dajac commented on code in PR #15717: URL: https://github.com/apache/kafka/pull/15717#discussion_r1572314707
########## jmh-benchmarks/src/main/java/org/apache/kafka/jmh/assignor/ServerSideAssignorBenchmark.java: ########## @@ -0,0 +1,266 @@ +/* + * 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.kafka.jmh.assignor; + +import org.apache.kafka.common.Uuid; +import org.apache.kafka.coordinator.group.assignor.AssignmentMemberSpec; +import org.apache.kafka.coordinator.group.assignor.AssignmentSpec; +import org.apache.kafka.coordinator.group.assignor.GroupAssignment; +import org.apache.kafka.coordinator.group.assignor.MemberAssignment; +import org.apache.kafka.coordinator.group.assignor.PartitionAssignor; +import org.apache.kafka.coordinator.group.assignor.RangeAssignor; +import org.apache.kafka.coordinator.group.assignor.SubscribedTopicDescriber; +import org.apache.kafka.coordinator.group.assignor.UniformAssignor; +import org.apache.kafka.coordinator.group.consumer.SubscribedTopicMetadata; +import org.apache.kafka.coordinator.group.consumer.TopicMetadata; +import org.openjdk.jmh.annotations.Benchmark; +import org.openjdk.jmh.annotations.BenchmarkMode; +import org.openjdk.jmh.annotations.Fork; +import org.openjdk.jmh.annotations.Level; +import org.openjdk.jmh.annotations.Measurement; +import org.openjdk.jmh.annotations.Mode; +import org.openjdk.jmh.annotations.OutputTimeUnit; +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.Threads; +import org.openjdk.jmh.annotations.Warmup; + +import java.util.ArrayList; +import java.util.Arrays; +import java.util.Collections; +import java.util.HashMap; +import java.util.HashSet; +import java.util.List; +import java.util.Map; +import java.util.Optional; +import java.util.Set; +import java.util.concurrent.TimeUnit; + +@State(Scope.Benchmark) +@Fork(value = 1) +@Warmup(iterations = 5) +@Measurement(iterations = 5) +@BenchmarkMode(Mode.AverageTime) +@OutputTimeUnit(TimeUnit.MILLISECONDS) +public class ServerSideAssignorBenchmark { + + public enum AssignorType { + RANGE(new RangeAssignor()), + UNIFORM(new UniformAssignor()); + + private final PartitionAssignor assignor; + + AssignorType(PartitionAssignor assignor) { + this.assignor = assignor; + } + + public PartitionAssignor assignor() { + return assignor; + } + } + + /** + * The subscription pattern followed by the members of the group. + * + * A subscription model is considered homogenous if all the members of the group + * are subscribed to the same set of topics, it is heterogeneous otherwise. + */ + public enum SubscriptionModel { + HOMOGENEOUS, HETEROGENEOUS + } + + /** + * The assignment type is decided based on whether all the members are assigned partitions + * for the first time (full), or incrementally when a rebalance is triggered. + */ + public enum AssignmentType { + FULL, INCREMENTAL + } + + @Param({"100", "500", "1000", "5000", "10000"}) + private int memberCount; + + @Param({"5", "10", "50"}) + private int partitionsToMemberRatio; + + @Param({"10", "100", "1000"}) + private int topicCount; + + @Param({"true", "false"}) + private boolean isRackAware; + + @Param({"HOMOGENEOUS", "HETEROGENEOUS"}) + private SubscriptionModel subscriptionModel; + + @Param({"RANGE", "UNIFORM"}) + private AssignorType assignorType; + + @Param({"FULL", "INCREMENTAL"}) + private AssignmentType assignmentType; + + private PartitionAssignor partitionAssignor; + + private static final int NUMBER_OF_RACKS = 3; + + private AssignmentSpec assignmentSpec; + + private SubscribedTopicDescriber subscribedTopicDescriber; + + private final List<Uuid> allTopicIds = new ArrayList<>(topicCount); + + @Setup(Level.Trial) + public void setup() { + Map<Uuid, TopicMetadata> topicMetadata = createTopicMetadata(); + subscribedTopicDescriber = new SubscribedTopicMetadata(topicMetadata); + + createAssignmentSpec(); + + partitionAssignor = assignorType.assignor(); + + if (assignmentType == AssignmentType.INCREMENTAL) { + simulateIncrementalRebalance(topicMetadata); + } + } + + private Map<Uuid, TopicMetadata> createTopicMetadata() { + Map<Uuid, TopicMetadata> topicMetadata = new HashMap<>(); + int partitionsPerTopicCount = (memberCount * partitionsToMemberRatio) / topicCount; + + Map<Integer, Set<String>> partitionRacks = isRackAware ? + mkMapOfPartitionRacks(partitionsPerTopicCount) : + Collections.emptyMap(); + + for (int i = 0; i < topicCount; i++) { + Uuid topicUuid = Uuid.randomUuid(); + String topicName = "topic" + i; + allTopicIds.add(topicUuid); + topicMetadata.put(topicUuid, new TopicMetadata( + topicUuid, + topicName, + partitionsPerTopicCount, + partitionRacks + )); + } + + return topicMetadata; + } + + private void createAssignmentSpec() { + Map<String, AssignmentMemberSpec> members = new HashMap<>(); + + // In the rebalance case, we will add the last member as a trigger. + // This is done to keep the total members count consistent with the input. + int numberOfMembers = assignmentType.equals(AssignmentType.INCREMENTAL) ? memberCount - 1 : memberCount; + + if (subscriptionModel.equals(SubscriptionModel.HOMOGENEOUS)) { + for (int i = 0; i < numberOfMembers; i++) { + addMemberSpec(members, i, new HashSet<>(allTopicIds)); + } + } else { + // Check minimum topics requirement + if (topicCount < 5) { + throw new IllegalArgumentException("At least 5 topics are recommended for effective bucketing."); + } + + // Adjust bucket count based on member count when member count < 5 + int bucketCount = Math.min(5, numberOfMembers); + int bucketSizeTopics = (int) Math.ceil((double) topicCount / bucketCount); + int bucketSizeMembers = (int) Math.ceil((double) numberOfMembers / bucketCount); + + // Define buckets for each member and assign topics from the same bucket + for (int bucket = 0; bucket < bucketCount; bucket++) { + int memberStartIndex = bucket * bucketSizeMembers; + int memberEndIndex = Math.min((bucket + 1) * bucketSizeMembers, numberOfMembers); + + int topicStartIndex = bucket * bucketSizeTopics; + int topicEndIndex = Math.min((bucket + 1) * bucketSizeTopics, topicCount); + + Set<Uuid> bucketTopics = new HashSet<>(allTopicIds.subList(topicStartIndex, topicEndIndex)); + + // Assign topics to each member in the current bucket + for (int i = memberStartIndex; i < memberEndIndex; i++) { + addMemberSpec(members, i, new HashSet<>(bucketTopics)); + } + } + } + + this.assignmentSpec = new AssignmentSpec(members); + } + + private Optional<String> rackId(int index) { + return isRackAware ? Optional.of("rack" + index % NUMBER_OF_RACKS) : Optional.empty(); + } + + private void addMemberSpec( + Map<String, AssignmentMemberSpec> members, + int memberIndex, + Set<Uuid> subscribedTopicIds + ) { + String memberId = "member" + memberIndex; + Optional<String> rackId = rackId(memberIndex); + + members.put(memberId, new AssignmentMemberSpec( + Optional.empty(), + rackId, + subscribedTopicIds, + Collections.emptyMap() + )); + } + + private static Map<Integer, Set<String>> mkMapOfPartitionRacks(int numPartitions) { + Map<Integer, Set<String>> partitionRacks = new HashMap<>(numPartitions); + for (int i = 0; i < numPartitions; i++) { + partitionRacks.put(i, new HashSet<>(Arrays.asList("rack" + i % NUMBER_OF_RACKS, "rack" + (i + 1) % NUMBER_OF_RACKS))); + } + return partitionRacks; + } + + private void simulateIncrementalRebalance(Map<Uuid, TopicMetadata> topicMetadata) { + GroupAssignment initialAssignment = partitionAssignor.assign(assignmentSpec, subscribedTopicDescriber); + Map<String, MemberAssignment> members = initialAssignment.members(); + + Map<String, AssignmentMemberSpec> updatedMembers = new HashMap<>(); + members.forEach((memberId, memberAssignment) -> { + AssignmentMemberSpec memberSpec = assignmentSpec.members().get(memberId); + updatedMembers.put(memberId, new AssignmentMemberSpec( + memberSpec.instanceId(), + memberSpec.rackId(), + memberSpec.subscribedTopicIds(), + memberAssignment.targetPartitions() + )); + }); + + Optional<String> rackId = rackId(memberCount - 1); + updatedMembers.put("newMember", new AssignmentMemberSpec( + Optional.empty(), + rackId, + topicMetadata.keySet(), Review Comment: I don't fully understand this one. Intuitively, I thought that we would use the last bucket to respect the topology of the group. Is there a reason not to do it? One way would be to reuse the subscribed topics of the last member in `members`. Would it work? ########## jmh-benchmarks/src/main/java/org/apache/kafka/jmh/assignor/TargetAssignmentBuilderBenchmark.java: ########## @@ -0,0 +1,202 @@ +/* + * 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.kafka.jmh.assignor; + +import org.apache.kafka.common.Uuid; +import org.apache.kafka.coordinator.group.assignor.AssignmentMemberSpec; +import org.apache.kafka.coordinator.group.assignor.AssignmentSpec; +import org.apache.kafka.coordinator.group.assignor.GroupAssignment; +import org.apache.kafka.coordinator.group.assignor.MemberAssignment; +import org.apache.kafka.coordinator.group.assignor.PartitionAssignor; +import org.apache.kafka.coordinator.group.assignor.UniformAssignor; +import org.apache.kafka.coordinator.group.consumer.Assignment; +import org.apache.kafka.coordinator.group.consumer.ConsumerGroupMember; +import org.apache.kafka.coordinator.group.consumer.SubscribedTopicMetadata; +import org.apache.kafka.coordinator.group.consumer.TargetAssignmentBuilder; +import org.apache.kafka.coordinator.group.consumer.TopicMetadata; +import org.apache.kafka.coordinator.group.consumer.VersionedMetadata; +import org.openjdk.jmh.annotations.Benchmark; +import org.openjdk.jmh.annotations.BenchmarkMode; +import org.openjdk.jmh.annotations.Fork; +import org.openjdk.jmh.annotations.Level; +import org.openjdk.jmh.annotations.Measurement; +import org.openjdk.jmh.annotations.Mode; +import org.openjdk.jmh.annotations.OutputTimeUnit; +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.Threads; +import org.openjdk.jmh.annotations.Warmup; + +import java.util.ArrayList; +import java.util.Collections; +import java.util.HashMap; +import java.util.HashSet; +import java.util.List; +import java.util.Map; +import java.util.Optional; +import java.util.Set; +import java.util.concurrent.TimeUnit; + +@State(Scope.Benchmark) +@Fork(value = 1) +@Warmup(iterations = 5) +@Measurement(iterations = 5) +@BenchmarkMode(Mode.AverageTime) +@OutputTimeUnit(TimeUnit.MILLISECONDS) +public class TargetAssignmentBuilderBenchmark { + + @Param({"100", "500", "1000", "5000", "10000"}) + private int memberCount; + + @Param({"5", "10", "50"}) + private int partitionsToMemberRatio; + + @Param({"10", "100", "1000"}) + private int topicCount; + + private static final String GROUP_ID = "benchmark-group"; + + private static final int GROUP_EPOCH = 0; + + private PartitionAssignor partitionAssignor; + + private Map<String, TopicMetadata> subscriptionMetadata = Collections.emptyMap(); + + private TargetAssignmentBuilder targetAssignmentBuilder; + + private AssignmentSpec assignmentSpec; + + private final List<String> allTopicNames = new ArrayList<>(topicCount); + + private final List<Uuid> allTopicIds = new ArrayList<>(topicCount); + + @Setup(Level.Trial) + public void setup() { + // For this benchmark we will use