The deployment plan of components is a very important factor that influences the throughput of the environment. Thus a good planning approach is needed to arrange the resources in such a manner that when the components are deployed on the resources, the maximum number of requests can be processed in a time step. We called this process deployment planning. We have shown that the optimal deployment on a cluster is a Complete Spanning d-ary (CSD) tree; a CSD tree is a tree that is both a complete d-ary tree and a spanning tree. This result conforms with results from the scheduling literature. More importantly, we have presented an approach for determining the optimal degree d for the tree. Finding the best deployment among heterogeneous resources is a hard problem since it amounts to finding the best broadcast tree on a general graph, which is known to be NP-complete. So we presented a deployment heuristic that predicts the maximum throughput that can be achieved by the use of available nodes. The main focus of the heuristic is to construct an hierarchy so as to maximize the throughput of each node, where the throughput depends on the number of children attached as children to the node in the hierarchy. The given heuristic provides a deployment that can meet the user request demand, if user demand is at most equal to the maximum throughput.
We have designed a mathematical model that can analyze an existing deployment and can improve the performance of the deployment by finding and then removing the bottlenecks. This is an heuristic approach for improving deployments of NES environments in heterogeneous grid environments. The heuristic is used to improve the throughput of a deployment that has been defined by other means. The approach is iterative: in each iteration, mathematical models are used to analyze the existing deployment to identify the primary bottleneck, and the bottleneck is then removed by adding resources in the appropriate area of the hierarchy. Using this model we can evaluate a virtual deployment before making a real deployment, provide a decision builder tool (i.e., designed to compare different hierarchies or add new resources) and take into account the hierarchies’ scalability.