An Unsupervised Approach to Infer Quality of Service for Large-Scale Wireless Networking
- PDF / 1,541,332 Bytes
- 20 Pages / 439.37 x 666.142 pts Page_size
- 42 Downloads / 144 Views
An Unsupervised Approach to Infer Quality of Service for Large‑Scale Wireless Networking Lucio Henrik A. Reis1 · Luiz Claudio S. Magalhães1 · Dianne Scherly V. de Medeiros1 · Diogo M. F. Mattos1 Received: 7 October 2019 / Revised: 10 March 2020 / Accepted: 30 March 2020 © Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract Inferring the quality of service experienced by wireless users is challenging, as network monitoring does not capture the service perception for each user individually. In this paper, we propose an unsupervised machine learning approach to infer the quality of service experienced by wireless users, based on the different usage profiles of a large-scale wireless network. To this end, our approach correlates the usage data of access points, and the summaries of connection flows passing through the access points in the network. Then, we apply the k-means clustering algorithm to infer different network usage profiles. We evaluate our proposed approach to infer QoS on a real large-scale wireless network, and the results show that discriminating the flows into five clusters allows identifying prevalent usage profiles of the degraded state of the network and overload conditions in access points, considering only the flow summaries. Keywords Network-usage profiling · k-means · Clustering · Network management · Network characterization
* Diogo M. F. Mattos [email protected] Lucio Henrik A. Reis [email protected] Luiz Claudio S. Magalhães [email protected] Dianne Scherly V. de Medeiros [email protected] 1
MídiaCom - PPGEET/TCE/UFF, Rua Passo da Pátria, 156, Niterói, RJ, Brazil
13
Vol.:(0123456789)
Journal of Network and Systems Management
1 Introduction The data usage on mobile devices is growing every year. In 2016, mobile networks increased 63%, reaching an accumulated growth of 18 times in the last 5 years. Outsourcing mobile data traffic to fixed wireless networks is an increasing trend, which makes up 60% of global mobile traffic [1]. Moreover, the popularization of the IEEE 802.11 standard for wireless networks makes them a constant presence in several environments [2, 3]. However, the popularization also fosters the challenge of controlling and managing wireless networks due to the superposition of the coverage of access points, which causes interference [4]. Such control and management are more challenging in large-scale wireless networks, as the number of access points is even higher than in ordinary wireless network environments. The absence of intelligence and active countermeasures mechanisms in the network core, often associated with limited views of flow control tools, depletes the Quality of Service (QoS) experienced by users in large-scale wireless network [5, 6]. The quality of service perceived by the users is subject to the accessed content and interference. Thus, QoS inference implies the need to identify wireless network usage profiles. The running flows on the network provide enough data to estimate the overall network QoS level. The
Data Loading...
