Dan Liao

A Performance Enhancing Proxy for terrestrial-satellite hybrid networks

In terrestrial-satellite hybrid networks, traditional TCP performance degrades a lot due to long propagation delays and high bit error rates link. Performance enhancing proxy(PEP) can shorten round trip time loop time and Peach can enhance the performance over high bit error rates link. In this paper, we propose a new scheme that combines PEP and TCP Peach for terrestrial-satellite hybrid networks. The scheme could reduce the negative impact on performance (in terms of throughput). To obtain an analysis result, we also introduce a mathematic model for our PEP. The analysis and simulation results demonstrate that our scheme could significantly improve the network performance than traditional TCP protocols.

On segment-shared protection for dynamic connections in multi-domain optical mesh networks

In this paper, the problem of how to provide survivability to multi-domain optical mesh networks is studied. Differing from the survivable research on single-domain networks, the whole network topology is unknown for all the source nodes. According to this point, we propose a novel protection algorithm called Enhanced Local Segment- Shared Protection (ELSSP) based on the simplified virtual topology for dynamic connections in multi-domain optical mesh networks. For the dynamic connections, ELSSP assigns rational link-cost for the virtual link on the virtual topology when calculating the primary path, and provides segment-shared protection in the passing domains of the primary path. Simulation results show the good network performance of ELSSP.

Study of Topological Dynamics and Periodicity of LEO Satellite Networks Based on Spectral Analysis

Topological dynamics and periodicity are two of the major differences between LEO satellite networks and terrestrial networks. In this work, we first propose MAPSS (Method to Analyze Periodicity of Satellite System) to obtain the system period of satellite networks based on spectral analysis. Then the topological dynamics of satellite networks, including the number of snapshots in one system period, is characterized by using this new method. Our method could be used in the development of routing algorithms as well as the design of satellite constellation.

A novel survivable traffic grooming algorithm with inter-layer sharing in IP/MPLS-over-VDM mesh networks

We propose a Mixed Sharing Auxiliary Graph (MSAG) for dynamic traffic grooming in heterogeneous WDM mesh networks. Based on MSAG model, a novel heuristic named BLSW-ILMS (Backup LSP Shared Working Lightpath with Inter-layer Mixed Sharing) is proposed. Simulation results show that the proposed algorithm can efficiently decrease the blocking probability.

An effective upgrading algorithm in incremental planning for optical networks

In this paper, we propose an effective upgrading algorithm in incremental planning for optical networks. This algorithm integrates making full use of existent available resource with upgrading capacity. Our objective is to minimize the newly added resource in the network. The Integer Linear Programming (ILP) formulation is presented. Since the ILP formulation is not adapted for large-scale network, a novel fast heuristic algorithm called Minimal Upgrading Resource (MUR) is proposed and evaluated. Computer simulation results show the MUR can achieve the performance near to the ILP results, and performs better than other two proposed approaches.

Minimum capacity dimensioning in diff-serv-aware MPLS networks

This paper focuses on minimizing the capacity of DiffServ/MPLS networks taking into account two widely deployed bandwidth constraint models: the maximum allocation model (MAM) and the Russion doll model (RDM). With the given physical network and traffic demands, by optimizing Label Switched Paths (LSPs) resource allocation, route distribution, our designs aim at minimizing the link capacity of the total network. The optimization problems are formulated as integer linear programming (ILP) models first, and then, since the ILP formulation is not adapted for large-scale network, two novel fast algorithms are proposed. Numerous simulation results for a variety of networks indicate that the performances of our proposed algorithms are close to the optimal results.