Abstract
A computer cluster consists of a number of computers that are coupled to solve complex problems in parallel. They provide a single front-end so that they can be managed and used from a single user interface. The cluster not only provides a excellent cost-performance trade-off compared to a super computer but it also provides the flexibility of using less power when the computational load is lower. Its performance is about the same as a super computer having computational power equal to combined computational capacity of all the computers in the cluster, but at a much lower cost.
This project provides a proof of concept in our capability to build a cluster using open source Message Passing Interface (MPI) at the NSU’s Network, Security and Communications laboratory. We have implemented a cluster using old machines, and verified the correct operation of the cluster by running a complex nonlinear dynamics problem. If faster machines are available, this team is ready to implement a powerful cluster that could be used to handle complex computational problems in areas such as prediction, data mining, physics, cryptography, number theory, and protein folding.
Faculty Sponsors
Dr. Saeed Rajput, Dr. Victor Castro
Project Type
Event
Location
Alvin Sherman Library
Start Date
4-4-2008 12:00 AM
End Date
4-4-2008 12:00 AM
Construction and Configuration of MPI Cluster
Alvin Sherman Library
A computer cluster consists of a number of computers that are coupled to solve complex problems in parallel. They provide a single front-end so that they can be managed and used from a single user interface. The cluster not only provides a excellent cost-performance trade-off compared to a super computer but it also provides the flexibility of using less power when the computational load is lower. Its performance is about the same as a super computer having computational power equal to combined computational capacity of all the computers in the cluster, but at a much lower cost.
This project provides a proof of concept in our capability to build a cluster using open source Message Passing Interface (MPI) at the NSU’s Network, Security and Communications laboratory. We have implemented a cluster using old machines, and verified the correct operation of the cluster by running a complex nonlinear dynamics problem. If faster machines are available, this team is ready to implement a powerful cluster that could be used to handle complex computational problems in areas such as prediction, data mining, physics, cryptography, number theory, and protein folding.
