End-to-End Delay Estimations in IP networks

Introduction

Quality of Service (QoS) is an important factor in telecommunications networks. For the time division multiplexing systems, comprehensive methods of the QoS attribute analysis were elaborated. For the networks based on the packet technologies, there is a number of new problems related to the QoS analysis. One of the complicated problems is an estimation of the packets delays between user's terminals. In this case, end-to-end delay of the transmitted IP packets has to be estimated. The purpose of the research is to create the analytical methods of the end-to-end delay estimation for the IP networks.

State of the art

Time delivery of IP packets can be considered as a random quantity. According to standardization of the QoS parameters proposed by ITU and ETSI, two characteristics of the time delivery are interesting. The first one is mean value of the analyzable random quantity. A quantile of the corresponding distribution function (jitter) is the second characteristic. In the tasks of end-to-end delay’s estimation the chain of routers between two terminals has to be considered as multiphase queuing system. 
NGN is an important sort of the IP networks. ITU recommendations related to the NGN assume the priority-service disciplines for the IP packets’ delivering. In case of priorities, an analysis of the multiphase queuing systems becomes more complex.
Model of the multiphase queuing system can be studied by the analytical methods or simulation. For the analysis of the model both approaches are widely used. Calculations of the mean value and distribution function are examples of the direct problem. For this problem, simulation appears to be efficient method for complicated models while analytical results can not be obtained. Typical example of the inverse problems is calculation of the service rate μ for the known offered traffic λ and established QoS parameters. For the inverse problems, analytical methods are preferable.

Research directions

First of all, main characteristics of the arrival process have to be defined. For telephone traffic arrival process can be described by the Poisson process. In cases of data and video traffic such assumption will be not acceptable. 
Secondly, main characteristics of the service time have to be selected. In general, a service time distribution is described by the step function. On the other hand, some approximations of this function are useful. 
In general, recently Pareto, Weibull and other self-similar processes are applied for a description of random processes in queuing systems. 
Finally, it is needed to choose priority-service disciplines. There are a number of solutions related to QoS provision. Undoubtedly, applied priority-service disciplines have to comply with international standards. 
The main mathematical method is based on the teletraffic theory. Laplace transform and some rough-and-ready methods will used as well. Simulation will be applied for the estimation of accuracy of the accepted assumptions. In addition, for estimation of accuracy the real measurement data of the IP traffic should be used.

Project results

Methods of the multiphase queuing systems analysis should have theoretical value. These methods will be useful for analysis of the complicated models describing a number of the telecommunications systems. As practical value some results will be useful for the NGN planning.

Contacts

Contact person: Andrew Sokolov