90 likes | 243 Views
Cluster Computing and Genetic Algorithms With ClusterKnoppix. David Tabachnick. What is a Cluster?. A Cluster is… A group of connected computers, acting as one to calculate computationally difficult problems. A Cluster is… Made up of a server , nodes , and networking hub(s) .
E N D
Cluster Computing and Genetic AlgorithmsWith ClusterKnoppix David Tabachnick
What is a Cluster? A Cluster is… A group of connected computers, acting as one to calculate computationally difficult problems. A Cluster is… Made up of a server, nodes, and networking hub(s).
Basics of Clustering • Server is the head of the cluster. • Nodes are all connected via hub. • Server is started first. The nodes are then powered on, and they look to the server for instructions. • The cluster is both scalable and highly available. • The cluster can be taken down and restarted quickly and easily.
Methods to Clustering • Load-balancing is done on the server by a software package, OpenMosix. • This results in breaking programs into processes that can be sent to the nodes for processing. • No special programming methods or languages must be used.
Features of this Cluster • Live Installwith configuration scripts • Remote access for authorized users • High scalability • Extremely secure • Local storage on the server and external storage for redundant data backup • Will run LISP
Genetic Algorithms on a Cluster • Server performs the string manipulation and sends the matrix calculations to the nodes. • Each node is given one fitness function to calculate. • Upon completion, a node will write back the resultant function and remain idle. • ≈80% of the calculations take place on the nodes.
Conclusion • This cluster… • Achieved relatively high performance when compared to other clusters of similar size and scope. • Was designed with computing genetic algorithms in mind. • Was built to be scalable, flexible, and easily accessible.