the Uniform Assignor + // and a group that has a homogeneous subscription model. + partitionAssignor = new UniformAssignor(); + + subscriptionMetadata = generateMockSubscriptionMetadata(); + Map<String, ConsumerGroupMember> members = generateMockMembers(); + Map<String, Assignment> existingTargetAssignment = generateMockInitialTargetAssignment(); + + // Add a new member to trigger a rebalance. + Set<String> subscribedTopics = new HashSet<>(subscriptionMetadata.keySet()); + + ConsumerGroupMember newMember = new ConsumerGroupMember.Builder("new-member") + .setSubscribedTopicNames(new ArrayList<>(subscribedTopics)) Review Comment: Could we use `allTopicNames` instead of creating a set to create an list? ########## jmh-benchmarks/src/main/java/org/apache/kafka/jmh/assignor/ServerSideAssignorBenchmark.java: ########## @@ -0,0 +1,266 @@ +/* + * 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.kafka.jmh.assignor; + +import org.apache.kafka.common.Uuid; +import org.apache.kafka.coordinator.group.assignor.AssignmentMemberSpec; +import org.apache.kafka.coordinator.group.assignor.AssignmentSpec; +import org.apache.kafka.coordinator.group.assignor.GroupAssignment; +import org.apache.kafka.coordinator.group.assignor.MemberAssignment; +import org.apache.kafka.coordinator.group.assignor.PartitionAssignor; +import org.apache.kafka.coordinator.group.assignor.RangeAssignor; +import org.apache.kafka.coordinator.group.assignor.SubscribedTopicDescriber; +import org.apache.kafka.coordinator.group.assignor.UniformAssignor; +import org.apache.kafka.coordinator.group.consumer.SubscribedTopicMetadata; +import org.apache.kafka.coordinator.group.consumer.TopicMetadata; +import org.openjdk.jmh.annotations.Benchmark; +import org.openjdk.jmh.annotations.BenchmarkMode; +import org.openjdk.jmh.annotations.Fork; +import org.openjdk.jmh.annotations.Level; +import org.openjdk.jmh.annotations.Measurement; +import org.openjdk.jmh.annotations.Mode; +import org.openjdk.jmh.annotations.OutputTimeUnit; +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.Threads; +import org.openjdk.jmh.annotations.Warmup; + +import java.util.ArrayList; +import java.util.Arrays; +import java.util.Collections; +import java.util.HashMap; +import java.util.HashSet; +import java.util.List; +import java.util.Map; +import java.util.Optional; +import java.util.Set; +import java.util.concurrent.TimeUnit; + +@State(Scope.Benchmark) +@Fork(value = 1) +@Warmup(iterations = 5) +@Measurement(iterations = 5) +@BenchmarkMode(Mode.AverageTime) +@OutputTimeUnit(TimeUnit.MILLISECONDS) +public class ServerSideAssignorBenchmark { + + public enum AssignorType { + RANGE(new RangeAssignor()), + UNIFORM(new UniformAssignor()); + + private final PartitionAssignor assignor; + + AssignorType(PartitionAssignor assignor) { + this.assignor = assignor; + } + + public PartitionAssignor assignor() { + return assignor; + } + } + + /** + * The subscription pattern followed by the members of the group. + * + * A subscription model is considered homogenous if all the members of the group + * are subscribed to the same set of topics, it is heterogeneous otherwise. + */ + public enum SubscriptionModel { + HOMOGENEOUS, HETEROGENEOUS + } + + /** + * The assignment type is decided based on whether all the members are assigned partitions + * for the first time (full), or incrementally when a rebalance is triggered. + */ + public enum AssignmentType { + FULL, INCREMENTAL + } + + @Param({"100", "500", "1000", "5000", "10000"}) + private int memberCount; + + @Param({"5", "10", "50"}) + private int partitionsToMemberRatio; + + @Param({"10", "100", "1000"}) + private int topicCount; + + @Param({"true", "false"}) + private boolean