Παραγωγή και ανάλυση κίνησης με έμφαση σε δίκτυα τεχνολογίας ΑΤΜ και IP

Abstract

The information volume and the application variety are constantly increasing imposing new demands on the telecommunication networks. Due to these reasons the problem of efficient network design, monitoring and control is more apparent than ever. Traffic generation and/or analysis tools, if properly applied, can assist in finding suitable solutions. The PhD thesis focuses on the development and usage methods of similar tools and is structured in ten chapters. In the first chapter the main components of the QoS delivery problem are mentioned while the emerging need for sufficient traffic generation and/or analysis tools is justified. The second chapter presents in detail the main characteristics of the environment where the proposed tools may be used. This environment consists mainly of ATM and IP topologies. The emphasis is put on the similarities between the discussed different platforms. At chapter 3 starts the detailed description of the proposed tools. More specifically in chapter 3 an advanced software architecture for ATM traffic generation is presented. This architecture exploits a stable and reliable hardware so as to provide the generation of traffic flows compliant with high level traffic model specifications. The basic advantage of the proposed generator is that it guarantees a fast and accurate traffic generation process. The software part introduces several innovations towards minimizing speed and memory constrains related to the hardware. Chapter 4 is dedicated to the presentation of the software supporting a prototype ATM traffic analyzer. The software exploits the experience acquired during the implementation of the ATM traffic generator counterpart. The software architecture of the proposed ATM analyzer exploits and controls the enhanced capabilities of the promising underlying hardware module the role of which is expanded in order to provide real time QoS metrics at relatively low computational cost. Apart from this, the software architecture directly creates models containing some cases of traffic being monitored. In chapter 5, an IP traffic generator is described. The remarkably successful use of the prototype ATM traffic generator, in conjunction with its well-balanced architecture led to the adoption of the presented generator’s logic in IP environments as well. It must be noticed that the whole effort does not ignore the idiosyncrasies of the Internet. The IP traffic generator, despite the simplicity of its first version, was proved as very useful to many testing cases where the exact reproduction of captured traffic was crucial or the hiring of real sources too expensive or complicated. Chapter 6 is dedicated to the description of traffic analyzers, especially designed for IP networks. As a first attempt, the ATM analyzer engine was further exploited, after the necessary modifications in its software architecture, while a sufficient «IP over ATM» mechanism was hired. At next stage, a native IP analyzer was designed and implemented. The later tool hires external clock units for solving the synchronization problem between source and destination nodes. The capabilities the two analyzers can be enhanced by injecting appropriate monitoring traffic into the network under evaluation. The «IP over ATM» based IP analyzer is capable of performing accurate and fast real time measurements without overloading the hosting system. One of its main advantages is the direct delivery of histograms presenting the inter-packet distance (or the packet size) distribution. The proposed native IP analyzer solves the problem performing real time reliable measurements of end-to-end delays or losses of IP packets. Chapter 7 is dedicated to another group of performance evaluation tools that work under looser real time constrains. They perform post-processing using log files captured by various traffic analysis tools. Via post-processing more testing cases can be assessed while more complex metrics can be incorporated into the analyzer’s logic. The computational load required for measurement processing is completely disconnected from the relevant load for data gathering. Chapter 8 presents an analyzer tool that is based on the logging capabilities exhibited by advanced network elements. Although the proposed tool does not differ in its high level architecture from the other traffic analysis tools being presented, data uploading mechanism is based on the SNMP MIBs supported by network elements like an IP router or an ATM switch. The specific tool is targeted towards assisting the rest of the analyzers. The main advantage of this tool is its flexibility as it can be transferred easily from one network platform to another. Chapter 9 presents several characteristic cases where the proposed tools are involved. Indeed, traffic generation and analysis tools have been exhaustively tested during a large number of experiments. The easy use of the tools and the complementarity of their features led to the appropriate solutions very fast in all cases. Furthermore, all the tools are cost-effective. The experiments being performed indicated for one more time that there are no solutions that are both integral and optimal, but only partial ones implied by the promised QoS and cost requirements. Finally, chapter 10 summarizes the thesis’s innovative points and contribution and indicates some open issues for further research.