Spark 3.5: Add DVReaderBenchmark (apache#11537)

spark/v3.5/spark/src/jmh/java/org/apache/iceberg/spark/source/DVReaderBenchmark.java

@@ -0,0 +1,267 @@
+/*
+ * 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.iceberg.spark.source;
+
+import com.google.errorprone.annotations.FormatMethod;
+import com.google.errorprone.annotations.FormatString;
+import java.io.IOException;
+import java.util.Collections;
+import java.util.List;
+import java.util.Random;
+import java.util.Set;
+import java.util.UUID;
+import java.util.concurrent.ThreadLocalRandom;
+import java.util.concurrent.TimeUnit;
+import org.apache.hadoop.conf.Configuration;
+import org.apache.iceberg.DeleteFile;
+import org.apache.iceberg.FileFormat;
+import org.apache.iceberg.FileGenerationUtil;
+import org.apache.iceberg.Table;
+import org.apache.iceberg.data.BaseDeleteLoader;
+import org.apache.iceberg.data.DeleteLoader;
+import org.apache.iceberg.deletes.BaseDVFileWriter;
+import org.apache.iceberg.deletes.DVFileWriter;
+import org.apache.iceberg.deletes.DeleteGranularity;
+import org.apache.iceberg.deletes.PositionDelete;
+import org.apache.iceberg.deletes.PositionDeleteIndex;
+import org.apache.iceberg.io.DeleteWriteResult;
+import org.apache.iceberg.io.FanoutPositionOnlyDeleteWriter;
+import org.apache.iceberg.io.OutputFileFactory;
+import org.apache.iceberg.relocated.com.google.common.collect.ImmutableList;
+import org.apache.iceberg.relocated.com.google.common.collect.Iterables;
+import org.apache.iceberg.relocated.com.google.common.collect.Lists;
+import org.apache.iceberg.relocated.com.google.common.collect.Sets;
+import org.apache.iceberg.spark.Spark3Util;
+import org.apache.iceberg.spark.SparkSessionCatalog;
+import org.apache.iceberg.util.ContentFileUtil;
+import org.apache.spark.sql.SparkSession;
+import org.apache.spark.sql.catalyst.InternalRow;
+import org.apache.spark.sql.catalyst.analysis.NoSuchTableException;
+import org.apache.spark.sql.catalyst.expressions.GenericInternalRow;
+import org.apache.spark.sql.catalyst.parser.ParseException;
+import org.apache.spark.unsafe.types.UTF8String;
+import org.openjdk.jmh.annotations.Benchmark;
+import org.openjdk.jmh.annotations.BenchmarkMode;
+import org.openjdk.jmh.annotations.Fork;
+import org.openjdk.jmh.annotations.Measurement;
+import org.openjdk.jmh.annotations.Mode;
+import org.openjdk.jmh.annotations.Param;
+import org.openjdk.jmh.annotations.Scope;
+import org.openjdk.jmh.annotations.Setup;
+import org.openjdk.jmh.annotations.State;
+import org.openjdk.jmh.annotations.TearDown;
+import org.openjdk.jmh.annotations.Threads;
+import org.openjdk.jmh.annotations.Timeout;
+import org.openjdk.jmh.annotations.Warmup;
+import org.openjdk.jmh.infra.Blackhole;
+
+/**
+ * A benchmark that compares the performance of DV and position delete readers.
