SPARK-1939 Refactor takeSample method in RDD to use ScaSRS

Modified the takeSample method in RDD to use the ScaSRS sampling technique to improve performance. Added a private method that computes sampling rate > sample_size/total to ensure sufficient sample size with success rate >= 0.9999. Added a unit test for the private method to validate choice of sampling rate.

Author: Doris Xin <[email protected]>
Author: dorx <[email protected]>
Author: Xiangrui Meng <[email protected]>

Closes apache#916 from dorx/takeSample and squashes the following commits:

5b061ae [Doris Xin] merge master
444e750 [Doris Xin] edge cases
3de882b [dorx] Merge pull request #2 from mengxr/SPARK-1939
82dde31 [Xiangrui Meng] update pyspark's takeSample
48d954d [Doris Xin] remove unused imports from RDDSuite
fb1452f [Doris Xin] allowing num to be greater than count in all cases
1481b01 [Doris Xin] washing test tubes and making coffee
dc699f3 [Doris Xin] give back imports removed by accident in rdd.py
64e445b [Doris Xin] logwarnning as soon as it enters the while loop
55518ed [Doris Xin] added TODO for logging in rdd.py
eff89e2 [Doris Xin] addressed reviewer comments.
ecab508 [Doris Xin] "fixed checkstyle violation
0a9b3e3 [Doris Xin] "reviewer comment addressed"
f80f270 [Doris Xin] Merge branch 'master' into takeSample
ae3ad04 [Doris Xin] fixed edge cases to prevent overflow
065ebcd [Doris Xin] Merge branch 'master' into takeSample
9bdd36e [Doris Xin] Check sample size and move computeFraction
e3fd6a6 [Doris Xin] Merge branch 'master' into takeSample
7cab53a [Doris Xin] fixed import bug in rdd.py
ffea61a [Doris Xin] SPARK-1939: Refactor takeSample method in RDD
1441977 [Doris Xin] SPARK-1939 Refactor takeSample method in RDD to use ScaSRS

Author: dorx

core/pom.xml

@@ -67,6 +67,11 @@
       <groupId>org.apache.commons</groupId>
       <artifactId>commons-lang3</artifactId>
     </dependency>
+    <dependency>
+      <groupId>org.apache.commons</groupId>
+      <artifactId>commons-math3</artifactId>
+      <scope>test</scope>
+    </dependency>
     <dependency>
       <groupId>com.google.code.findbugs</groupId>
       <artifactId>jsr305</artifactId>

core/src/main/scala/org/apache/spark/rdd/RDD.scala

@@ -42,7 +42,7 @@ import org.apache.spark.partial.PartialResult
 import org.apache.spark.storage.StorageLevel
 import org.apache.spark.util.{BoundedPriorityQueue, Utils}
 import org.apache.spark.util.collection.OpenHashMap
-import org.apache.spark.util.random.{BernoulliSampler, PoissonSampler}
+import org.apache.spark.util.random.{BernoulliSampler, PoissonSampler, SamplingUtils}
 
 /**
  * A Resilient Distributed Dataset (RDD), the basic abstraction in Spark. Represents an immutable,
@@ -378,46 +378,56 @@ abstract class RDD[T: ClassTag](
     }.toArray
   }
 
-  def takeSample(withReplacement: Boolean, num: Int, seed: Long = Utils.random.nextLong): Array[T] =
-  {
-    var fraction = 0.0
-    var total = 0
-    val multiplier = 3.0
-    val initialCount = this.count()
-    var maxSelected = 0
+  /**
+   * Return a fixed-size sampled subset of this RDD in an array
+   *
+   * @param withReplacement whether sampling is done with replacement
+   * @param num size of the returned sample
+   * @param seed seed for the random number generator
+   * @return sample of specified size in an array
+   */
+  def takeSample(withReplacement: Boolean,
+      num: Int,
+      seed: Long = Utils.random.nextLong): Array[T] = {
+    val numStDev =  10.0
 
     if (num < 0) {
       throw new IllegalArgumentException("Negative number of elements requested")
+    } else if (num == 0) {
+      return new Array[T](0)
     }
 
+    val initialCount = this.count()
     if (initialCount == 0) {
       return new Array[T](0)
     }
 
-    if (initialCount > Integer.MAX_VALUE - 1) {
-      maxSelected = Integer.MAX_VALUE - 1
-    } else {
-      maxSelected = initialCount.toInt
+    val maxSampleSize = Int.MaxValue - (numStDev * math.sqrt(Int.MaxValue)).toInt
+    if (num > maxSampleSize) {
+      throw new IllegalArgumentException("Cannot support a sample size > Int.MaxValue - " +
+        s"$numStDev * math.sqrt(Int.MaxValue)")
     }
 
-    if (num > initialCount && !withReplacement) {
-      total = maxSelected
-      fraction = multiplier * (maxSelected + 1) / initialCount
-    } else {
-      fraction = multiplier * (num + 1) / initialCount
-      total = num
+    val rand = new Random(seed)
+    if (!withReplacement && num >= initialCount) {
+      return Utils.randomizeInPlace(this.collect(), rand)
     }
 
