Mathematical model for horizontal on-demand vEPC scalability in SDN-based environment

Alexey Tsarev, Pavel Abaev
Next generation networks such as 5G provide a new approach for building highly scalable network infrastructure owing to NFV/VNF and SDN technologies.Infrastructure scalability allows a network operator to offer fine-graded on-demand services deployment. We consider virtualized Evolved Packet Core which functions are represented in form of group consisting of several independent Virtual Network Functions. Computational capacity of every vEPC group couldbe extended via scaling mechanism that follows the hysteretic algorithm. In this paper we build a mathematical model of horizontal scaling for vEPC functions in terms of queuing model with finite buffer size, several group of servers and queue thresholds. As a result, analytical formulas were derived and QoS metrics were proposed. Analytical evaluation shows that correct thresholds selection allows to significantly improve system performance.