+ *
+ * <p>To run this benchmark for spark-3.5: <code>
+ *   ./gradlew -DsparkVersions=3.5 :iceberg-spark:iceberg-spark-3.5_2.12:jmh
+ *       -PjmhIncludeRegex=DVReaderBenchmark
+ *       -PjmhOutputPath=benchmark/iceberg-dv-reader-benchmark-result.txt
+ * </code>
+ */
+@Fork(1)
+@State(Scope.Benchmark)
+@Warmup(iterations = 3)
+@Measurement(iterations = 15)
+@Timeout(time = 20, timeUnit = TimeUnit.MINUTES)
+@BenchmarkMode(Mode.SingleShotTime)
+public class DVReaderBenchmark {
+
+  private static final String TABLE_NAME = "test_table";
+  private static final int DATA_FILE_RECORD_COUNT = 2_000_000;
+  private static final long TARGET_FILE_SIZE = Long.MAX_VALUE;
+
+  @Param({"5", "10"})
+  private int referencedDataFileCount;
+
+  @Param({"0.01", "0.03", "0.05", "0.10", "0.2"})
+  private double deletedRowsRatio;
+
+  private final Configuration hadoopConf = new Configuration();
+  private final Random random = ThreadLocalRandom.current();
+  private SparkSession spark;
+  private Table table;
+  private DeleteWriteResult dvsResult;
+  private DeleteWriteResult fileDeletesResult;
+  private DeleteWriteResult partitionDeletesResult;
+
+  @Setup
+  public void setupBenchmark() throws NoSuchTableException, ParseException, IOException {
+    setupSpark();
+    initTable();
+    List<InternalRow> deletes = generatePositionDeletes();
+    this.dvsResult = writeDVs(deletes);
+    this.fileDeletesResult = writePositionDeletes(deletes, DeleteGranularity.FILE);
+    this.partitionDeletesResult = writePositionDeletes(deletes, DeleteGranularity.PARTITION);
+  }
+
+  @TearDown
+  public void tearDownBenchmark() {
+    dropTable();
+    tearDownSpark();
+  }
+
+  @Benchmark
+  @Threads(1)
+  public void dv(Blackhole blackhole) {
+    DeleteLoader loader = new BaseDeleteLoader(file -> table.io().newInputFile(file), null);
+    DeleteFile dv = dvsResult.deleteFiles().get(0);
+    CharSequence dataFile = dv.referencedDataFile();
+    PositionDeleteIndex index = loader.loadPositionDeletes(ImmutableList.of(dv), dataFile);
+    blackhole.consume(index);
+  }
+
+  @Benchmark
+  @Threads(1)
+  public void fileScopedParquetDeletes(Blackhole blackhole) {
+    DeleteLoader loader = new BaseDeleteLoader(file -> table.io().newInputFile(file), null);
+    DeleteFile deleteFile = fileDeletesResult.deleteFiles().get(0);
+    CharSequence dataFile = ContentFileUtil.referencedDataFile(deleteFile);
+    PositionDeleteIndex index = loader.loadPositionDeletes(ImmutableList.of(deleteFile), dataFile);
+    blackhole.consume(index);
+  }
+
+  @Benchmark
+  @Threads(1)
+  public void partitionScopedParquetDeletes(Blackhole blackhole) {
+    DeleteLoader loader = new BaseDeleteLoader(file -> table.io().newInputFile(file), null);
+    DeleteFile deleteFile = Iterables.getOnlyElement(partitionDeletesResult.deleteFiles());
+    CharSequence dataFile = Iterables.getLast(partitionDeletesResult.referencedDataFiles());
+    PositionDeleteIndex index = loader.loadPositionDeletes(ImmutableList.of(deleteFile), dataFile);
+    blackhole.consume(index);
+  }
+
+  private FanoutPositionOnlyDeleteWriter<InternalRow> newWriter(DeleteGranularity granularity) {
+    return new FanoutPositionOnlyDeleteWriter<>(
+        newWriterFactory(),
+        newFileFactory(FileFormat.PARQUET),
+        table.io(),
+        TARGET_FILE_SIZE,
+        granularity);
+  }
+
+  private SparkFileWriterFactory newWriterFactory() {
+    return SparkFileWriterFactory.builderFor(table).dataFileFormat(FileFormat.PARQUET).build();
+  }
+
+  private OutputFileFactory newFileFactory(FileFormat format) {
+    return OutputFileFactory.builderFor(table, 1, 1).format(format).build();
+  }
+
+  private List<InternalRow> generatePositionDeletes() {
+    int numDeletesPerFile = (int) (DATA_FILE_RECORD_COUNT * deletedRowsRatio);
+    int numDeletes = referencedDataFileCount * numDeletesPerFile;
+    List<InternalRow> deletes = Lists.newArrayListWithExpectedSize(numDeletes);
+
+    for (int pathIndex = 0; pathIndex < referencedDataFileCount; pathIndex++) {
+      UTF8String dataFilePath = UTF8String.fromString(generateDataFilePath());
+      Set<Long> positions = generatePositions(numDeletesPerFile);
+      for (long pos : positions) {
+        deletes.add(new GenericInternalRow(new Object[] {dataFilePath, pos}));
+      }
+    }
+
+    Collections.shuffle(deletes);
+
+    return deletes;
+  }
+
+  private DeleteWriteResult writeDVs(Iterable<InternalRow> rows) throws IOException {
+    OutputFileFactory fileFactory = newFileFactory(FileFormat.PUFFIN);
+    DVFileWriter writer = new BaseDVFileWriter(fileFactory, path -> null);
+    try (DVFileWriter closableWriter = writer) {
+      for (InternalRow row : rows) {
+        String path = row.getString(0);
+        long pos = row.getLong(1);
+        closableWriter.delete(path, pos, table.spec(), null);
+      }
+    }
+    return writer.result();
+  }
+
+  private DeleteWriteResult writePositionDeletes(
+      Iterable<InternalRow> rows, DeleteGranularity granularity) throws IOException {
+    FanoutPositionOnlyDeleteWriter<InternalRow> writer = newWriter(granularity);
+    try (FanoutPositionOnlyDeleteWriter<InternalRow> closableWriter = writer) {
+      PositionDelete<InternalRow> positionDelete = PositionDelete.create();
+      for (InternalRow row : rows) {
+        String path = row.getString(0);
+        long pos = row.getLong(1);
+        positionDelete.set(path, pos, null /* no row */);
+        closableWriter.write(positionDelete, table.spec(), null);
+      }
+    }
+    return writer.result();
+  }
+
+  public Set<Long> generatePositions(int numPositions) {
+    Set<Long> positions = Sets.newHashSet();
+
+    while (positions.size() < numPositions) {
+      long pos = random.nextInt(DATA_FILE_RECORD_COUNT);
+      positions.add(pos);
+    }
+
+    return positions;
+  }
+
+  private String generateDataFilePath() {
+    String fileName = FileGenerationUtil.generateFileName();
+    return table.locationProvider().newDataLocation(table.spec(), null, fileName);
+  }
+
+  private void setupSpark() {
+    this.spark =
+        SparkSession.builder()
+            .config("spark.ui.enabled", false)
+            .config("spark.serializer", "org.apache.spark.serializer.KryoSerializer")
+            .config("spark.sql.catalog.spark_catalog", SparkSessionCatalog.class.getName())
+            .config("spark.sql.catalog.spark_catalog.type", "hadoop")
+            .config("spark.sql.catalog.spark_catalog.warehouse", newWarehouseDir())
+            .master("local[*]")
+            .getOrCreate();
+  }
+
+  private void tearDownSpark() {
+    spark.stop();
+  }
+
+  private void initTable() throws NoSuchTableException, ParseException {
+    sql("CREATE TABLE %s (c1 INT, c2 INT, c3 STRING) USING iceberg", TABLE_NAME);
+    this.table = Spark3Util.loadIcebergTable(spark, TABLE_NAME);
+  }
+
+  private void dropTable() {
+    sql("DROP TABLE IF EXISTS %s PURGE", TABLE_NAME);
+  }
+
+  private String newWarehouseDir() {
+    return hadoopConf.get("hadoop.tmp.dir") + UUID.randomUUID();
+  }
+
+  @FormatMethod
+  private void sql(@FormatString String query, Object... args) {
+    spark.sql(String.format(query, args));
+  }
+}