Examples with TopologyException org.ow2.proactive.resourcemanager.frontend.topology.TopologyException used on opensource projects

Example 1 with TopologyException

use of org.ow2.proactive.resourcemanager.frontend.topology.TopologyException in project scheduling by ow2-proactive.

the class HAC method select.

/**
 * Selects a set of closest nodes from specified list.
 * The proximity between nodes is defined by distance function.
 *
 * @param number desired nodes number
 * @param from list of "free" nodes
 * @return list of nodes to be provided to the client
 */
public List<Node> select(int number, List<Node> from) {
    if (from.size() == 0) {
        // return empty list
        return new LinkedList<>();
    }
    // initializing cluster distances map
    // cluster is a group of nodes, initially each cluster consist of one node
    logger.debug("Initializing clusters map");
    HashMap<Cluster<Node>, HashMap<Cluster<Node>, Long>> clusterDistances = initClusterDistances(from);
    // no topology information for provided nodes
    if (from.size() > 0 && clusterDistances.size() == 0) {
        throw new TopologyException("Topology information is not available");
    }
    Cluster<Node> target = null;
    if (pivot.size() > 0) {
        // fixed orientation clustering
        Iterator<Node> it = pivot.iterator();
        Node pivotNode = it.next();
        target = new Cluster<>(getNodeId(pivotNode), pivotNode);
        // merging pivot nodes into one cluster and recalculating distances
        logger.debug("Merging pivot nodes into one cluster");
        while (it.hasNext()) {
            // merging clusters and recalculating distances between others
            pivotNode = it.next();
            Cluster<Node> pivotCluster = new Cluster<>(getNodeId(pivotNode), pivotNode);
            target = recalculateDistances(target, pivotCluster, clusterDistances);
        }
        // clustering centralized to the pivot
        logger.debug("Begin centralized hierarchical agglomerative clustering");
        while (clusterDistances.size() > 1 && target.size() < (number + pivot.size())) {
            Cluster<Node> closest = findClosestClustersTo(target, clusterDistances);
            if (closest == null) {
                // no clusters found => cannot merge anything => stop where we are
                break;
            }
            // merging clusters and recalculating distances between others
            target = recalculateDistances(target, closest, clusterDistances);
        }
        // removing pivot nodes from the result
        target.remove(pivot);
    } else {
        logger.debug("Begin hierarchical agglomerative clustering");
        target = (Cluster<Node>) clusterDistances.keySet().iterator().next();
        Cluster<Node> largest = target;
        // floating clustering
        while (clusterDistances.size() > 1) {
            // finding two clusters to merge according
            Cluster<Node>[] clustersToMerge = findClosestClusters(clusterDistances);
            if (clustersToMerge == null) {
                // stop the process
                break;
            }
            // merging clusters and recalculating distances between others
            target = recalculateDistances(clustersToMerge[0], clustersToMerge[1], clusterDistances);
            if (target.size() >= largest.size()) {
                largest = target;
            }
            if (target.size() == number) {
                // found all the nodes we need
                break;
            } else if (target.size() > number) {
                // found more nodes that we need,
                // target cluster contains all nodes from another cluster
                // largest is the target here
                logger.debug("Number of node in the cluster exceeded required node number " + target.size() + " vs " + number);
                Cluster<Node> anotherCluster = clustersToMerge[0] == target ? clustersToMerge[1] : clustersToMerge[0];
                target.removeLast(anotherCluster.size());
                final Cluster<Node> finalTarget = target;
                Comparator<Node> nodeDistanceComparator = new Comparator<Node>() {

                    public int compare(Node n1, Node n2) {
                        long res = getDistance(n1, finalTarget) - getDistance(n2, finalTarget);
                        if (res < 0) {
                            return -1;
                        } else if (res > 0) {
                            return 1;
                        } else {
                            return 0;
                        }
                    }
                };
                // sorting nodes in the smaller cluster according to their distances to target
                Collections.sort(anotherCluster.getElements(), nodeDistanceComparator);
                int neededNodesNumber = number - target.size();
                target.add(anotherCluster.getElements().subList(0, neededNodesNumber));
                break;
            }
        }
        target = largest;
    }
    if (logger.isDebugEnabled()) {
        logger.debug("Found " + target.size() + " nodes out of " + number + ": " + target);
    }
    return target.getElements();
}