无线通信系统面向突发流量的时延分析

Delay Analysis of Wireless Communication Systems for Bursty Traffic

摘要: 本文对无线通信系统的时延和可靠性这类的问题，提出了一种解决问题的新的模型——鞅论。针对有时延可靠性要求的业务，提出单跳时延性能分析鞅框架。根据鞅的基本性质推导出对随机过程可以进行鞅构造的特性。由于实际的5G网络中存在大量突发的以及自相似的流量，本文以马尔可夫调制开关过程为例，通过对突发到达流和服务过程进行鞅构造，得出时延违反概率和时延之间的关系。对突发到达流和服务过程分别进行了基于鞅论和基于有效带宽/容量理论的实验。仿真结果表明，对于突发到达流来说，采用鞅构造对比采用传统的有效带宽/容量理论而言得到的时延违反概率更加紧致。

Abstract: This paper proposes a new model to solve the problems of delay and reliability in wireless commu-nication systems, known as martingale theory. For reliability-aware applications with delay requirements, a single-hop delay performance analysis reinforcement learning framework is pro-posed. The basic properties of martingales are used to derive the characteristics of martingale construction for random processes. Considering the existence of large amounts of bursty and self-similar traffic in actual 5G networks, this paper takes the Markov-modulated switching process as an example, and martingale constructions are performed on the bursty arrival process and service process to analyze the relationship between delay violation probability and delay. Experimental results based on martingale theory and effective bandwidth/capacity theory for the bursty arrival process and service process are presented. The simulation results show that compared to traditional effective bandwidth/capacity theory, the martingale construction method obtains a more compact delay violation probability for bursty arrival flows.