Examples with ComputeGraph edu.iu.dsc.tws.api.compute.graph.ComputeGraph used on opensource projects

Example 61 with ComputeGraph

use of edu.iu.dsc.tws.api.compute.graph.ComputeGraph in project twister2 by DSC-SPIDAL.

the class KMeansCheckpointingWorker method execute.

/**
 * First, the execute method invokes the generateDataPoints method to generate the datapoints file
 * and centroid file based on the respective filesystem submitted by the user. Next, it invoke
 * the DataObjectSource and DataObjectSink to partition and read the partitioned data points
 * respectively through data points task graph. Then, it calls the DataFileReader to read the
 * centroid values from the filesystem through centroid task graph. Next, the datapoints are
 * stored in DataSet \(0th object\) and centroids are stored in DataSet 1st object\). Finally, it
 * constructs the kmeans task graph to perform the clustering process which computes the distance
 * between the centroids and data points.
 */
@SuppressWarnings("unchecked")
@Override
public void execute(WorkerEnvironment workerEnv) {
    int workerId = workerEnv.getWorkerId();
    Config config = workerEnv.getConfig();
    IWorkerController workerController = workerEnv.getWorkerController();
    ComputeEnvironment taskEnv = ComputeEnvironment.init(workerEnv);
    CheckpointingWorkerEnv checkpointingEnv = CheckpointingWorkerEnv.newBuilder(config, workerId, workerController).registerVariable(I_KEY, IntegerPacker.getInstance()).registerVariable(CENT_OBJ, ObjectPacker.getInstance()).build();
    Snapshot snapshot = checkpointingEnv.getSnapshot();
    TaskExecutor taskExecutor = taskEnv.getTaskExecutor();
    LOG.info("Task worker starting: " + workerId + " Current snapshot ver: " + snapshot.getVersion());
    int parallelismValue = config.getIntegerValue(DataObjectConstants.PARALLELISM_VALUE);
    int dimension = config.getIntegerValue(DataObjectConstants.DIMENSIONS);
    int numFiles = config.getIntegerValue(DataObjectConstants.NUMBER_OF_FILES);
    int dsize = config.getIntegerValue(DataObjectConstants.DSIZE);
    int csize = config.getIntegerValue(DataObjectConstants.CSIZE);
    int iterations = config.getIntegerValue(DataObjectConstants.ARGS_ITERATIONS);
    String dataDirectory = config.getStringValue(DataObjectConstants.DINPUT_DIRECTORY) + workerId;
    String centroidDirectory = config.getStringValue(DataObjectConstants.CINPUT_DIRECTORY) + workerId;
    String type = config.getStringValue(DataObjectConstants.FILE_TYPE);
    KMeansUtils.generateDataPoints(config, dimension, numFiles, dsize, csize, dataDirectory, centroidDirectory, type);
    long startTime = System.currentTimeMillis();
    /* First Graph to partition and read the partitioned data points **/
    ComputeGraph datapointsTaskGraph = KMeansComputeJob.buildDataPointsTG(dataDirectory, dsize, parallelismValue, dimension, config, type);
    // Get the execution plan for the first task graph
    ExecutionPlan datapointsExecutionPlan = taskExecutor.plan(datapointsTaskGraph);
    // Actual execution for the first taskgraph
    taskExecutor.execute(datapointsTaskGraph, datapointsExecutionPlan);
    // Retrieve the output of the first task graph
    DataObject<Object> dataPointsObject = taskExecutor.getOutput(datapointsTaskGraph, datapointsExecutionPlan, "datapointsink");
    DataObject<Object> centroidsDataObject;
    if (!snapshot.checkpointAvailable(CENT_OBJ)) {
        /* Second Graph to read the centroids **/
        ComputeGraph centroidsTaskGraph = KMeansComputeJob.buildCentroidsTG(centroidDirectory, csize, parallelismValue, dimension, config, type);
        // Get the execution plan for the second task graph
        ExecutionPlan centroidsExecutionPlan = taskExecutor.plan(centroidsTaskGraph);
        // Actual execution for the second taskgraph
        taskExecutor.execute(centroidsTaskGraph, centroidsExecutionPlan);
        // Retrieve the output of the first task graph
        centroidsDataObject = taskExecutor.getOutput(centroidsTaskGraph, centroidsExecutionPlan, "centroidsink");
    } else {
        centroidsDataObject = (DataObject<Object>) snapshot.get(CENT_OBJ);
    }
    long endTimeData = System.currentTimeMillis();
    /* Third Graph to do the actual calculation **/
    ComputeGraph kmeansTaskGraph = KMeansComputeJob.buildKMeansTG(parallelismValue, config);
    // Perform the iterations from 0 to 'n' number of iterations
    IExecutor ex = taskExecutor.createExecution(kmeansTaskGraph);
    for (int i = 0; i < iterations; i++) {
        // actual execution of the third task graph
        ex.execute(i == iterations - 1);
    }
    DataPartition<?> centroidPartition = centroidsDataObject.getPartition(workerId);
    double[][] centroid = (double[][]) centroidPartition.getConsumer().next();
    long endTime = System.currentTimeMillis();
    if (workerId == 0) {
        LOG.info("Data Load time : " + (endTimeData - startTime) + "\n" + "Total Time : " + (endTime - startTime) + "Compute Time : " + (endTime - endTimeData));
    }
    LOG.info("Final Centroids After\t" + iterations + "\titerations\t" + Arrays.deepToString(centroid));
    taskEnv.close();
}

