MohammadReza Majma - Islamic Azad University, Pardis Branch
Hossein Pedram - Amirkabir University of Technology, Computer Engineering Department
Bijan Fallahi - National Cloud Seeding Research Center (NACSER)
Sanaz Almassi - Iran University of Science and Technology, Computer engineering Department

Parallel processing and distributed systems, have efficient role in different sciences such as meteorology, medicine, nuclear, physics, chemistry, etc. These sciences need parallel processing systems based on cluster to solve and run their algorithms with a higher speed. Some tools and models exist for parallel programming such as shared memory, message passing, oriented model, hybrid model, java, but message passing by MPI library plays an essential role in cluster systems. In this paper, we have presented WRF model with a message passing interface. We have implemented parallel algorithm of this model in a real cluster and evaluated the results in comparison with different forecasting models. Experimental results demonstrated that by increasing the number of processors to 24, run time decreases from 2500 seconds to 600 seconds, however the performance decrease to 17%.

Cluster, Interface WRF model, Parallel Processing, Message Passing, Meteorology