Resilience and Efficiency of Multi-layered Networks

多層網絡的彈性與效率

Student thesis: Doctoral Thesis

View graph of relations

Author(s)

Related Research Unit(s)

Detail(s)

Awarding Institution
Supervisors/Advisors
Award date10 Jan 2020

Abstract

In recent decades, we have experienced an unabated enormous growth of telecom- munications applications and services. The demand for voice and data services over telecommunication networks continues to rise in the past few years. The evolving networks with 5G and cloud technologies which include complex data generated by digital services and content further bring new opportunities and challenges to communications service providers (CSPs). To keep pace with the increase of the demands, several Internet technologies were developed and some are widely adopted, such as IP, MPLS, WDM. Network operators aim to provide quality guaranteed services by efficient and effective provision and management of the resources in the network. They also aim to maintain its service quality during peak hours and even if the networks are confronted with failures (i.e., man-made fault or natural disasters). The need for developing a high speed, quality guaranteed and resilient network becomes paramount. In accordance with the current growth in complexity and size of the Internet, new methodologies are required to be developed for cost-effective and resilient multi-layered telecommunications networks. By deploying sufficient resources using a cost-effective combination of the technologies, network operators can ensure that customers obtain services that meet required quality of service (QoS) within a reasonable budget that includes capital expenditure (CAPEX) and operating expense (OPEX).

In this thesis, we first provide a cost-based polynomial-time heuristic algorithm called multi-layered market algorithm (MMA). MMA can efficiently solve resource provisioning problems for multi-layered, multi-technology and realistically-sized networks. MMA solution is based on realistic traffic model where each end-to-end traffic stream is modelled either by constant bit-rate (CBR) or variable bit-rate (VBR). A VBR traffic stream is modelled either by a short-range dependent Gaussian process, or by a Poisson Pareto burst process (PPBP) which is long-range dependent (LRD) under certain parameter values. Since PPBP exhibits the LRD, this model is known to capture the behavior of internet traffic demand. The consideration of VBR traffic models in a multi-layered network optimization is a key novel aspect of MMA.

MMA considers different sharing properties of different switching technologies to allow traffic streams to share the capacity and the cost of the resources they use. For each end-to-end traffic stream in the network, MMA aims to provide near optimal solution for resource provisioning based on different routing strategies depending on flow-size and the choice of transport technologies. In particular, it implements statistical multiplexing and flow-size based routing which each end-to-end traffic stream was split and routed independently according to flow sizes. As routing influences resource allocation in the network, such routing strategy has an important effect on cost-effectiveness for resource provisioning. The complexity resulting from the considerations including layering and PPBP traffic requires a simplified design philosophy. This is achieved by adopting shortest path routing in each layer. MMA is based on an iterative algorithm, and resource provisioning that is performed link-by-link in all layers.

MMA has been further extend to resilient design of multi-layered networks. In particular, under this thesis, MMA has been extended for survivable design considering a series of failure scenarios, e.g., node failures, link failures. In resilient design, MMA aims to maximize earnings before interest and tax (EBIT). EBIT is defined as annual revenues minus annual costs. The costs include amortized CAPEX, OPEX, and penalties (compensation to the customers when the service is degraded or interrupted). Resilient MMA provides resource provisioning solutions for survivable and cost-effective design of multi-layered networks where sufficient resources are allocated under different grade of service (GoS) levels so that acceptable QoS of services is still maintained even under conditions of failures. The novel aspects of MMA include the incorporation of VBR traffic streams in survivable multilayered network design together with the aim to maximize EBIT.

MMA runs on a Java-based software called network mark-up language (Netml). Netml provides a network editor platform which is simple for users to create and store a multi-layered network and can apply network design algorithms to arbitrarily large networks. Netml provides a series of visualization tools, which help understanding of network design solutions. Netml also provides a method to verify the optimization results by using a method from accounting systems called double-entry book-keeping. As the benchmark, we also provide results from integer linear programming (ILP) formulations with deterministic (CBR) traffic streams. MMA is validated by com- parison between its solutions and those ILP benchmarks using six-node, three-layer networks. We also provide optimization solutions for a wider range of networks to gain further insight and understanding of the benefit of MMA. Comprehensive results show that MMA can provide resilient and cost-effective resource provisioning solutions for network design by considering statistical multiplexing, dynamic links and different GoS levels under failures. MMA can assist CSPs for future multi-layered network design including resource provisioning, network dimensioning and network evolution.