-    val rand = new Random(seed)
+    val fraction = SamplingUtils.computeFractionForSampleSize(num, initialCount,
+      withReplacement)
+
     var samples = this.sample(withReplacement, fraction, rand.nextInt()).collect()
 
     // If the first sample didn't turn out large enough, keep trying to take samples;
     // this shouldn't happen often because we use a big multiplier for the initial size
-    while (samples.length < total) {
+    var numIters = 0
+    while (samples.length < num) {
+      logWarning(s"Needed to re-sample due to insufficient sample size. Repeat #$numIters")
       samples = this.sample(withReplacement, fraction, rand.nextInt()).collect()
+      numIters += 1
     }
 
-    Utils.randomizeInPlace(samples, rand).take(total)
+    Utils.randomizeInPlace(samples, rand).take(num)
   }
 
   /**

core/src/main/scala/org/apache/spark/util/random/RandomSampler.scala

@@ -70,7 +70,7 @@ class BernoulliSampler[T](lb: Double, ub: Double, complement: Boolean = false)
   }
 
   /**
-   *  Return a sampler with is the complement of the range specified of the current sampler.
+   *  Return a sampler that is the complement of the range specified of the current sampler.
    */
   def cloneComplement():  BernoulliSampler[T] = new BernoulliSampler[T](lb, ub, !complement)
 

core/src/main/scala/org/apache/spark/util/random/SamplingUtils.scala

@@ -0,0 +1,55 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one or more
+ * contributor license agreements.  See the NOTICE file distributed with
+ * this work for additional information regarding copyright ownership.
+ * The ASF licenses this file to You 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.apache.spark.util.random
+
+private[spark] object SamplingUtils {
+
+  /**
+   * Returns a sampling rate that guarantees a sample of size >= sampleSizeLowerBound 99.99% of
+   * the time.
+   *
+   * How the sampling rate is determined:
+   * Let p = num / total, where num is the sample size and total is the total number of
+   * datapoints in the RDD. We're trying to compute q > p such that
+   *   - when sampling with replacement, we're drawing each datapoint with prob_i ~ Pois(q),
+   *     where we want to guarantee Pr[s < num] < 0.0001 for s = sum(prob_i for i from 0 to total),
+   *     i.e. the failure rate of not having a sufficiently large sample < 0.0001.
+   *     Setting q = p + 5 * sqrt(p/total) is sufficient to guarantee 0.9999 success rate for
+   *     num > 12, but we need a slightly larger q (9 empirically determined).
+   *   - when sampling without replacement, we're drawing each datapoint with prob_i
+   *     ~ Binomial(total, fraction) and our choice of q guarantees 1-delta, or 0.9999 success
+   *     rate, where success rate is defined the same as in sampling with replacement.
+   *
+   * @param sampleSizeLowerBound sample size
+   * @param total size of RDD
+   * @param withReplacement whether sampling with replacement
+   * @return a sampling rate that guarantees sufficient sample size with 99.99% success rate
+   */
+  def computeFractionForSampleSize(sampleSizeLowerBound: Int, total: Long,
+      withReplacement: Boolean): Double = {
+    val fraction = sampleSizeLowerBound.toDouble / total
+    if (withReplacement) {
+      val numStDev = if (sampleSizeLowerBound < 12) 9 else 5
+      fraction + numStDev * math.sqrt(fraction / total)
+    } else {
+      val delta = 1e-4
+      val gamma = - math.log(delta) / total
+      math.min(1, fraction + gamma + math.sqrt(gamma * gamma + 2 * gamma * fraction))
+    }
+  }
+}

core/src/test/scala/org/apache/spark/rdd/RDDSuite.scala

@@ -505,55 +505,56 @@ class RDDSuite extends FunSuite with SharedSparkContext {
   }
 
   test("takeSample") {
-    val data = sc.parallelize(1 to 100, 2)
+    val n = 1000000
+    val data = sc.parallelize(1 to n, 2)
 
