In this paper, we propose three Deflection-Routing-based Multicast (DRM) schemes for a bufferless NoC. The DRM scheme without packets replication (DRM_noPR) sends multicast packet through a non-deterministic path. The DRM schemes with adaptive packets replication (DRM_PR_src and DRM_PR_all) replicate multicast packets at the source or intermediate node according to the destination position and the state of output ports to reduce the average multicast latency. We also provide fault-tolerant supporting in these schemes through a reinforcement-learning-based method to reconfigure the routing table to tolerate permanent faulty links in the network. Simulation results illustrate that the DRM_PR_all scheme achieves 41%, 43% and 37% less latency on average than that of the DRM_noPR scheme and 27%, 29% and 25% less latency on average than that of the DRM_PR_src scheme under three synthetic traffic patterns respectively. In addition, all three fault-tolerant DRM schemes achieve acceptable performance degradation at various link fault rates without any packet lost.

An important issue in the realization of optical packet-switched (OPS) networks is the resolution of packet contention caused by the lack of RAM-like optical buffering. Although an optical buffer using fiber delay lines (FDLs) has been proposed, its capacity is extremely limited. There have been several studies of this problem. One approach is deflection routing, which is widely used in electronic packet-switched networks or optical burst-switched (OBS) networks. However, in OPS networks, packet lengths are short, so that the speed requirement for route lookup is very stringent. If the network topology is geometric, such as a Manhattan Street Network (MSN), hop-by-hop routing can be implemented by simple optical logic devices without an electronic routing table. However, if the topology is not geometric, it is hard to implement deflection routing electronically or optically. Another approach is reflection routing, which is easy to implement but has a higher probability of packet loss than does deflection routing. In this paper, we propose a packet contention resolution scheme, reflection-based deflection routing, which is based on reflection routing and enables switching the reflected packet to an alternate path if its primary path remains congested. Our method alleviates the time limitation on setting an alternate path by making use of the packet reflection latency and also reduces the probability of packet loss. We evaluate the performance of the proposed method by simulation experiments and show its effectiveness.

Deflection routing is one of the promising approaches to resolve contention in the optical burst switching networks. In the conventional deflection routing scheme, optical bursts may be unable to traverse the route evaluated to select an outgoing link because of the contention at succeeding downstream transit nodes. As a result, the optical bursts may traverse a different route resulting in a long distance or decreased performance. This paper proposes a deflection routing scheme that considers the possibility of the contention at downstream nodes. This scheme utilizes the "expected route distance" instead of the static route distance toward a destination node. The expected route distance considers the possibility of contention at each downstream transit node and is calculated using measured link blocking probabilities at each downstream transit node. The selection priority of each outgoing link is given dynamically based on its expected route distance toward a destination node. By considering the possibility of contention at downstream nodes, a routing scheme with high performance can be realized. The loss rate of optical bursts is improved when an imbalanced load is applied to the network, and the loss rate of optical bursts is also improved when the network includes links with extremely different distances.

Optical Burst Switching (OBS) is one of the most important switching technologies in future optical Internet. One of critical design issues in OBS is how to reduce burst dropping resulting from resource contention. Especially when traffic load is high, there should be frequent deflection routing as well as more contentions in an optical burst-switched network. The burst loss performance can be improved by implementing a proper deflection routing scheme. In this paper, we propose a limited deflection routing scheme to prevent injudicious deflection routing. The proposed scheme reduces unnecessary contentions resulting from deflection routing itself, increasing the utilization of network resource such as channels. Simulation tests were performed to evaluate the performance of the proposed scheme.

We propose an efficient, low cost, multicast ATM switch which is fair to all inputs. The switch consists of a novel copy network which creates unicast packets in a fair manner, followed by a network that routes packets to their correct Address Translation Tables (ATT's) and ultimately a unicast routing network which ensures sequencing. The copy and routing networks are based on deflection routing. We show that our switch requires O(log N) stages and can be designed for any arbitrarily low level of packet loss. The theoretical results are backed up by simulations. Switching elements in both the copying and routing networks have O(1) bit complexity, making the overall bit level hardware complexity of the network O(N log N). The latency of the switch is proportional to the number of stages O(log N). Unlike other existing copy networks, our copy network drops packets in a fair manner and hence can provide quality of service (QoS) support. The switch is output queued and allows the delivery of multiple packets to the same destination during a time slot.

In this paper, we investigate the performance characteristics of parallel switching architectures constructed by a stack of multistage switching networks. We first find that the performances of the previously proposed parallel switching architectures are much worse than the expected ones from analytic models which are based on the assumption that traffic is uniformly distributed at each stage of a switching network. We show that this phenomenon is closely related to a traffic-distribution capability of a parallel switching system and has a large influence on the performance. From these results, we then propose an architectural solution based on the Generalized Shuffle Network (GSN) and analyze its performance by proposing a new iterative analysis method. The proposed architecture uses self-routing and deflection routing, and inherently has a traffic-distribution capability to improve switch performances such as cell loss and delay in a cost-effective manner. From the comparison of simulation and analysis results, it is shown that the developed models are quite accurate in predicting the performance of a new parallel switching system.