F. Ionescu and R. Dogaru (Romania)
high performance computing, distributed algorithms, cellular neural networks, CORBA middleware.
The design problem of multi-nested Cellular Neural Network cells is to find a proper set of parameters (gene) for each Boolean function specified by an identifier. This is a very intensive computational task that can be cost effective solved by exploiting available resources in a distributed platform. The architecture of the framework consists of a group of components (servers) distributed on the hosts of a network under the control of CORBA middleware. All servers access a logical shared structure defined as software distributed-shared memory using multiple replicas (copies), updated through remote methods invocation. The communications between distributed components is dynamically controlled in order to avoid the network congestion. The experimental results demonstrate the utility and advantage of this approach that allows significant speedup in solving the required problem with very low costs.
Important Links:
Go Back
