Research On Network QoS Analysis And Guarantee Under The Background Of Heterogeneous Services

The new types of services are emerging with the coming of the 5G era,which imposes stricter requirements on QoS(Quality of Service).However,the existing QoS theory seriously lags behind the flourish development of communication networks.As the network scale continues to expand and the service type continues to diversify,the differences of traffic become greater and greater,which brings more difficulties and challenges to the guarantee of QoS.At the same time,the contradiction between bandwidth efficiency and QoS guarantee is increasingly significant.It is an urgent problem that how to evaluate QoS accurately and provide QoS guarantee efficiently.Therefore,it is of practical significance and important value to explore new theories and methods to analyze and guarantee QoS under the background of heterogeneous services.In this paper,the new methods of network QoS analysis based on martingale theory are proposed by establishing network queuing models with multiple arrivals.Meanwhile,it is explored that how to guarantee QoS with a low cost of bandwidth under the background of differentiated services.When the communication network is modeled as the queuing system,the QoS performance is mainly reflected by the buffer behavior and delay,etc.Relying on the martingale theory,this paper analyzes the complex queuing system.More perspective and more accurate QoS evaluation is obtained.On the basis of supermartingale-based delay analysis,a bandwidth estimation algorithm under delay QoS constraint is designed.For the carrier sense random access networks,considering the QoS performance degradation caused by hidden terminals,and considering the heterogeneity of terminal traffic,we propose a grouping algorithm based on game theory.Further,we analyze the delay performance for each group based on supermartingale theory.The main content and innovation of this paper are as follows:(1)For the network queuing system with the arrivals aggregated by different distributed services,this paper introduces martingale theory and proposes an instantaneous analysis method to analyze queuing behavior.In the proposed model,the two-dimension characteristics of probability and time are both analyzed.In our work,by integrating multiple arrival processes and service process in the martingale framework,we construct a martingale relative to queue length.Relying on the high order statistical characteristics of arrival processes and service process,another martingale is constructed.By establishing the expectation inequality,the new martingale is also related to queue length.The first buffer overflow probability and the expectation of buffer stopping time are obtained based on the stopping time theory.According to the instantaneous observation queue length,the evolution tendency of queue buffer can be predicted in probability.The time expectation of the target event occurrence in the process of queue evolution is obtained.Simulations present the transient characteristics and two-dimensional time probability characteristics of the buffer behavior.This work provides a new perspective to evaluate buffer behavior for the queuing system with aggregate traffic,which provides a guidance for buffer design and buffer management.(2)For the queuing system with multiple arrivals containing bursty services,a tight delay performance bound is obtained based on supermartingale theory.We propose a bandwidth estimation algorithm under delay QoS constraint.It is the first work using supermartingale to evaluate network traffic load.In our studies,we first establish the supermartingale for arrival processes and service process,then construct the supermartingale relative to queue length.Using the stopping time theory,a tight bound of delay violation probability for aggregate traffic is derived.Due to the implicit relationship between delay QoS and service rate in the supermartingale model,we design a binary search algorithm to solve the minimum required service rate under delay QoS constraint,which is defined as the supermartingale bandwidth.Simulations prove that the supermartingale bandwidth is smaller than the classical effective bandwidth,especially when the traffic contains bursty services.Therefore,the supermartingale bandwidth estimation can improve the utilization of bandwidth resources.(3)For the widely used CSMA/CA(Carrier Sense Multiple Access/Collision Avoidance)service mechanism in random access networks,considering the QoS performance degradation caused by hidden terminals and the difference of terminal traffic simultaneously,this paper proposes a grouping algorithm based on game theory to mitigate the hidden terminal collision and achieve traffic load balance.Firstly,according to the traffic of each terminal,the algorithm forms an initial group with basic load balance,then the adjustment of game grouping is performed with the goal of minimizing hidden terminals and maximizing load balance.After grouping,we establish a queuing system with heterogeneous arrivals from multiple terminals and CSMA/CA service.Relying on the supermartingale theory and establishing new QoS indicators,the delay performance of each group is evaluated.Simulations prove the effectiveness of the proposed grouping algorithm.The grouping algorithm is beneficial to the network service fairness guarantee and the efficient QoS guarantee.The results of delay analysis could provide a guidance for bandwidth allocation and adjustment.This paper mainly focuses on the related research of QoS in the network scenarios with heterogeneous arrivals.Our research methods are also applicable to the scenarios with multiple homogeneous arrivals,which is a special case of the model.The QoS performance analysis and bandwidth estimation method based on martingale theory are also applicable to the scenarios with a single arrival.The work of this paper could provide valuable references for the design and improvement of network service.