11, 15] for k ary n cubes, and [12, 16, 18] for arbitrary networks. Simple hardware is added to the router to enable it to copy the content of a worm while forwarding the worm to the next router. The header of such a worm can carry a number of destination addresses for the worm to visit. So such worms ....

....by enhancing routers with multi destination capability [6, 13] Examples include This research is supported by the National Science Council of the Republic of China under Grant # NSC86 2213 E 008 029 and Grant # NSC86 2213 E 216 021. 6, 13] for meshes, 12, 7, 11, 15] for k ary n cubes, and [12, 16, 18] for arbitrary networks. Simple hardware is added to the router to enable it to copy the content of a worm while forwarding the worm to the next router. The header of such a worm can carry a number of destination addresses for the worm to visit. So such worms are also termed as multidestination ....

[Article contains additional citation context not shown here]

R. Libeskind-Hadas, K. Watkins, and T. Hehre. Fault-tolerant multicast routing in the mesh with no virtual channels. In HighPerformance Computer Arch. Conf., pages 180--190, 1996.

....such approach tends to have communication deadlock if multiple multicasts are performed simul4 taneously. One major advance in solving this problem is the recently proposed pathbased solutions, by enhancing routers with multi destination capability [11, 20] Examples include [11, 20] for meshes, [19, 12, 17, 23] for k ary n cubes, and [19, 24, 30] for arbitrary networks. Simple hardware is added to the router to enable it to copy the content of a worm while forwarding the worm to the next router. The header of such a worm can carry a number of destination addresses for the worm to visit. So such worms ....

....deadlock if multiple multicasts are performed simul4 taneously. One major advance in solving this problem is the recently proposed pathbased solutions, by enhancing routers with multi destination capability [11, 20] Examples include [11, 20] for meshes, 19, 12, 17, 23] for k ary n cubes, and [19, 24, 30] for arbitrary networks. Simple hardware is added to the router to enable it to copy the content of a worm while forwarding the worm to the next router. The header of such a worm can carry a number of destination addresses for the worm to visit. So such worms are also termed as multidestination ....

[Article contains additional citation context not shown here]

R. Libeskind-Hadas, K. Watkins, and T. Hehre. Fault-tolerant multicast routing in the mesh with no virtual channels. In High-Performance Computer Arch. Conf., pages 180--190, 1996.

....as message authentication and atomicity need not be provided. Possible 5 implementations of a reliable multicast service include sending multiple copies of a message along node disjoint paths (e.g. as in [6, 22, 24, 25] or relying on an NCCN with adaptive fault tolerant routing capabilities [14, 15, 17]. As discussed in Subsection 2.6, choice of a particular type of reliable multicast service depends on the class of RP faults that must be tolerated. In each step of m ICV, disjoint subsets of TSPs in T use multicasts to implement an a clock reading method. The goal of such a reading method is to ....

....as in [6, 22, 24, 25] This requires a network with connectivity (M) 2E arb max 1, which limits the degree of fault tolerance to arbitrary RP faults to E arb max , b( M) Gamma 1) 2c. If RP faults are restricted to crash RP faults, one can use adaptive fault tolerant multicast protocols [14, 15, 17]. These will employ O( M) fewer messages than protocols using multiple node disjoint paths. In most cases, E crash max , M) Gamma 1 crash RP faults can be tolerated. In addition, adaptive protocols that handle congestion may also tolerate faults such as RPs that relay messages too slowly. ....

R. Libeskind-Hadas, K. Watkins, and T. Hehre, "Fault-Tolerant Multicast Routing in the Mesh with No Virtual Channels," Proceedings of the 2nd International Symposium on High-Performance Computer Architecture, February 3-7, 1996, San Jose, CA, pp. 180-190.

....and combined unicast multicast traffic conditions. 2 Related Work Recent research on hardware supported multicast in wormhole routed networks can be broadly divided into path based and tree based approaches. Path based approaches either construct a single path spanning all nodes of the network [1, 4, 7] or use an underlying deadlock free unicast routing algorithm to construct multiple paths which cover the destinations of the multicast [6] One drawback of the path based approaches is that the path lengths can be extremely long, especially in the former approach, and this leads to high latency ....

R. Libeskind-Hadas, K. Watkins, and T. Hehre, "Fault-Tolerant Multicast Routing in the Mesh with No Virtual Channels," Proceedings of the IEEE Symposium on High-Performance Computer Architecture, pp. 180--190, 1996.

....e.g. Panda, et al. 15, 18] In the former approach, path lengths can be extremely long while in the latter approach, multiple message startups are required and this may cause worm transmissions to be serialized. Only a few previous papers have addressed the issue of fault tolerant multicast [3, 10, 16]. Most of this work is based on construction of single paths in a network but one of the algorithms of [3] uses multiple paths. In addition to the problems cited above with path based multicast, other limitations of previous approaches include the need for nodes to have global fault information ....

