DOI: 10.17587/prin.8.195-206

Estimation of Maximum Number of Simulated Information Flows for Parallel Software Implementation of a Hybrid Fluid Model of Internet Traffic using GPGPU

D. A. Basavin, basavind@gmail.com, Belgorod State Agricultural University named after V. Gorin, Mayskiy, Belgorod region, 308503, Russian Federation, S. V. Porsnev, sergey_porshnev@mail.ru, Ural Federal University named after the first President of Russia B. N. Yeltsin, Ekaterinburg, 620002, Russian Federation, D. A. Petrosov, scorpionss2002@mail.ru, Belgorod State Agricultural University named after V. Gorin, Mayskiy, Belgorod region, 308503, Russian Federation

Corresponding author: Basavin Dmitry A., Assistant, Belgorod State Agricultural University named after V. Gorin, Mayskiy, Belgorod region, 308503, Russian Federation, E-mail: basavind@gmail.com

The article discusses the hybrid fluid model (HFM) of Internet traffic, which describes the information flows in modern computer networks. Information flows in HFM are described by a system of ordinary differential equations (ODE). From the physical point of view this system of ODEs is the result of balance of information flows, which comes into the corresponding node of the network, and goes out of it. Due to the lack of analytical solutions, it is required to develop software tools to obtain numerical solutions of this system of ODEs included in HFM. The experience of practical usage of the well-known software HFM implementation with sequentially executable code showed that the time of calculating the information flows characteristics, even for relatively simple network configurations is quite large, and for the networks with a complex architecture it is unacceptably high. In this regard, it has been hypothesized that it is possible to apply parallel computing techniques based on GPGPU to software HFM implementation development. It should also be noted that the GPGPU technology can accelerate the complex calculations in a few dozen times in a variety of cases. Also, the technology is now one of the leading and the publicly available ones. However, one of its limitations is stiff hardware limitation of available GPU memory. It is therefore necessary to establish consistency between the hardware parameters (the GPU) and the maximum number of simulated information flows (the size of the network and its architecture, respectively). The paper proposes an estimation of the maximum number of simulated information flows via the parallel software HFM implementation based on GPGPU technology.

Keywords: Internet traffic, computer networks, parallel hybrid fluid model, modeling, simulation, GPGPU

pp. 195–206

For citation:
Basavin D. A., Porsnev S. V., Petrosov D. A. Estimation of Maximum Number of Simulated Information Flows for Parallel Software Implementation of a Hybrid Fluid Model of Internet Traffic using GPGPU, Programmnaya Ingeneria, 2017, vol. 8, no. 5, pp. 195-206.