Super Cluster system
Based on our studies and experience, we designed a Super Cluster system which
will be the most powerful supercomputer have been ever made in Iran. It will
consist of 256 computing nodes and some special purpose nodes for management,
check-pointing, parallel file system and so on. Its peak performance will be
about 1500 billion operations per second, and it will serve remote clients to
run their parallel programs. Considered aspects of design are: Architecture
issue, Workload management, Parallel I/O, High availability, Hardware management
and packaging.
The main aims of building this Super Cluster are as follows:
1. Establishment a service center to serve high performance computing needs.
In many countries, there are some service centers that serve other institutes or research centers to run their time-consuming programs on their supercomputer. We are going to establish such a center in Iran to serve institutes that have temporary high performance computing needs, or can not afford to buy a supercomputer.
2. Improvement of parallel processing knowledge
By building the Super Cluster system, our parallel processing knowledge will be evolved. On the other hand it can be an ideal test-bed for parallel algorithms, so it will help to evolve computational science in the country on the whole.3. Developing useful parallel applications
Some useful and valuable applications need supercomputing power to be developed. The designed Super Cluster will help many research centers to develop their applied programs.