Elastic Local Breakout Strategy and Implementation for Delay-Sensitive Packets with Local Significance

TL;DR

This paper proposes a real-time local breakout strategy for delay-sensitive packets with local significance, using a Markov decision process to optimize traffic routing and significantly reduce end-to-end delay in wireless networks.

Contribution

It introduces a novel local breakout framework with a real-time decision rule based on Markov decision processes for delay-sensitive traffic.

Findings

The framework effectively reduces end-to-end delay for local significance packets.

Real-time decision rule balances load between core network and local breakout link.

Test-bed results confirm feasibility and delay reduction benefits.

Abstract

Explosion of mobile traffic will bring a heavy burden to the core network, and rapid growth of mobile devices, as well as increasing demand for delay-sensitive services poses severe challenges to future wireless communication systems. In this regard, local breakout is a promising solution to save core network load and, more importantly, to reduce end-to-end (e2e) delay of packets with local significance. However, the capacity of local breakout link is limited, resulting in excessive delay when the traffic load through the local link is high. Therefore, the decision on whether the traffic flows should be transmitted through core network or by local breakout link has great practical significance. In this paper, we propose and implement a novel local breakout framework to deliver low e2e delay packets with local significance. A real-time local breakout rule based on the solution to a Markov decision process is given, showing that some packets with local significance should pass through core network rather than being delivered by local breakout link to meet the delay requirements. To test our proposed framework, a long-term-evolution (LTE) based test-bed with virtual base stations is implemented, by which we show the proposed framework is feasible and the e2e delay is significantly reduced.