Optical burst switching (OBS) is widely believed to be the technology for the future core network in the Internet. Traditionally, the burst header packet (BHP) is served by the control processor on first-come-first-serve basis which increases the burst loss probability (BLP) at the core nodes closer

An important issue of research in optical burst switching (OBS) networks is to minimize the loss of bursts due to contention at the intermediate nodes. These contention losses can be minimized with the design of efficient path selection algorithms at the ingress node. Path selection algorithms that learn the optimal path dynamically with the changing traffic conditions outperform the deterministic path selection algorithms. Usually in the single agent path selection algorithms, a path is selected by the agent based on the feedback received at the ingress node which does not capture the effect of the paths selected by the other nodes in the network. We develop a multi-agent approach for path selection that includes the effect of the selection made by all the other nodes in the network. The proposed path selection algorithm uses agents at different source nodes to collectively learn the network dynamics and select the best outgoing path for each burst. We present simulation results to demonstrate the effectiveness of the proposed algorithm over the other similar algorithms in the literature.

Contention losses which usually do not indicate congestion is a major issue that hinders the deployment of optical burst switching (OBS) networks. Development of efficient path and wavelength selection algorithms is crucial to minimize the burst loss probability (BLP) in OBS networks. In this paper, we handle path selection and wavelength selection in a joint fashion. We formulate the problem of selecting a pair of path and wavelength jointly as a multi-armed bandit problem (MABP) and discuss the difficulties in solving MABP directly. We then rewrite the Q-learning formalism to solve the MABP without explicit model in an online fashion and propose an algorithm to solve the problem near-optimally. The proposed algorithm selects a pair of path and wavelength at each ingress node to minimize the BLP on the long run. Simulation results demonstrate the effectiveness of our algorithm in minimizing the BLP with better link utilization compared to the other proposals in the literature.

Optical burst switching (OBS) is considered as a contending technology for the core of the Internet in future. However, due to lack of the buffers, losses occur due to contention among simultaneously arriving bursts at the core nodes. Contention losses do not necessarily indicate a situation of congestion in the network. Thus differentiation (classification) of losses is essential in many applications to avoid false identification of congestion. In this paper, we propose a loss classification technique for the OBS networks based on machine learning techniques. We devise a new measure to differentiate between congestion and contention losses, which is derived from the observed losses, called the number of bursts between failures (NBBF). We observe that the NBBF follows a Gaussian distribution with different parameters for contention and congestion losses. This feature is used in differentiation. We use both a supervised learning technique (hidden Markov model (HMM)) and an unsupervised learning technique (expectation maximization (EM) clustering) on the observed losses and classify them into a set of states (clusters) after which an algorithm differentiates between the congestion and contention losses. We also demonstrate the use of loss differentiation in improving the performance of transport control protocol (TCP) over OBS networks. We modify congestion control mechanism of TCP suitably to arrive at two variants of TCP, HMM-TCP and EM-TCP. Their performance is compared with TCP NewReno, TCP SACK, and Burst TCP (X. Yu et al., Mar. 2004). Simulation results demonstrate the effectiveness and accuracy of the loss classification technique in different network scenarios.

It is widely accepted that retransmitting bursts in Optical Burst Switching networks improves the throughput at higher layers. Since the data is transported in large bursts and there are burst losses due to contention, indiscriminate retransmission can defeat the purpose of burst-level retransmissions. For many applications, retransmission after a certain time would be of no use and it may not be necessary to retransmit all the lost packets. We propose a framework for retransmission with parameters to control the retransmission rate and thus the increase in the network load. We propose a network model for controlled retransmission and a modified functional architecture of the ingress node. The existing work lacks an accurate analysis to estimate the impact of retransmissions on the network load while studying the improvement in the end-to-end packet recovery. We provide theoretical analysis to evaluate the load at each node due to both fresh and retransmitted bursts. We propose some metrics to quantify the benefit of retransmission and the impact of the retransmission parameters proposed on the network performance. We show that with controlled retransmission, the buffer requirement at the ingress node is proportional to the steady-state rate of retransmitted bursts. We validate the analytical model arid also study the impact of retransmission on the network performance with extensive simulations.

