The Research Of Queuing System Model With Heterogeneous Servers Under Blending Services Arrival

In recent years, Internet of Things is an attractive concept in the worldwide, which isall-inclusive and covers "electron+computer+communication" industry. At present, thenumber of Chinese mobile users of the Internet of Things is beyond50million, and the rapiddevelopment of IoT in China leads to over500billion IoT devices by2015, which willundoubtedly inject new vitality to our communication industry. M2M service is thecommunication between machines; it plays an important role in the early development stageof the IoT. Most M2M devices are located in the places not suitable for the erection of cable,therefore, wireless network used for M2M services to access has a significant advantage.The behavior characteristics and flow characteristics of M2M services are different fromthose of H2H services, as a result, when M2M services share wireless network resourceswith existing H2H services, the services arrival in the network side will change. Thesechanges not only lead to the increase of the network load, but also change the characteristicsof aggregate traffic, and then affect the allocation and management of network resources.Model the blending services that reach the network side with the data collected from thenetwork simulation platform, and the entire network can be modeled more accurately usingthe established flow model as system input. At present, the Internet of things is in rapid development stage, and the market size is also very limited. A lot of M2M services are stillin trial stage, which have not launched on the market massively. Therefore, we can’t collectthe related data of M2M services from the real network now. NS-2is a useful networksimulation software. In this paper, we set up the network environment on the NS-2simulation platform, choose a few kinds of typical M2M small data services and H2H dataservice which are closely related to life as service source. According to the behaviorcharacteristics of all kinds of services, we rationally configure service terminals and collectthe data of blending services such as packet arrival intervals in the network side.At present, there have been many achievements about traffic modeling. To get accurateflow modes, the aggregate traffic is modeled as MMPP model, FGN model and Generaldistribution, and the accuracy of different traffic models are derived by QQ plot andleast-square method. MMPP models in different time scales and states are established byLAMBDA algorithm. Then, a single comprehensive MMPP model is established fromMMPP models in different time scales by Kronecker sum and Kronecker product.. Finally,considering the advantages of strong controllability and easy theory solving, MMPP model ischosen as the input of the system model.Heterogeneous networks have always been the research hot spot, and the rapiddevelopment of wireless technology such as WCDMA, WLAN and LTE provides a morespacious space for services to choose the access network. The popularity of multi-modeterminal further speeds up the development that heterogeneous networks act as access networks. In this paper, heterogeneous servers is used to describe heterogeneous networks,and queuing theory is used to establish the queuing system model with heterogeneousservers and the blending services arrival. According to the birth and death process theory, wedefine the state of the system and derive the markov state transition process. The model issolved with the matrix geometric method, and some system performances such as theaverage queue length, system efficiency, delay and packet loss rate are obtained. Ourresearch is done from three aspects: Firstly, the queuing system with two heterogeneousservers is modeled, and the shortest waiting time is guaranteed by defining queuing rulesreasonably. In order to describe the parallel service characteristics of the network more really,the heterogeneous servers can be converted to homogeneous server groups (each groupcontains a different number of small homogeneous servers), and we solve the queuingsystem model under two kinds of access rules, that is the random access and the balancedaccess. Then we derive the system performance differences under different access rules.Finally, considering that there are not only data services but also voice and video serviceswhose QoS demands are higher in real network, on the basis of the research of homogeneousserver groups, every arrival service no longer simply requests a small server, but can requesta few small servers. Through solving the correction system model, the system performancessuch as the average waiting queue length and the system efficiency are derived. Under thecondition of invariable total service rate, we change the service rate of the small server (thenumber of small servers in the homogeneous server group also changes), and observe the change of the system performance. The accuracy of the model can be validated through thecombination of theoretical analysis and simulation method.