....16] Most of this work is based on construction of single paths in a network but one of the algorithms of [3] uses multiple paths. In addition to the problems cited above with path based multicast, other limitations of previous approaches include the need for nodes to have global fault information [10, 16], the need to have multiple virtual data channels per physical link [3, 16] and the sharing of paths by distinct multicast messages in the network [10, 16] Simulation results have been published for the algorithms of Libeskind Hadas, et al. 10] and Panda, et al. 16] but not for the algorithms ....

[Article contains additional citation context not shown here]

R. Libeskind-Hadas, et al., "Fault-Tolerant Multicast Routing in the Mesh with No Virtual Channels," HPCA Symp. Proceedings, pp. 180--190, 1996.

....the improved performance achieved by our algorithm. 2 Related Work Recent research on hardware supported multicast in wormhole routed networks can be broadly divided into path based and tree based approaches. Path based approaches either construct a single path spanning all nodes of the network [1, 4, 7] or use an underlying deadlockfree unicast routing algorithm to construct multiple paths which cover the destinations of the multicast [6] One drawback of the pathbased approaches is that the path lengths can be extremely long, especially in the former approach, and this leads to high latency ....

R. Libeskind-Hadas, et. al, "Fault-Tolerant Multicast Routing in the Mesh with No Virtual Channels," HPCA'96, pp. 180--190.

....combine to reach all destinations, e.g. Panda, et al. 12] In the former approach, path lengths can be extremely long while in the latter approach, multiple message startups are required. To our knowledge, there have been only a few papers that have addressed the issue of faulttolerant multicast [3, 7, 13]. Most of this work is based on construction of single paths in a network [3, 7, 13] which can become extremely long for multicast messages. One of the algorithms of [3] uses multiple paths, which may be relatively short but require multiple message startups and, in some cases, can cause worm ....

....path lengths can be extremely long while in the latter approach, multiple message startups are required. To our knowledge, there have been only a few papers that have addressed the issue of faulttolerant multicast [3, 7, 13] Most of this work is based on construction of single paths in a network [3, 7, 13] which can become extremely long for multicast messages. One of the algorithms of [3] uses multiple paths, which may be relatively short but require multiple message startups and, in some cases, can cause worm transmissions to be serialized. Other limitations of previous approaches include the ....

[Article contains additional citation context not shown here]

R. Libeskind-Hadas, K. Watkins, and T. Hehre, "Fault-Tolerant Multicast Routing in the Mesh with No Virtual Channels," Proceedings of the IEEE Symposium on High-Performance Computer Architecture, pp. 180--190, 1996.

....great deal of recent research has investigated how to implement deadlock free multicast efficiently with wormhole routing. Most of this work has focused This research was supported by the National Science Foundation under Grant CCR 9803741. on path based approaches that use multidestination worms [8, 11, 14, 16]. Path based approaches suffer from either extremely long paths (leading to high latency and large latency variation) or multiple costly message startups. Tree based multicast algorithms [5, 9] use up down routing. A spanning tree is constructed in the network and all links are classified into ....

R. Libeskind-Hadas, et al., "Fault-Tolerant Multicast Routing in the Mesh with No Virtual Channels," Proc. HPCA '96, pp. 180--190.

....such as message authentication and atomicity need not be provided. Possible implementations of a reliable multicast service include sending multiple copies of a message along node disjoint paths (e.g. as in [7, 27, 29, 30] or relying on an NCCN with adaptive fault tolerant routing capabilities [18, 19, 20]. As discussed in Subsection 3.6, choice of a particular type of reliable multicast service depends on the class of RP faults that must be tolerated. In each step of m ICV, disjoint subsets of TSPs in T use multicasts to implement an a clock reading method. The goal of such a reading method is to ....

....as in [7, 27, 29, 30] This requires a network with connectivity (M) 2E arb max 1, which limits the degree of fault tolerance to arbitrary RP faults to E arb max , b( M) Gamma 1) 2c. If RP faults are restricted to crash RP faults, one can use adaptive fault tolerant multicast protocols [18, 19, 20]. These will employ O( M) fewer messages than protocols using multiple node disjoint paths. In most cases, E crash max , M) Gamma 1 crash RP faults can be tolerated. In addition, adaptive protocols that handle congestion may also tolerate faults such as RPs that relay messages too slowly. ....

R. Libeskind-Hadas, K. Watkins, and T. Hehre, "Fault-Tolerant Multicast Routing in the Mesh with No Virtual Channels," Proceedings of the 2nd International Symposium on High-Performance Computer Architecture, pp. 180--190, 1996.

....then vertex v is added to the cycle as shown in Figure 9(e) The result of applying this step to the cycles in Figure 10(e) is shown in Figure 10(f) The last remaining case is that all edges (u i ; u (i 1)mod8 ) are used. The analysis for this case is rather involved and the reader is referred to [15] for details. However, in over 100; 000 fault patterns generated by our simulator for each of 16 Theta 16, 32 Theta 32, and 64 Theta 64 meshes, this case never arose. Finally, since this algorithm considers each node only a finite number of times, the total time to construct the ....

R. Libeskind-Hadas, K. Watkins, and Tom Hehre. Fault-tolerant multicast routing in the mesh with no virtual channels. Technical Report HMCCS -95-03, Harvey Mudd College, Department of Computer Science, June 1995.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC