光因特网管理技术研究

随着Internet的迅速普及,以IP为主的业务在飞速增长,未来的网络将要应付这种不断增长的IP业务,光因特网是未来Internet的发展方向。对光因特网的管理技术进行了研究,讨论了基于策略和基于SNMP的网络管理的原理和实现方式。     

A Reliable Method of Image Registration Based on Optical Flow Field and Feature Extraction

A novel method for solving the automated registration problems based on the global optical flow field and feature extracted from images is proposed. The method can be preformed with three steps. Firstly, the global optical flow field is computed, which provides a platform of background registration. Secondly, the optical flow field and the edge feature of moving target are integrated to segment the whole target. Finally, the registration of moving targets is finished through the transformation computation of perspective projection model. The simulation experiments for infrared images show that the method proposed is not accurate, but robust to motion of scene, which is ideally suited for the application of automated image registration.     

Contour-Based Multi-Source Information Fusion for Motion Segmentation

A motion segmentation framework that effectively exploited the multiple sources of image information and fused these sources of the information synergisti-cally was proposed to serve the purpose of motion segmen- tation. A Markov process was formulated for motion seg- mentation in which two feature spaces were established to estimate the state transition Probability density function （PDF） and the initial state, respectively. An information fusion space was developed such that each motion struc-ture was described as a single distribution in this space. The proposed framework can naturally embed the evolution equations of the active contour methods into the seg-mentation to achieve contour-based segmentation results. Extensive empirical evaluations demonstrate the robust-ness and the promise of this framework.     

Multicast with a new switching structure in optical networks

The emergence of new services demands for a multicast function in optical networks. At the same time, wavelength converters are introduced to increase the efficiency of wavelength usage. It is because of the high cost and complex architecture of optical multicast and wavelength conversion technology, that a new switch structure is introduced, in which optical splitters and wavelength converters are shared per-node. In order to accommodate this architecture, a multicast routing and wavelength assignment algorithm in a splitter–converter-sharing optical network and a changing link weight policy to balance network traffic are proposed. By extending RSVP-TE (Resource ReSerVation Protocol-Traffic Engineering) and OSPF-TE (Open Shortest Path First-Traffic Engineering), an optical multicast mechanism is provided, and message type, signaling flow, and finite state machine model are given. Simulations of NSFNET show that, when the number of splitters and converters are 50% and 12.5% of the full equipment respectively, the performance is close to the ideal case. Using a changing link weight policy can improve performance greatly, when there are enough splitters and converters.     

Analysis of blocking probability in optical burst switched networks

In this article, we study the blocking probability in a wavelength division multiplexing (WDM) based asynchronous bufferless optical burst switched (OBS) network, equipped with a bank of wavelength converters. Our analysis encloses two wavelength reservation schemes JIT (just-in-time) and JET (just-enough-time), and two-class data rate. The contribution of our work includes: (i) derivation of an accurate model for blocking probability of lower priority bursts in case of a non-preempted model; (ii) provision of the analytical model for blocking probability calculation in the OBS network, which includes these variables: two signaling schemes, partial wavelength conversion, two-class data, traffic intensity, cross-connect speed, number of wavelengths in WDM fiber, number of fibers in the node, number of wavelength converters, and number of nodes in the path; (iii) simulation results, which show that partial wavelength conversion provide quite satisfactory quality of service. We compare performance in a single OBS node, under various sets of parameter values. The OBS network shows great flexibility in terms of used multiclass data, and there is no dependence on the used higher layer protocol.     

Cross-virtual concatenation for Ethernet-over-SONET/SDH networks

Ethernet-over-SONET/SDH (EoS) is a popular approach for interconnecting geographically distant Ethernet segments using a SONET/SDH transport infrastructure. It typically uses virtual concatenation (VC) for dynamic bandwidth management. The aggregate SONET/SDH bandwidth for a given EoS connection is obtained by “concatenating” a number of equal-capacity virtual channels. Together, these virtual channels form a virtually concatenated group (VCG). In this article, we introduce a new concatenation technique, referred to as cross-virtual concatenation (CVC), which involves the concatenation of virtual channels of heterogeneous capacities. We show that CVC can be implemented through a simple upgrade at the end node, thus utilizing the existing legacy SDH infrastructure. By employing CVC for EoS systems, we show that the SDH bandwidth can be harvested more efficiently than in conventional VC. We consider two problems associated with routing CVC connections: the connection establishment problem and the connection upgrade problem. The goal of the first problem is to compute a set of paths between two EoS end systems such that a total bandwidth demand and a constraint on the differential delay between the paths are satisfied. Among all feasible sets, the one that consumes the least amount of network bandwidth is selected. For this problem, we develop an integer linear program (ILP) and an efficient algorithm based on the sliding-window approach. For the connection upgrade problem, the goal is to augment an existing set of paths so as to increase the aggregate bandwidth, while continue to meet the differential-delay constraint. We model this problem as a flow-maximization problem with a constraint on the delay of the virtual channels with positive flow. We then consider the problem of path selection under imprecise network state information. Simulations are conducted to demonstrate the advantages of employing CVC and to evaluate the performance of the proposed algorithms.

On capacity planning for the GMPLS network control plane

This article presents capacity planning rules for the control plane of all-optical networks featuring GMPLS and RSVP-TE as a connection setup protocol. As per RSVP standard, a refresh message mechanism is incorporated to RSVP such that the state is periodically refreshed on a link per link basis. We provide analytical expressions for the bandwidth and buffer sizes to be provided such that no flows are torn down due to lack of refresh messages. Our findings show that small buffers (several KBytes) suffice to sustain the signaling load for as much as 400 RSVP flows per link, with the simplest RSVP refresh mechanism (neither using link bundling nor acknowledgments). On the other hand, we also find the packet drop probability per link for a given network topology for the case that the flow survival probability is larger than a given threshold. We provide numerical examples based on the COST 239 european network topology and real RSVP traffic traces from early-commercial switching equipment.

Service differentiation scheme in OBS networks

In order to achieve service differentiation, especially loss differentiation, in optical burst switching (OBS) networks, we propose a dynamic fiber delay line (FDL) partitioning algorithm, which divides FDLs into several groups over a feed-forward output buffering architecture. In the proposed scheme, a plurality of traffic classes and FDL groups can be considered, and each FDL group is assigned to traffic classes, so that the target loss probabilities of classes are guaranteed. Also, the optimal number of FDLs for each FDL group by the proposed algorithm is decided in Poisson traffic environments. The extensive simulation results validate the effectiveness of the proposed dynamic FDL partitioning algorithm for the loss differentiation in OBS networks.     

Packet loss analysis in optical packet-switched networks with limited deflection routing

We present an approximate analytical method for the evaluation of packet loss probability in synchronous optical packet-switched networks which operate under limited deflection routing with the contention resolution method based on priorities. Packets are lost because they are removed by nodes. They are removed because they experience too many deflections and stay prohibitively long in the network. Such packets have to be removed because they will be ignored by the transmission protocols (like TCP) and because the quality of their optical signal is unacceptable. Presented are results for the network in the topology of the torus of the two-dimensional grid, which operates at a steady state with the uniform load u, . The strength of our analysis is its novel mathematical approach, which is capable of providing very low packet loss probabilities. For the network composed of 100 nodes, we predict the packet loss probability as low as 10⁻⁹ or lower, while simulation provided results only at the order of 10⁻⁶. For a given permissible packet loss probability, our analysis provides the maximal network load and the number of allowed deflections. We verify the analysis with simulation in the cases for which simulation gave results.

几种全光逻辑门浅析

全光逻辑门是实现高速光分组交换、全光地址识别、数据编码、奇偶校验、信号再生、光计算和未来高速大容量全光信号处理的关键器件,本文介绍了几种基于半导体光放大器非线性效应的全光逻辑门,分析了它们的系统结构和工作原理,并对各自的特点进行了比较,展望了光逻辑门在未来的发展方向.