Mobile user type aware load balancing algorithm in SD-RAN

Abstract

Under extreme increase on video contents in eMBB applications; the 5G requirements cannot be handled by the conventional self-organizing in 4G infrastructure. While executing load balancing in 5G RAN, mobile user type for eMBB applications should be considered. Nowadays, eMBB has been carried by QUIC and HTTP2.0 protocol for Android and iOS users, respectively. In mobile user aware load balancing, Deep Packet Inspection (DPI) up to application layer for packet routing is required. This can be only handled by Software-Defined Network (SDN) without any hardware expenditure in physical infrastructure. Therefore, this paper proposed Software-Defined Radio Access Network (SD-RAN) with two novel functions: Waiting Time Function (WTF) and Load Balancing Function (LBF). In WTF; the queuing inspired approach is proposed for the low complex implementation of the mobile user aware load balancing in 5G-RAN. Waiting Time parameters for iOS and Android users are analytically defined by M/G/1 and G/G/1 Markov queues. It is also executed by M/M/c/K Markov model SD-RAN topology. In LBF; a novel Mixed Integer Linear Problem is defined for waiting time optimization. To overcome NP hardness, a local search for the eMBB load threshold analysis is performed and determined as 0.79 and 0.94 for UMas and UMis. A low complex load balancing algorithm is proposed in the light of these thresholds. According to performance results; SD-RAN outperforms nearly 40% QoS than the conventional SON according to received packet count. It can serve 40% more users than the conventional one without any extra expenditure on physical infrastructure. As a result, it can handle eMBB flows with an acceptable waiting time under 2 milliseconds level