CONTRIBUTION TO ADAPTIVE SAMPLING OF QoS PARAMETERS IN COMPUTER NETWORKS 1
Juraj Giertl, Rudolf Andoga, Michal Mirilovič · 2006
SUMMARY This paper focuses on adaptive sampling that is used by measurement and evaluation of operational parameters of computer networks. These parameters characterize networks' ability to maintain quality of services (QoS) defined by volume and time operational characteristics. Process of measurement can be performed using several methods. From the measurement efficiency point of view, the most appropriate are the so-called passive methods. No testing traffic needs to be generated while using them, and parameter values are derived from ongoing network activity. This area's forming standard that has appeared recently is referred to as the standard for exporting information on IP flows (IPFIX). However, there are several problems related to measurement of network's QoS parameters. The most important ones include a huge amount of data that needs to be processed on the measurement platform and afterwards to be stored for potential future use. This issue has become critical as the speed of currently used network technologies increased. However, there are several methods of reducing the amount of processed data. One of them is sampling which makes possible to select a characteristic sample of the overall packet population. Recent sampling methods are inefficient because they do not reflect on the nature of ongoing network activity. This problem, though, can be overcome by the so-called selective sampling. The paper intends to outline a model for adaptive sampling based on modification of the sampling interval by fuzzy logic controller. Experiments leading to the optimization of its parameters are also presented. At the end, the proposed methodology is compared to conventional sampling methods.