Deflection routing is a feasible approach to resolve the output contention problem in packet-switched networks when buffering of packets is not practical. In this paper, we investigate the performance of multichannel deflection-routed networks with no packet injection control, strict packet injection control, and a simple token-bucket-based packet injection control. The analytical performance models of multichannel deflection-routed networks with strict packet injection control are derived. Simulation results show that the analytical models can accurately predict the performance regardless of the network topology, number of channels, and packet injection control methods. We observed that the end-to-end throughput-delay and the packet re-transmission performance at sources can be largely improved by using simple packet injection control mechanisms such as the proposed token-bucket-based method.

With the recently proposed multi-slot batch-transfer (MSBT) architecture, we can build optical packet switches using slow switching fabrics with reconfiguration time larger than the guard time between packets. Since MSBT switches can provide multichannel capability with no additional hardware, we propose to combine the multichannel and deflection routing approaches for packet contention resolution in MSBT networks. As there is no analytical performance model available, we derive the required model in this paper. Simulations show that the model is very accurate.

In optical burst switching (OBS) system, the throughput decreases rapidly with increase in control packet processing time Tcp. The negative impact of Tcpwill become significant as the optical fiber transmission rate increases. By analyzing the relationship between the throughput and Tcp, we attempt to improve the throughput. Different possible solutions are discussed. We found that using extra random offset time can significantly improve the throughput at the expense of increase in data burst delay.

In optical packet switching, the reconfiguration time of optical switches should only be a small fraction of the packet transmission time for efficient utilization of the bandwidth. This constraint puts a very stringent demand on the switch reconfiguration time as the transmission rate of optical fiber increases. By using batch transfer of packets, we propose an optical switch architecture that can significantly relax the requirement on switch reconfiguration time. The utilization of the transmission links is greatly improved because the guard time between packets is no longer related to the switch reconfiguration time. We have also demonstrated analytically that the proposed switch architecture can significantly reduce the packet loss probability, at the expense of an increase in the delay variations, if the packets are not restricted to their original time slots at the switch output.

In this paper, we investigate the integration of passive optical networks (PONs) with optical burst switched (OBS) networks. Owing to the decomposition of the operations between the PONs and OBS nodes, serious problems have been observed. One of them is data burst assembly problem. In some situations, burst assembly delay exceeding tens seconds may be required if the general purpose PON access schemes are used. To solve this problem, we propose to have the optical network units (ONUs) take over the burst assembly function from the OBS nodes. Large reduction of the burst assembly delay is obtained. Four ONU data burst scheduling schemes have also been investigated. We observe that a simple scheduling scheme is itself adequate in most situations.

The optical burst switching (OBS) schemes to date assume that the switching overhead at intermediate nodes is either negligible or can be considered as part of the processing delay of the control packet. In this paper, we will show that the switching overhead can have a significant impact on the performance of OBS. We have also proposed methods to alleviate the problem.

We propose to improve the throughput performance of optical burst switching by using regional controller nodes and window-based reservation. Both methods increase the information available to the intermediate nodes during scheduling decisions. Simulations show that the proposed reservation schemes provide significant improvement in the throughput performance compared with the original optical burst switching when the network is heavily loaded.

Optical time division multiplexing (OTDM) is an effective approach to improve the performance of wavelength-routed (WR) networks. Implementation of all-optical time-slot routing in OTDM-WR networks is difficult owing to the lack of practical optical buffer and sophisticated optical processing devices. In this paper, a wavelength-switched time-slot routing scheme that can be implemented with fast wavelength converters only is proposed. Simulation results demonstrate that the performance of the proposed scheme is much better than that of WR networks and is comparable to OTDM-WR networks with time-shared space switching.

A deflection routing scheme for small to medium size future all-optical networks with arbitrary topologies is proposed. The proposed scheme assumes only single-bit all-optical processing and no buffers. The primary output selection and the alternate output choices by a packet at each node are encoded in the packet header in order to reduce the signal processing requirement. Additional features such as priority and time-to-live fields have also been defined. The performance of the deflection routing scheme is studied using the AT&T North America OC-48 optical fiber network topology.

Two general-purpose decomposition methods to calculate the blocking probabilities of connection-oriented networks are presented. The methods are based on either the call status or the link status of the networks, and can significantly reduce the required computational times. A heuristic is presented to simplify the application of the proposed decomposition methods on networks with irregular topologies. Numerical examples are given to demonstrate the applications of the proposed methods.

MEMS optical switch technology is one of the key technologies in wavelength division multiplexing (WDM) optical networks. Although the 2-D MEMS optical switch technology is mature, the commonly used crossbar architecture is not amenable to building large switches. In this paper, we propose a design of 2x2 switching modules, and use it to build large 2-D MEMS optical switches with architectures such as Spanke-Benes and Benes networks.

Two multicast protocols are proposed for deflection-routed all-optical packet-switched networks. One scheme sends a deflected multicast packet back to the root node while the other sends it back to the deflection point. Both schemes can be implemented using demonstrated optical signal processing technology. The performance of the two proposed multicast schemes are compared using Manhattan street networks. We found that the back-to-the-root-node scheme performed better than the back-to-the-deflection-node scheme. A hybrid approach can further improve the system performance.

A deflection routing algorithm that can be applied to a novel self-routing address scheme for networks with arbitrary topology is proposed. The proposed deflection routing algorithm can be implemented all-optically using bitwise optical logic gates. Besides the primary output link selection, alternate output link choices by a packet at each node in case of deflection are also encoded in the address header. Priority classes can also be defined in the proposed address scheme. The performance of the deflection routing algorithm is studied using the AT&T North America OC-48 optical fiber network topology.

Due to limited available photonic devices, optical networks in the near future will likely employ routing schemes that do not require sophisticated processing of optical packets. In this paper, we propose a novel self-routing scheme for all-optical packet networks that can be applied to networks with arbitrary topology. The proposed routing scheme requires only single bit processing and can be implemented with existing technologies.