User-centric versus network-centric load balancing: How to provide good QoE for gold users

Olli Mämmelä (Corresponding Author), Petteri Mannersalo

Research output: Contribution to journalArticleScientificpeer-review

Abstract

The predicted increase in mobile data traffic and in the number of mobile users poses challenges for network operators. Future mobile networks should support traffic volumes of several orders of magnitude larger than today. This may create situations in which the transmission capacity of the network is inadequate for the traffic demand. Intelligent and robust network management methods are thus needed to handle the upcoming changes. Because quite often the bottleneck lies in the wireless part of the network, it is essential to have methods in place to select appropriate wireless access points for the users or their applications. This can be performed either by a central entity or in a distributed manner in which the intelligence is placed in the user equipment or hybrid combinations of those. The selection criteria may vary from selfish optimization of a single user's quality of experience to global network load balancing and class-wise prioritization. This work explores the wireless access selection scheme from a system perspective in scenarios where users of different subscription classes are using progressive video streaming. Evaluation is carried out between a Q-learning based user-centric algorithm, a user class aware network-centric algorithm and a setting without handovers. Simulation results show that both user-centric and network-centric algorithms are able to improve the network performance and the quality of experience of the users. The gains of the network-centric algorithm are more evident during high congestion. In a lightly congested network, the distributed selfish optimization is also qualitatively a good solution for gold users.
Original languageEnglish
Pages (from-to)242-259
JournalInternational Journal of Network Management
Volume25
Issue number4
DOIs
Publication statusPublished - 2015
MoE publication typeA1 Journal article-refereed

    Fingerprint

Keywords

  • wireless access selection
  • Q-learning
  • load balancing
  • network management framework
  • video streaming
  • simulation

Cite this