Hasso-Plattner-Institut25 Jahre HPI
Hasso-Plattner-Institut25 Jahre HPI

Thi Than Mai Hoang

Planning of Multi-service Computer Networks

This habilitation thesis deals with various aspects of planning multi-service computer networks. These aspects are network applications, protocol multiplicity, topology complexity, traffic characterization, quality of service requirements, network costs, network architectures, communication forms, and communication layers. With these aspects in mind, methods and algorithms for traffic control and service quality control are presented and analyzed. The analysis of existing methods for planning computer networks shows that most existing approaches focus only on traffic, topologies, routing and network costs. The protocol multiplicity and communication forms are ignored. Furthermore, end-to-end abstraction and mapping are not covered in detail in most existing approaches. Based on methods for traffic control and service quality control as well as techniques for network planning, models, algorithms and architectures for planning multi-service computer networks are developed and investigated in a corresponding approach (Multi-Service Network Planning MSNEP).

With the presented approach MSNEP it is possible for the first time to realize the common consideration of all above mentioned aspects (especially protocol multiplicity, service quality requirements, network architectures, communication forms as well as communication layers) in the planning and optimization of network structures. The goal is to achieve a global optimum that corresponds to the cost-effective structure of an IP/MPLS backbone while adhering to the constraints of the given aspects. MSNEP enables traffic control and service quality control mechanisms for different traffic classes (data, voice, video) to be considered together with unicast and multicast communication services within the network planning process.

A prototype of the MSNEP approach is implemented in Java. It has been investigated for the planning of unicast and multicast IP backbones, for the analysis and forecasting of networks, and for the planning of MPLS VPNs and multicast VPNs. Simulation results in NS2 show that our approach can be used not only for cost optimization but also for quality of service improvement of multimedia applications in IP-based networks.