LTE user plane congestion detection and analysis

Abstract

In Long Term Evolution (LTE) systems both radio and transport congestion might occur due to the increasing data traffic generated by smart devices accessing popular Internet applications. Packets arriving at a congested resource experience increased delay, random discards and eventually tail drops as a consequence of network side buffer overflows. The LTE system delegates congestion management to the end-to-end protocols such as the Transmission Control Protocol (TCP), which is not an efficient solution in mobile systems. Reduced TCP transmission rate and significant degradations such as retransmission timeouts (RTO) are commonly present upon congestion. Therefore, instead of relying only on the congestion management capabilities of the TCP, efficient network side mechanisms are needed to achieve harmonic system operation and prevent unnecessary TCP degradations. This was recognized by 3GPP resulting in the specification of the User Plane Congestion management (UPCON) framework. Nevertheless, the real time congestion detection and management mechanisms are not in the scope of UPCON. As the prerequisite of any efficient congestion management technique is the capability of accurate congestion detection, this paper provides a survey of the existing solutions, analyses their shortcomings and proposes a congestion detection method that correlates the delay skewness and the pattern of burst losses collected by user traffic monitoring. The method is presented via measurements from an LTE lab network.