use of edu.iu.dsc.tws.api.resource.WorkerEnvironment in project twister2 by DSC-SPIDAL.
the class BKeyedPartitionExample method compute.
@Override
protected void compute(WorkerEnvironment workerEnv) {
LogicalPlanBuilder logicalPlanBuilder = LogicalPlanBuilder.plan(jobParameters.getSources(), jobParameters.getTargets(), workerEnv);
// create the communication
partition = new BKeyedPartition(workerEnv.getCommunicator(), logicalPlanBuilder, MessageTypes.INTEGER, MessageTypes.INTEGER_ARRAY, new PartitionReceiver(), new SimpleKeyBasedSelector());
Set<Integer> tasksOfExecutor = logicalPlanBuilder.getSourcesOnThisWorker();
// now initialize the workers
this.resultsVerifier = new ResultsVerifier<>(inputDataArray, (ints, args) -> {
int lowestTarget = logicalPlanBuilder.getTargets().stream().min(Comparator.comparingInt(o -> (Integer) o)).get();
int target = Integer.parseInt(args.get("target").toString());
Set<Integer> keysRoutedToThis = new HashSet<>();
for (int i = 0; i < jobParameters.getTotalIterations(); i++) {
if (i % logicalPlanBuilder.getTargets().size() == target - lowestTarget) {
keysRoutedToThis.add(i);
}
}
List<Tuple<Integer, int[]>> expectedData = new ArrayList<>();
for (Integer key : keysRoutedToThis) {
for (int i = 0; i < logicalPlanBuilder.getSources().size(); i++) {
expectedData.add(new Tuple<>(key, ints));
}
}
return expectedData.iterator();
}, new IteratorComparator<>(new TupleComparator<>(// any int
(d1, d2) -> true, IntArrayComparator.getInstance())));
LOG.log(Level.INFO, String.format("%d Sources %s target %d this %s", workerId, logicalPlanBuilder.getSources(), 1, tasksOfExecutor));
for (int t : tasksOfExecutor) {
// the map thread where data is produced
Thread mapThread = new Thread(new KeyedBenchWorker.MapWorker(t));
mapThread.start();
}
}
use of edu.iu.dsc.tws.api.resource.WorkerEnvironment in project twister2 by DSC-SPIDAL.
the class BKeyedGatherExample method compute.
@Override
protected void compute(WorkerEnvironment workerEnv) {
LogicalPlanBuilder logicalPlanBuilder = LogicalPlanBuilder.plan(jobParameters.getSources(), jobParameters.getTargets(), workerEnv);
// create the communication
keyedGather = new BKeyedGather(workerEnv.getCommunicator(), logicalPlanBuilder, MessageTypes.INTEGER, MessageTypes.INTEGER_ARRAY, new FinalReduceReceiver(), new SimpleKeyBasedSelector());
Set<Integer> tasksOfExecutor = logicalPlanBuilder.getSourcesOnThisWorker();
for (int t : tasksOfExecutor) {
finishedSources.put(t, false);
}
if (tasksOfExecutor.size() == 0) {
sourcesDone = true;
}
this.resultsVerifier = new ResultsVerifier<>(inputDataArray, (ints, args) -> {
int lowestTarget = logicalPlanBuilder.getTargets().stream().min(Comparator.comparingInt(o -> (Integer) o)).get();
int target = Integer.valueOf(args.get("target").toString());
Set<Integer> keysRoutedToThis = new HashSet<>();
for (int i = 0; i < jobParameters.getTotalIterations(); i++) {
if (i % logicalPlanBuilder.getTargets().size() == target - lowestTarget) {
keysRoutedToThis.add(i);
}
}
List<int[]> dataForEachKey = new ArrayList<>();
for (int i = 0; i < logicalPlanBuilder.getSources().size(); i++) {
dataForEachKey.add(ints);
}
List<Tuple<Integer, Iterator<int[]>>> expectedData = new ArrayList<>();
for (Integer key : keysRoutedToThis) {
expectedData.add(new Tuple<>(key, dataForEachKey.iterator()));
}
return expectedData.iterator();
}, new IteratorComparator<>(new TupleComparator<>(// any int
(d1, d2) -> true, new IteratorComparator<>(IntArrayComparator.getInstance()))));
LOG.log(Level.INFO, String.format("%d Sources %s target %d this %s", workerId, logicalPlanBuilder.getSources(), 1, tasksOfExecutor));
// now initialize the workers
for (int t : tasksOfExecutor) {
// the map thread where data is produced
Thread mapThread = new Thread(new MapWorker(t));
mapThread.start();
}
}
use of edu.iu.dsc.tws.api.resource.WorkerEnvironment in project twister2 by DSC-SPIDAL.
the class BReduceExample method compute.
