We designed and implemented a Cluster System for Iranian Meteorological Organization (IRIMO). Based on the needs of IRIMO, it consists of 48 computing nodes, 2 special nodes, and 4 spare nodes. Special nodes are a “Master & Configure node” and a “Storage & Login node”. The main application of the system is MM5 program which fully supports both MPI and OpenMP; hence, it can effectively deploy the computing resources of the system.

Compute nodes are the computational heart of the cluster system. The duty of Master and Configure node is to monitor and manage the whole activities of cluster and the related components. In addition to above, the administrator of the system can install and configure dedicated software on compute nodes through this node. The Storage and Login Node with a storage space of about 1TB is assigned dual duties. First, not only can the system’s users benefit from this space as storage space, but also the compute nodes can use it as a shared storage. RAID 5 was selected to improve the reliability of the system while keeping the cost of the design low. Moreover, this node plays the role of login node for the system. In other words, it is a gateway from outside to the cluster. Users must login to this node to compile or run their jobs.

A variety of useful software is installed on the system to ease effective operations. Since we have employed standard OS, compiler, and MPI library for our system, all supporting weather applications including ALADIN, VGA, ARPS, and MM5 can be installed easily. Moreover, we have implemented a user interface which is a graphical web-based interface to monitor and manage the cluster system. The core engine of this GUI is developed by php, and enjoys two modes: Administrator and User.