Example 62 with ComputeGraph

use of edu.iu.dsc.tws.api.compute.graph.ComputeGraph in project twister2 by DSC-SPIDAL.

the class WordCountJob method execute.

@Override
public void execute(WorkerEnvironment workerEnv) {
    ComputeEnvironment cEnv = ComputeEnvironment.init(workerEnv);
    // source and aggregator
    WordSource source = new WordSource();
    WordAggregator counter = new WordAggregator();
    // build the task graph
    ComputeGraphBuilder builder = ComputeGraphBuilder.newBuilder(workerEnv.getConfig());
    builder.addSource("word-source", source, 4);
    builder.addCompute("word-aggregator", counter, 4).keyedReduce("word-source").viaEdge(EDGE).withReductionFunction(new ReduceFn(Op.SUM, MessageTypes.INTEGER_ARRAY)).withKeyType(MessageTypes.OBJECT).withDataType(MessageTypes.INTEGER_ARRAY);
    builder.setMode(OperationMode.BATCH);
    // execute the graph
    ComputeGraph graph = builder.build();
    ExecutionPlan plan = cEnv.getTaskExecutor().plan(graph);
    cEnv.getTaskExecutor().execute(graph, plan);
}

Example 63 with ComputeGraph

use of edu.iu.dsc.tws.api.compute.graph.ComputeGraph in project twister2 by DSC-SPIDAL.

the class TwoDataFlowsExample method runFirstJob.

private static void runFirstJob(Config config, CDFWEnv cdfwEnv, int parallelism, DataFlowJobConfig jobConfig) {
    FirstSourceTask firstSourceTask = new FirstSourceTask();
    ConnectedSink connectedSink = new ConnectedSink("first_out");
    ComputeGraphBuilder graphBuilderX = ComputeGraphBuilder.newBuilder(config);
    graphBuilderX.addSource("source1", firstSourceTask, parallelism);
    ComputeConnection partitionConnection = graphBuilderX.addCompute("sink1", connectedSink, parallelism);
    partitionConnection.partition("source1").viaEdge("partition").withDataType(MessageTypes.OBJECT);
    graphBuilderX.setMode(OperationMode.BATCH);
    ComputeGraph batchGraph = graphBuilderX.build();
    DataFlowGraph job = DataFlowGraph.newSubGraphJob("first_graph", batchGraph).setWorkers(4).addDataFlowJobConfig(jobConfig).setGraphType("non-iterative");
    cdfwEnv.executeDataFlowGraph(job);
}

Example 64 with ComputeGraph

use of edu.iu.dsc.tws.api.compute.graph.ComputeGraph in project twister2 by DSC-SPIDAL.

the class TwoDataFlowsExample method runSecondJob.

private static void runSecondJob(Config config, CDFWEnv cdfwEnv, int parallelism, DataFlowJobConfig jobConfig) {
    ConnectedSource connectedSource = new ConnectedSource("reduce", "first_out");
    ConnectedSink connectedSink = new ConnectedSink();
    ComputeGraphBuilder graphBuilderX = ComputeGraphBuilder.newBuilder(config);
    graphBuilderX.addSource("source2", connectedSource, parallelism);
    ComputeConnection reduceConn = graphBuilderX.addCompute("sink2", connectedSink, 1);
    reduceConn.reduce("source2").viaEdge("reduce").withReductionFunction(new Aggregator()).withDataType(MessageTypes.OBJECT);
    graphBuilderX.setMode(OperationMode.BATCH);
    ComputeGraph batchGraph = graphBuilderX.build();
    DataFlowGraph job = DataFlowGraph.newSubGraphJob("second_graph", batchGraph).setWorkers(4).addDataFlowJobConfig(jobConfig).setGraphType("non-iterative");
    cdfwEnv.executeDataFlowGraph(job);
}

