Modeling And Optimization Of Short-lived M2M Services Based On TCP Over IEEE802.11WLAN

Along with the rapid development of the information technology and the fast increaseof human needs, everything in the world has the demand to access to the network andexchanges information with each other. So the concept of Internet of Things (IoT) has beenproposed and drawn worldwide attention. Internet of Things will greatly change people’sway of life, improve the efficiency of production, and bring huge economic benefits. As themain manifestation of the Internet of Things in current stage, machine to machine (M2M) isone of the popular communication technologies, which has tremendous market potential.The M2M industry chain develops rapidly. More and more equipments support networkaccessing and intelligent control. The3G+WLAN wireless broadband data networkdeployed by operators will be one of the best bearing platforms for M2M applications, for itswide coverage and high bandwidth. Especially, WLAN has been widely used for M2Mapplications because of its high data rate, convenient, and free for the spectrum. There aremany M2M applications used in all areas of society range from oil industry, public traffic, toenvironmental monitoring, and so on.In this paper we analysis and summary the characteristics of M2M applications. Theservice which transmits short-lived business based on TCP is one of the most commonapplications of M2M. For its popular use, there are large numbers of short-lived TCPconnections in the network, which will make a significant impact on the network. Therefore,the research into modeling and analyzing the short-live M2M services based on TCPprotocol over IEEE802.11WLAN has important theoretical significance and applicationvalue.The framework of the system model consists of two components: the TCP source modeland the IEEE802.11WLAN queuing model. First, we use a Markov chain to describe thebehavior of congestion windows of short-lived TCP. Mimicking the feature of M2Mshort-lived services, we use an exponential On/Off source as the application level servicesource. By comparing and analyzing, we fit the On/Off source pass through TCP with an IPPsource and used it as an arrival process for the network. Then we take WLAN as thebottleneck link of the network and model it as an IPP/G/1/k queuing system. To get theservice time distribution of MAC layer, our research use a two-dimensional Markov chainmodel the CSMA/CA mechanism. Solving the queuing model, we can get the packet loss rate and delay of queuing, then feed the system packet loss rate and RTT back to the TCPmodel. The parameters in the TCP model and queuing model are tuned by iterative analysiswith feedback from each other. The procedure is continued until a stable solution is obtained,then we can get the stable performance of the system.NS2network simulator is used to simulate the system, and the corresponding systemperformance parameters are obtained. The simulation results are compared with thetheoretical analysis in order to verify the correctness of mathematical model and simulationalgorithm. Because of the idealized of the802.11MAC layer queuing model of AP, there aredifferences between the theoretical result and simulation result. But the theoretical modelstill provides a basis for the theoretical studies. Then according to the result of themathematical model and simulation results, we find that the restriction of TCP congestionwindow make great influence on TCP throughput. Aiming at short-lived M2M services, wepropose a cross-layer TCP optimization scheme, which can change the max threshold ofcongestion window according to the information provided by application layer and MAClayer. The simulation result show that the throughput is improved when there are morestations in the WLAN system.