Optical burst switching (OBS) is envisioned as the paradigm for a high-bandwidth Internet. Losses due to contention is a serious problem in OBS networks. Since the core nodes have limited capabilities, loss rate cannot be determined at the nodes. Thus, estimating the losses on a path purely based on end-to-end measurements and thereafter develop proactive measures for loss reduction is an attractive option. For the first time, we apply a tomographic technique that can estimate the losses at the core nodes from end-to-end measurement between the edge nodes. We use passive unicast tomography to minimize the network overhead. We model the problem of estimating loss rate at the nodes from path-level measurements as a maximum likelihood problem and solve it using the expectation-maximization algorithm. In simulations, we use a multiple source, multiple destination embedding on the NSFNET topology and observe losses on multiple paths to infer losses at the core nodes. The estimated losses are found to match closely with actual losses measured.

Study of the performance of transmission control protocol (TCP) over optical burst switching (OBS) networks has been an important problem of research lately. In this work, we propose an analytical model for a TCP NewReno source to derive the steady-state throughput in presence of burst assembly process and burst losses. The source model uses a Markov chain based evolution of congestion window and the network model characterizes the distribution of burst size for a general assembly process which is used to estimate the impact of a burst loss on the number of packets lost. A fixed-point iteration method is then used to jointly solve the source model and the network model to obtain the TCP send rate. We validate the proposed analytical model through simulations. Results highlight the importance of accounting for slow start and fast retransmit phases in the model.

Optical burst switching (OBS) is a promising technology that exploits the benefits of optical communication and supports statistical multiplexing of data traffic at a fine granularity making it a suitable technology for the next generation Internet. Contention among the bursts that arrive simultaneously at a core node leads to burst loss which affects the throughput of higher layer traffic. Development of efficient algorithms for path selection and wavelength selection is crucial to minimize the burst loss probability (BLP) in OBS networks. In this paper, we formulate path selection and wavelength selection in OBS networks as a multi-armed bandit problem and discuss the difficulties to solve them optimally. We propose algorithms based on Q-learning to solve these problems near-optimally. At an egress node, the path selection algorithm evaluates the Q values for a set of precomputed paths and chooses a path that corresponds to minimum BLP. Similarly, Q-learning algorithm for wavelength selection selects a wavelength in a pre-routed path such that the BLP is minimized. We do not assume wavelength conversion and buffering at the core nodes and hence, selection of path and wavelength is done only at the edge nodes. We simulate the proposed algorithms under dynamic load to demonstrate that they reduce the BLP compared to the other adaptive algorithms available in the literature.

Optical burst switching (OBS) is a promising technology that exploits the benefits of optical communication and supports statistical multiplexing of data traffic at a fine granularity making it a suitable technology for the next generation Internet. Development of efficient algorithms for path selection and wavelength selection is crucial in minimizing the burst loss probability (BLP) in OBS networks. In this paper, we present novel Reinforcement Learning algorithms for path selection and wavelength selection in the context of OBS networks. We develop an online path selection algorithm based on Q-learning to minimize the BLP by choosing an optimal path among a set of predetermined routes between every pair of ingress and egress nodes. We also propose a Q-learning algorithm for wavelength selection that selects an optimal wavelength among the available wavelengths in a pre-routed path with an objective of minimizing the BLP. We assume no wavelength conversion and buffering to be available at the core nodes of the OBS network. We simulate the proposed algorithms under dynamic load to demonstrate that they reduce the BLP compared to the best known adaptive techniques for path selection and wavelength selection available in the literature.

In this paper, we discuss issues involved in developing autonomic Optical Burst Switched (OBS) networks. We develop an OBS network system, the first of its kind, which is self-aware, self-protecting, and self-optimizing, which are essential requirements of an autonomic network system. We use learning automata to autonomously learn the network state and make intelligent choices of route and wavelength, for burst transmission. We develop, for the first time, a self-protecting mechanism, to guard against contention losses and to adapt to network component (link/node) failures. For each connection (flow), at any point of time, this system either works without protection or chooses from one of many available protection mechanisms, based on the current network conditions and the performance requirements. Further, we develop a self-restoration mechanism based on deflection routing, wherein learning automata are used to identify an efficient alternate route to the destination, when there is a failure on the primary route. We show through extensive simulation studies that our mechanisms significantly improve burst loss probability over their existing counterparts.