Dynamic Arrival Process for TCP Using a Novel Time-Dependant Statistical Calculation Approach

This paper proposes a dynamic arrival process to represent a typical TCP source in Queueing theory performance modeling. This process utilises a novel time-dependant statistical calculation approach, which modifies and extends the traditional method of modeling and calculating Geo/Geo/1/J (discrete-time M/M/1/J) queues. The main contributions of this paper are: (1) The limitation of traditional method of modeling and calculation is analysed and possibility for extension is discussed, (2) The new calculation approach is demonstrated in detail, (3) The new dynamic arrival process and the related modeling system is illustrated thoroughly, (4) The effectiveness of the proposed random process is evaluated by a set of extensive comparisons between groups of closely matching results of MATLAB numerical solutions of a model which employ the new dynamic arrival process, and NS-2 simulations of equivalent scenarios, (5) The limitation of the current work and future extension plan is discussed and summarised.