@Override
protected void compute(WorkerEnvironment workerEnv) {
LogicalPlanBuilder logicalPlanBuilder = LogicalPlanBuilder.plan(jobParameters.getSources(), jobParameters.getTargets(), workerEnv).withFairDistribution();
// create the communication
reduce = new BReduce(workerEnv.getCommunicator(), logicalPlanBuilder, new ReduceOperationFunction(Op.SUM, MessageTypes.INTEGER_ARRAY), new FinalSingularReceiver(), MessageTypes.INTEGER_ARRAY);
Set<Integer> tasksOfExecutor = logicalPlanBuilder.getSourcesOnThisWorker();
for (int t : tasksOfExecutor) {
finishedSources.put(t, false);
}
if (tasksOfExecutor.size() == 0) {
sourcesDone = true;
}
this.resultsVerifier = new ResultsVerifier<>(inputDataArray, (ints, args) -> GeneratorUtils.multiplyIntArray(ints, jobParameters.getTotalIterations() * logicalPlanBuilder.getSources().size()), IntArrayComparator.getInstance());
LOG.log(Level.INFO, String.format("%d Sources %s target %d this %s", workerId, logicalPlanBuilder.getSources(), logicalPlanBuilder.getTargets().iterator().next(), tasksOfExecutor));
// now initialize the workers
for (int t : tasksOfExecutor) {
// the map thread where data is produced
Thread mapThread = new Thread(new MapWorker(t));
mapThread.start();
}
}
use of edu.iu.dsc.tws.api.resource.WorkerEnvironment in project twister2 by DSC-SPIDAL.
the class ArrowTSetSourceExample method execute.
@Override
public void execute(WorkerEnvironment workerEnv) {
BatchEnvironment env = TSetEnvironment.initBatch(workerEnv);
Config config = env.getConfig();
String csvInputDirectory = config.getStringValue(DataObjectConstants.DINPUT_DIRECTORY);
String arrowInputDirectory = config.getStringValue(DataObjectConstants.ARROW_DIRECTORY);
String arrowFileName = config.getStringValue(DataObjectConstants.FILE_NAME);
int workers = config.getIntegerValue(DataObjectConstants.WORKERS);
int parallel = config.getIntegerValue(DataObjectConstants.PARALLELISM_VALUE);
int dsize = config.getIntegerValue(DataObjectConstants.DSIZE);
LOG.info("arrow input file:" + arrowFileName + "\t" + arrowInputDirectory + "\t" + csvInputDirectory + "\t" + workers + "\t" + parallel);
Schema schema = makeSchema();
SourceTSet<String[]> csvSource = env.createCSVSource(csvInputDirectory, dsize, parallel, "split");
SinkTSet<Iterator<Integer>> sinkTSet = csvSource.direct().map((MapFunc<String[], Integer>) input -> Integer.parseInt(input[0])).direct().sink(new ArrowBasedSinkFunction<>(arrowInputDirectory, arrowFileName, schema.toJson()));
env.run(sinkTSet);
// Source Function Call
env.createArrowSource(arrowInputDirectory, arrowFileName, parallel, schema.toJson()).direct().compute((ComputeFunc<Iterator<Object>, List<Integer>>) input -> {
List<Integer> integers = new ArrayList<>();
input.forEachRemaining(i -> integers.add((Integer) i));
return integers;
}).direct().forEach(s -> LOG.info("Integer Array Size:" + s.size() + "\tvalues:" + s));
}
use of edu.iu.dsc.tws.api.resource.WorkerEnvironment in project twister2 by DSC-SPIDAL.
the class ReduceExample method execute.
@Override
public void execute(WorkerEnvironment workerEnv) {
BatchEnvironment env = TSetEnvironment.initBatch(workerEnv);
int start = env.getWorkerID() * 100;
SourceTSet<Integer> src = dummySource(env, start, COUNT, PARALLELISM);
ReduceTLink<Integer> reduce = src.reduce(Integer::sum);
LOG.info("test foreach");
reduce.forEach(i -> LOG.info("foreach: " + i));
LOG.info("test map");
reduce.map(i -> i.toString() + "$$").withSchema(PrimitiveSchemas.STRING).direct().forEach(s -> LOG.info("map: " + s));
LOG.info("test flat map");
reduce.flatmap((i, c) -> c.collect(i.toString() + "##")).withSchema(PrimitiveSchemas.STRING).direct().forEach(s -> LOG.info("flat:" + s));
LOG.info("test compute");
reduce.compute((ComputeFunc<Integer, String>) input -> "sum=" + input).withSchema(PrimitiveSchemas.STRING).direct().forEach(s -> LOG.info("compute: " + s));
LOG.info("test computec");
reduce.compute((ComputeCollectorFunc<Integer, String>) (input, output) -> output.collect("sum=" + input)).withSchema(PrimitiveSchemas.STRING).direct().forEach(s -> LOG.info("computec: " + s));
LOG.info("test map2tup");
reduce.mapToTuple(i -> new Tuple<>(i, i.toString())).keyedDirect().forEach(i -> LOG.info("mapToTuple: " + i.toString()));
LOG.info("test sink");
SinkTSet<Integer> sink = reduce.sink((SinkFunc<Integer>) value -> {
LOG.info("val =" + value);
return true;
});
env.run(sink);
}
Aggregations