The Gravity Of The Distinction Between Business Model And Its Application In Traffic Matrix Estimation

As an important input of the network traffic engineering, Traffic matrix is the basisof network capacity planning, network optimal design, network monitoring and trafficanomaly detection. In order to better design, monitor and manage network, networkmanagers need to understand and master the network traffic matrix. Direct measurementof the network traffic matrix is very difficult or even infeasible, and so traffic matrixestimation is a very popular research field by indirectly observing link data. There islinear relationship among the measurable link data routing matrix and traffic matrix. Asthe number of measurable link is much larger than the number of OD(Original-Destination), tomography traffic matrix estimation is the underdeterminedinverse problem solution. The existing traffic matrix estimation method has itslimitations. First, existing methods generally assume that all the traffic flow between thesame node pair choose the same route. However, in order to provide better servicequality to different network service, different traffic actually choose different route.Secondly, the existing methods can only give the total traffic between OD traffic pair ofthe network, struggling to provide different service traffic between OD traffic pair. Inorder to plan and optimization of the routing of network traffic, the network managerswant to get the important service traffic between OD traffic pair.Distribution of different service traffic in the entire network is inconsistent andeach has certain structural features, this paper give the service traffic OD existinstruction matrix, using it to indicate service traffic in which the source node anddestination node pair being OD traffic pair. Based on distinguish between business ofgravity model is put forward in this paper, unlike the traditional gravity model don’tdifferentiate the service traffic in the network, distinguish between business of gravitymodel reflect the distribution of different service traffic in the network using servicetraffic OD exist instruction matrix, according to the access link traffic information ofdifferent service, the proportion that OD traffic accounted for the total traffic into thenetwork from the OD pair’s source node is estimated, different service traffic matrixestimated value is gained. The distinguish between business of gravity model can reflect the distribution of service traffic in the network better, service traffic matrix has moresatisfactory estimated result than the simple gravity model.The distinguish between business of gravity model doesn’t use the network internallink flow and routing policy information, the initial estimated value general doesn’tsatisfy the constraints of network tomography equation. Based on the service trafficmatrix of the distinguish between business of gravity model, combining the networktomography and inequality constraints and other constraints, the constraint condition ofill-posed inverse problem is increased and the solution space is reduced. Iterativeconstraint algorithm is used to compute the optimal approximation solution that iscloser to the distinguishing between businesses of gravity model initial solution fromthe constraint space.Differentiated service traffic matrix estimation software is accomplished, includingthe data I/O subsystem, the algorithm subsystem and the show I/O subsystem. The dataI/O subsystem get the routing matrix, service link flow matrix and the known OD trafficinformation by reading，writing and processing the user data. The algorithm subsystemincludes distinguish between business of gravity model module, the constraint algorithmmodule and error calculation module. The show subsystem includes data storagemodule and GUI module that is responsible for graphical display of estimation solutionand function of customer interactive.