org.jppf.load.balancer.impl.AutoTunedBundler Maven / Gradle / Ivy

Go to download
Show more of this group Show more artifacts with this name
Show all versions of jppf-common Show documentation
JPPF, the open source grid computing solution
There is a newer version: 6.3-alpha
/*
 * JPPF.
 * Copyright (C) 2005-2015 JPPF Team.
 * http://www.jppf.org
 *
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 *   http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */
package org.jppf.load.balancer.impl;

import org.jppf.load.balancer.*;
import org.slf4j.*;

/**
 * This class implements a self tuned bundle size algorithm. It starts using the
 * bundle size defined in property file and starts changing it to find a better
 * performance. The algorithm starts making The algorithm waits for some
 * execution to get a mean execution time, and them make a change in bundle size
 * Each time the change is done, it is done over a smaller range randomly
 * selected (like Monte Carlo algorithm).
 * 
 * @author Domingos Creado
 * @author Laurent Cohen
 * @exclude
 */
public class AutoTunedBundler extends AbstractAutoTunedBundler {
  /**
   * Logger for this class.
   */
  private static Logger log = LoggerFactory.getLogger(AutoTunedBundler.class);
  /**
   * Determines whether trace level is set for logging.
   */
  private static boolean traceEnabled = log.isTraceEnabled();

  /**
   * Creates a new instance with the initial size of bundle as the start size.
   * @param profile the parameters of the auto-tuning algorithm,
   * grouped as a performance analysis profile.
   */
  public AutoTunedBundler(final LoadBalancingProfile profile) {
    super((AnnealingTuneProfile) profile);
  }

  /**
   * This method performs the actual bundle size computation, based on current and past
   * performance data.

   * Depending on the the performance samples and profile parameters, the following actions
   * may be triggered in this method:
   * 
   * samples collection (unconditional)
   * detection of performance profile changes, if not currently being done
   * when a performance profile change is detected, recompute the bundle size.
   * 
   * @param bundleSize bundle size of the new performance sample.
   * @param time total execution time of the new sample.
   */
  @Override
  public void feedback(final int bundleSize, final double time) {
    assert bundleSize > 0;
    if (traceEnabled) {
      log.trace("Bundler#" + bundlerNumber + ": Got another sample with bundleSize=" + bundleSize + " and totalTime=" + time);
    }

    // retrieving the record of the bundle size
    BundlePerformanceSample bundleSample;
    synchronized (samplesMap) {
      bundleSample = samplesMap.get(bundleSize);
      if (bundleSample == null) {
        bundleSample = new BundlePerformanceSample();
        samplesMap.put(bundleSize, bundleSample);
      }
    }

    long samples = bundleSample.samples + bundleSize;
    synchronized (bundleSample) {
      bundleSample.mean = (time + bundleSample.samples * bundleSample.mean) / samples;
      bundleSample.samples = samples;
    }
    if (samples > ((AnnealingTuneProfile) profile).getMinSamplesToAnalyse()) {
      performAnalysis();
      if (traceEnabled) log.trace("Bundler#" + bundlerNumber + ": bundle size = " + bundleSize);
    }
  }

  /**
   * Recompute the bundle size after a performance profile change has been detected.
   */
  private void performAnalysis() {
    double stableMean = 0;
    synchronized (samplesMap) {
      int bestSize = searchBestSize();
      int max = maxSize();
      if ((max > 0) && (bestSize > max)) bestSize = max;
      int counter = 0;
      while (counter < ((AnnealingTuneProfile) profile).getMaxGuessToStable()) {
        int diff = ((AnnealingTuneProfile) profile).createDiff(bestSize, samplesMap.size(), rnd);
        if (diff < bestSize) {
          // the second part is there to ensure the size is > 0
          if (rnd.nextBoolean()) diff = -diff;
        }
        bundleSize = bestSize + diff;
        if (samplesMap.get(bundleSize) == null) {
          if (traceEnabled) log.trace("Bundler#" + bundlerNumber + ": The next bundle size that will be used is " + bundleSize);
          return;
        }
        counter++;
      }

      bundleSize = Math.max(1, bestSize);
      BundlePerformanceSample sample = samplesMap.get(bundleSize);
      if (sample != null) {
        stableMean = sample.mean;
        samplesMap.clear();
        samplesMap.put(bundleSize, sample);
      }
    }
    if (traceEnabled) log.trace("Bundler#" + bundlerNumber + ": The bundle size converged to " + bundleSize + " with the mean execution of " + stableMean);
  }

  /**
   * Lookup the best bundle size in the current samples map.
   * @return the best bundle size as an int value.
   */
  private int searchBestSize() {
    int bestSize = 0;
    double minorMean = Double.POSITIVE_INFINITY;
    for (Integer size : samplesMap.keySet()) {
      BundlePerformanceSample sample = samplesMap.get(size);
      if (sample.mean < minorMean) {
        bestSize = size;
        minorMean = sample.mean;
      }
    }
    if (traceEnabled) log.trace("Bundler#" + bundlerNumber + ": best size found = " + bestSize);
    return bestSize;
  }

  /**
   * Make a copy of this bundler.
   * @return a new AutoTunedBundler instance.
   * @see org.jppf.load.balancer.Bundler#copy()
   */
  @Override
  public Bundler copy() {
    return new AutoTunedBundler(profile.copy());
  }

  @Override
  protected int maxSize() {
    if (jppfContext == null) throw new IllegalStateException("jppfContext not set");
    return jppfContext.getMaxBundleSize() / 2;
  }
}