Authors:
Yazeed A. Al-Sbou
;
Reza Saatchi
;
Samir Al-Khayatta
and
Rebecca Strachan
Affiliation:
Sheffield Hallam University, Faculty of Art, Computing, Engineering and Science, United Kingdom
Keyword(s):
Quality of Service (QoS), network measurements, traffic-sampling methods, network performance analysis.
Related
Ontology
Subjects/Areas/Topics:
Communication and Software Technologies and Architectures
;
Data Communication Networking
;
e-Business
;
Enterprise Information Systems
;
Fairness Scheduling and Qos Guarantee
;
Information and Systems Security
;
Network and Service Management
;
Network Management
;
Telecommunications
;
Wireless Information Networks and Systems
Abstract:
As networks grow in complexity and scale, the importance of network performance monitoring and measurement also increases significantly. High data rates often lead to large amount of measurement results. Therefore, in order to prevent an exhaustion of the network resources and to reduce the measurement cost, a reduction of the collected data is required. A performance measurement method for estimating the actual network performance, experienced by the user, has been proposed. This study focuses on monitoring the network performance and estimates its main Quality of Service (QoS) parameters (delay, throughput, and jitter) through the use of a non-intrusive passive measurement method based on sampling methodologies. This method will overcome the drawbacks of both active and passive monitoring methods. That is because it measures the actual performance experienced by the user and requires reduced calculations of QoS parameters from the sampled packets. The validation of this approach wa
s analysed and verified through simulations. Three different sampling techniques (systematic, random, and stratified) were investigated. The study indicated that an accurate estimation of the QoS parameters could be obtained without the need to measure across the whole packets of traffic information. As a result, the scheme has shown an estimation of the detailed characteristics of performance for each user. For a bottleneck based network topology and traffic conditions used, the random sampling showed the best overall performance.
(More)