Queue Models And Overload Control For Machine Type Communications

Machine type communications (MTC) is defined as machine to machine (M2M)communications over cellular mobile networks. It is a subclass of M2Mcommunications which are automatic applications involving machines or devicescommunicating through a network without human intervention. In contrast to human tohuman (H2H) communications, MTC applications are essentially characterized byvarious application types, diverse business models, huge number of terminals,infrequent small amount data transmission, time controlled transmission, time tolerantquality of service (QoS), terminals with low mobility and etc... The typical MTCapplications include smart grid, Internet of vehicles, smart home, telemetering, wirelesssensor network and etc... These applications are the integral part of future ubiquitousnetwork, which have broad application prospects and market potentials. Conventionalcellular mobile networks have been optimally designed for H2H communications. Toaccommodate the new challenges brought by MTC and guarantee the QoS of both MTCand H2H communications, it should be further developed and re-optimized.Prior to any technology enhancement for MTC, an appropriate traffic modelenabling preliminary performance evaluation is of prime important. Classic Markovianqueue models under the assumption of Poisson arrival have been long-term used forstochastic traffic modeling in telecommunications, as Poisson process can model thenumber of occurrences of a rare event in a very large population. However, thepossibility of massive MTC devices generating their access attempts in a short timeperiod makes the arrival pattern of MTC more bursty. The burstiness of MTC trafficinvalidates the commonly-used Markovian models under the assumption of Poissonarrival which is characterized by the negative exponentially distributed inter-arrival time(IAT). In case of this, a non-Markovian statistical traffic model may be more appropriate.Beta distribution is a distribution with two shape parameters and limited support rangewhich has the flexibility to model a lot of MTC devices starting their access in a shorttime period. It has been proposed by the3rdGeneration Partnership Project (3GPP) tomodel the IAT of MTC.On the basis of this, this dissertation builds four traffic prediction models for localcellular, two queue models for MTC and presents some insights to access control forMTC, which serves as a preliminary study on traffic modeling and network performance evaluation and management for MTC by the use of queue theory andrenewal theory. The main contributions are as follows:①According to the main application scenarios and the typical features of MTCdescribed in3GPP TS22.368, this paper establishes four traffic prediction models forlocal celluar in the context of massive MTC access, namely the no-mobility model,extended no-mobility model, low mobility model, full mobility model, in which themethod to calculate the volume of burst data and signaling overhead is given out.②By the useage of order statistics, the differences between Beta distribution andexponential distribution which is the inter-arrival time distribution of Poisson processare elaborated as well as the reason why Beta distribution is more suitable for MTC. Tocarry out the performance analysis of MTC under Beta arrival, the moment generationfunctions (MGFs) of Beta distributions with integral paramerters, which are theconfluent hyper-geometric functions of first kind, are given out analytically. By theuseage of renewal theory and Volterra integral equation of the second kind withdifference kernel, the methodology to deduce the access intensity of MTC that isdefined as the mean number of renewals by time t, is presented. Numerical results arepresented by useage of numerical Laurent series expansion and then present the maincharacteristics of the arrival process of MTC. Two queue models are built for MTC.One is about pure MTC traffic, namely, Beta/M/1model. The other is about blendingH2H and MTC traffic, namely, Beta+M/M/1model. These two models associated withBeta distribution are special cases of G/M/1model. With the MGFs of Betadistributions deduced above, they can be solved by the standard method of G/M/1model. Numerical results show that as the burstiness of MTC is extremely larger thanthat of H2H communications, it could potentially increase the mean sojourn time andthe mean waiting time of queueing syetems and degrade the performance of networkand decrease the QoS of H2H communications. These works provide researchers andengineers a basis to appropriately choose traffic models for different MTC applicationsand promptly judge the effectiveness of newly designed control measures.③To enhance radio access network overload control, the underlying mathematicalmeanings of3GPP proposals are firstly elaborated. On the basis of aforementioned queuemodels, four overload control measures are proposed. They are inter-class grouping,IAT shaping, ACB scheme dedicated to Beta distribution and MTC specific back-offmechanism. Among them, only the last one has the potential to put into practice. MonteCarlo simulations show that MTC specific back-off mechanism can effectively reduce the collision probability of random access process of network in context of MTC andimprove the throughput by2%-5%with additional100-200packets access delay whichis tolerable for most of the MTC applications.