Example 65 with ComputeGraph

use of edu.iu.dsc.tws.api.compute.graph.ComputeGraph in project twister2 by DSC-SPIDAL.

the class KMeansComputeJob method execute.

/**
 * First, the execute method invokes the generateDataPoints method to generate the datapoints file
 * and centroid file based on the respective filesystem submitted by the user. Next, it invoke
 * the DataObjectSource and DataObjectSink to partition and read the partitioned data points
 * respectively through data points task graph. Then, it calls the DataFileReader to read the
 * centroid values from the filesystem through centroid task graph. Next, the datapoints are
 * stored in DataSet \(0th object\) and centroids are stored in DataSet 1st object\). Finally, it
 * constructs the kmeans task graph to perform the clustering process which computes the distance
 * between the centroids and data points.
 */
@SuppressWarnings("unchecked")
@Override
public void execute(WorkerEnvironment workerEnv) {
    int workerId = workerEnv.getWorkerId();
    Config config = workerEnv.getConfig();
    LOG.log(Level.FINE, "Task worker starting: " + workerId);
    ComputeEnvironment cEnv = ComputeEnvironment.init(workerEnv);
    TaskExecutor taskExecutor = cEnv.getTaskExecutor();
    int parallelismValue = config.getIntegerValue(DataObjectConstants.PARALLELISM_VALUE);
    int dimension = config.getIntegerValue(DataObjectConstants.DIMENSIONS);
    int numFiles = config.getIntegerValue(DataObjectConstants.NUMBER_OF_FILES);
    int dsize = config.getIntegerValue(DataObjectConstants.DSIZE);
    int csize = config.getIntegerValue(DataObjectConstants.CSIZE);
    int iterations = config.getIntegerValue(DataObjectConstants.ARGS_ITERATIONS);
    String dataDirectory = config.getStringValue(DataObjectConstants.DINPUT_DIRECTORY) + workerId;
    String centroidDirectory = config.getStringValue(DataObjectConstants.CINPUT_DIRECTORY) + workerId;
    String type = config.getStringValue(DataObjectConstants.FILE_TYPE);
    KMeansUtils.generateDataPoints(config, dimension, numFiles, dsize, csize, dataDirectory, centroidDirectory, type);
    long startTime = System.currentTimeMillis();
    /* First Graph to partition and read the partitioned data points **/
    ComputeGraph datapointsTaskGraph = buildDataPointsTG(dataDirectory, dsize, parallelismValue, dimension, config, type);
    /* Second Graph to read the centroids **/
    ComputeGraph centroidsTaskGraph = buildCentroidsTG(centroidDirectory, csize, parallelismValue, dimension, config, type);
    /* Third Graph to do the actual calculation **/
    ComputeGraph kmeansTaskGraph = buildKMeansTG(parallelismValue, config);
    // Get the execution plan for the first task graph
    ExecutionPlan firstGraphExecutionPlan = taskExecutor.plan(datapointsTaskGraph);
    // Actual execution for the first taskgraph
    taskExecutor.execute(datapointsTaskGraph, firstGraphExecutionPlan);
    // Get the execution plan for the second task graph
    ExecutionPlan secondGraphExecutionPlan = taskExecutor.plan(centroidsTaskGraph);
    // Actual execution for the second taskgraph
    taskExecutor.execute(centroidsTaskGraph, secondGraphExecutionPlan);
    long endTimeData = System.currentTimeMillis();
    // Perform the iterations from 0 to 'n' number of iterations
    IExecutor ex = taskExecutor.createExecution(kmeansTaskGraph);
    for (int i = 0; i < iterations; i++) {
        // actual execution of the third task graph
        ex.execute(i == iterations - 1);
    }
    cEnv.close();
    long endTime = System.currentTimeMillis();
    LOG.info("Total K-Means Execution Time: " + (endTime - startTime) + "\tData Load time : " + (endTimeData - startTime) + "\tCompute Time : " + (endTime - endTimeData));
}