     for (num <- List(5, 20, 100)) {
       val sample = data.takeSample(withReplacement=false, num=num)
       assert(sample.size === num)        // Got exactly num elements
       assert(sample.toSet.size === num)  // Elements are distinct
-      assert(sample.forall(x => 1 <= x && x <= 100), "elements not in [1, 100]")
+      assert(sample.forall(x => 1 <= x && x <= n), s"elements not in [1, $n]")
     }
     for (seed <- 1 to 5) {
       val sample = data.takeSample(withReplacement=false, 20, seed)
       assert(sample.size === 20)        // Got exactly 20 elements
       assert(sample.toSet.size === 20)  // Elements are distinct
-      assert(sample.forall(x => 1 <= x && x <= 100), "elements not in [1, 100]")
+      assert(sample.forall(x => 1 <= x && x <= n), s"elements not in [1, $n]")
     }
     for (seed <- 1 to 5) {
-      val sample = data.takeSample(withReplacement=false, 200, seed)
+      val sample = data.takeSample(withReplacement=false, 100, seed)
       assert(sample.size === 100)        // Got only 100 elements
       assert(sample.toSet.size === 100)  // Elements are distinct
-      assert(sample.forall(x => 1 <= x && x <= 100), "elements not in [1, 100]")
+      assert(sample.forall(x => 1 <= x && x <= n), s"elements not in [1, $n]")
     }
     for (seed <- 1 to 5) {
       val sample = data.takeSample(withReplacement=true, 20, seed)
       assert(sample.size === 20)        // Got exactly 20 elements
-      assert(sample.forall(x => 1 <= x && x <= 100), "elements not in [1, 100]")
+      assert(sample.forall(x => 1 <= x && x <= n), s"elements not in [1, $n]")
     }
     {
       val sample = data.takeSample(withReplacement=true, num=20)
       assert(sample.size === 20)        // Got exactly 100 elements
       assert(sample.toSet.size <= 20, "sampling with replacement returned all distinct elements")
-      assert(sample.forall(x => 1 <= x && x <= 100), "elements not in [1, 100]")
+      assert(sample.forall(x => 1 <= x && x <= n), s"elements not in [1, $n]")
     }
     {
-      val sample = data.takeSample(withReplacement=true, num=100)
-      assert(sample.size === 100)        // Got exactly 100 elements
+      val sample = data.takeSample(withReplacement=true, num=n)
+      assert(sample.size === n)        // Got exactly 100 elements
       // Chance of getting all distinct elements is astronomically low, so test we got < 100
-      assert(sample.toSet.size < 100, "sampling with replacement returned all distinct elements")
-      assert(sample.forall(x => 1 <= x && x <= 100), "elements not in [1, 100]")
+      assert(sample.toSet.size < n, "sampling with replacement returned all distinct elements")
+      assert(sample.forall(x => 1 <= x && x <= n), s"elements not in [1, $n]")
     }
     for (seed <- 1 to 5) {
-      val sample = data.takeSample(withReplacement=true, 100, seed)
-      assert(sample.size === 100)        // Got exactly 100 elements
+      val sample = data.takeSample(withReplacement=true, n, seed)
+      assert(sample.size === n)        // Got exactly 100 elements
       // Chance of getting all distinct elements is astronomically low, so test we got < 100
-      assert(sample.toSet.size < 100, "sampling with replacement returned all distinct elements")
+      assert(sample.toSet.size < n, "sampling with replacement returned all distinct elements")
     }
     for (seed <- 1 to 5) {
-      val sample = data.takeSample(withReplacement=true, 200, seed)
-      assert(sample.size === 200)        // Got exactly 200 elements
+      val sample = data.takeSample(withReplacement=true, 2 * n, seed)
+      assert(sample.size === 2 * n)        // Got exactly 200 elements
       // Chance of getting all distinct elements is still quite low, so test we got < 100
-      assert(sample.toSet.size < 100, "sampling with replacement returned all distinct elements")
+      assert(sample.toSet.size < n, "sampling with replacement returned all distinct elements")
     }
   }
 

core/src/test/scala/org/apache/spark/util/random/SamplingUtilsSuite.scala

@@ -0,0 +1,46 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one or more
+ * contributor license agreements.  See the NOTICE file distributed with
+ * this work for additional information regarding copyright ownership.
+ * The ASF licenses this file to You 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.apache.spark.util.random
+
+import org.apache.commons.math3.distribution.{BinomialDistribution, PoissonDistribution}
+import org.scalatest.FunSuite
+
+class SamplingUtilsSuite extends FunSuite {
+
+  test("computeFraction") {
+    // test that the computed fraction guarantees enough data points
+    // in the sample with a failure rate <= 0.0001
+    val n = 100000
+
+    for (s <- 1 to 15) {
+      val frac = SamplingUtils.computeFractionForSampleSize(s, n, true)
+      val poisson = new PoissonDistribution(frac * n)
+      assert(poisson.inverseCumulativeProbability(0.0001) >= s, "Computed fraction is too low")
+    }
+    for (s <- List(20, 100, 1000)) {
+      val frac = SamplingUtils.computeFractionForSampleSize(s, n, true)
+      val poisson = new PoissonDistribution(frac * n)
+      assert(poisson.inverseCumulativeProbability(0.0001) >= s, "Computed fraction is too low")
+    }
+    for (s <- List(1, 10, 100, 1000)) {
+      val frac = SamplingUtils.computeFractionForSampleSize(s, n, false)
+      val binomial = new BinomialDistribution(n, frac)
+      assert(binomial.inverseCumulativeProbability(0.0001)*n >= s, "Computed fraction is too low")
+    }
+  }
+}

project/SparkBuild.scala

@@ -349,6 +349,7 @@ object SparkBuild extends Build {
     libraryDependencies ++= Seq(
         "com.google.guava"           % "guava"            % "14.0.1",
         "org.apache.commons"         % "commons-lang3"    % "3.3.2",
+        "org.apache.commons"         % "commons-math3"    % "3.3" % "test",
         "com.google.code.findbugs"   % "jsr305"           % "1.3.9",
         "log4j"                      % "log4j"            % "1.2.17",
         "org.slf4j"                  % "slf4j-api"        % slf4jVersion,