Faster numerical weather forecasting using parallel computing with multi-mesh topology

Abstract

Today for different Meteorological observations attempts are being made for a faster parallel computing method for the purpose of numerical weather forecasting. Highly faster computation technique is required to study raw observation data into easily understandable and accurate weather forecast today. Faster technique will make it possible for meteorological data investigations successfully as well as to run complex weather forecasting models in lesser time, using highly complex set of weather models. The methodology that is used by meteorological study of weather forecasting consists usually of technique called numerical weather prediction. Numerical modeling of atmosphere is a complex process of solving a number of equations involving the future changes of weather variables, like temperature, pressure, humidity, wind speed etc. A simultaneous calculation of different parameters is a necessary part for accurate weather prediction. The multi-mesh topology is getting popular for its efficient topological properties for example: existence of Hamiltonian cycle, simple routing etc. In an n  n multi-mesh network used n 2 meshes of size n  n each, which themselves are again arranged in n rows and n columns so that there will be n 4 processors in total. Each n  n mesh in this network is termed as a block. Using this topology different numerical equation for weather prediction can be solved very faster. This paper presents an efficient and faster way for solving different numerical equation using multi-mesh topology with multiple processors acting simultaneously.