Trade-off analysis in monitoring IP networks

Abstract

Efficiently monitoring the behaviours of IP networks is a formidable challenge for network management. With comprehensive perspectives of trade-off through multiple hierarchical dimensions, a Hierarchical Trade-off Analysis (HTA) model and its applications are presented in this study that offer operation strategies to quantitatively analyze and solve multiple objective and subjective decision making problems in network monitoring. The work is presented in two parts, i.e. a theoretical explanation of the proposed model and its applications, and a series of experimental simulation studies to verify and evaluate various aspects of the proposed model and the applications. The theoretical explanation includes three sections. Firstly the HTA model is defined and explained in Section 1. In Section 2, an HTA-based monitoring scheme is provided and analyzed with the goal of finding an efficient solution to the network monitoring problem. This strategy transforms the conventional network monitoring problem, which is provably NP-hard, into a multi-objective optimization problem within the category of "decision-making before search" strategy. In this scheme, high-level abstract qualitative management objectives, network-wide practical objectives, and low-level quantitative measurement metrics are taken into account to formulate quantitatively an aggregation for decision-making. Lastly, in Section 3, before examining the practicability of the scheme, an abstraction of management quality assessment based on the same HTA model is introduced to help evaluate the efficiency of management efforts in IP networks. Being derived from the standardized software quality model, the management quality model and its related Management Quality Index (MQI) express the objective and subjective management demands and detailed measurements in a numerical index. The experimental simulation part also contains three sections. Firstly, with the introduction of the ns-2 simulator as a systematic performance evaluation tool, a series of network scenarios, with an Internet Service Provider (ISP)-like topology and various traffic patterns, are simulated and evaluated to assess the management quality model. It is found that the MQI is quite acceptable in reflecting overall management impact on network performance. In the second section, an analytical modelling of the HTA monitoring scheme, specialized on polling frequency differentiation of network monitoring, is investigated to verify its appropriateness and limitations. Within this section, a simulation study is further conducted to evaluate the HTA scheme. The simulation results show that using knowledge of the global information in monitoring produces a sensible reduction in the number of management polling traffic at acceptable monitoring performance. Finally, due to current prevailing research and implementations of Quality of Service (QoS) mechanisms in IP networks, empirical studies of the HTA-based monitoring scheme on a QoS-enabled IP network are performed in the third section. Detailed evaluation insights how the HTA model and its applications, the HTA-based monitoring algorithm and the HTA-based MQI, can be applied in the fast developing real-time networked systems.