isRackAware; + + @Param({"HOMOGENEOUS", "HETEROGENEOUS"}) + private SubscriptionModel subscriptionModel; + + @Param({"RANGE", "UNIFORM"}) + private AssignorType assignorType; + + @Param({"FULL", "INCREMENTAL"}) + private AssignmentType assignmentType; + + private PartitionAssignor partitionAssignor; + + private static final int NUMBER_OF_RACKS = 3; + + private AssignmentSpec assignmentSpec; + + private SubscribedTopicDescriber subscribedTopicDescriber; + + private final List<Uuid> allTopicIds = new ArrayList<>(topicCount); + + @Setup(Level.Trial) + public void setup() { + Map<Uuid, TopicMetadata> topicMetadata = createTopicMetadata(); + subscribedTopicDescriber = new SubscribedTopicMetadata(topicMetadata); + + createAssignmentSpec(); + + partitionAssignor = assignorType.assignor(); + + if (assignmentType == AssignmentType.INCREMENTAL) { + simulateIncrementalRebalance(topicMetadata); + } + } + + private Map<Uuid, TopicMetadata> createTopicMetadata() { + Map<Uuid, TopicMetadata> topicMetadata = new HashMap<>(); + int partitionsPerTopicCount = (memberCount * partitionsToMemberRatio) / topicCount; + + Map<Integer, Set<String>> partitionRacks = isRackAware ? + mkMapOfPartitionRacks(partitionsPerTopicCount) : + Collections.emptyMap(); + + for (int i = 0; i < topicCount; i++) { + Uuid topicUuid = Uuid.randomUuid(); + String topicName = "topic" + i; + allTopicIds.add(topicUuid); + topicMetadata.put(topicUuid, new TopicMetadata( + topicUuid, + topicName, + partitionsPerTopicCount, + partitionRacks + )); + } + + return topicMetadata; + } + + private void createAssignmentSpec() { + Map<String, AssignmentMemberSpec> members = new HashMap<>(); + + // In the rebalance case, we will add the last member as a trigger. + // This is done to keep the total members count consistent with the input. + int numberOfMembers = assignmentType.equals(AssignmentType.INCREMENTAL) ? memberCount - 1 : memberCount; + + if (subscriptionModel.equals(SubscriptionModel.HOMOGENEOUS)) { + for (int i = 0; i < numberOfMembers; i++) { + addMemberSpec(members, i, new HashSet<>(allTopicIds)); + } + } else { + // Check minimum topics requirement + if (topicCount < 5) { + throw new IllegalArgumentException("At least 5 topics are recommended for effective bucketing."); + } + + // Adjust bucket count based on member count when member count < 5 + int bucketCount = Math.min(5, numberOfMembers); Review Comment: Should we define a const for `5`? ########## jmh-benchmarks/src/main/java/org/apache/kafka/jmh/assignor/TargetAssignmentBuilderBenchmark.java: ########## @@ -0,0 +1,202 @@ +/* + * 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.kafka.jmh.assignor; + +import org.apache.kafka.common.Uuid; +import org.apache.kafka.coordinator.group.assignor.AssignmentMemberSpec; +import org.apache.kafka.coordinator.group.assignor.AssignmentSpec; +import org.apache.kafka.coordinator.group.assignor.GroupAssignment; +import org.apache.kafka.coordinator.group.assignor.MemberAssignment; +import org.apache.kafka.coordinator.group.assignor.PartitionAssignor; +import org.apache.kafka.coordinator.group.assignor.UniformAssignor; +import org.apache.kafka.coordinator.group.consumer.Assignment; +import org.apache.kafka.coordinator.group.consumer.ConsumerGroupMember; +import org.apache.kafka.coordinator.group.consumer.SubscribedTopicMetadata; +import org.apache.kafka.coordinator.group.consumer.TargetAssignmentBuilder; +import org.apache.kafka.coordinator.group.consumer.TopicMetadata; +import org.apache.kafka.coordinator.group.consumer.VersionedMetadata; +import org.openjdk.jmh.annotations.Benchmark; +import org.openjdk.jmh.annotations.BenchmarkMode; +import org.openjdk.jmh.annotations.Fork; +import org.openjdk.jmh.annotations.Level; +import org.openjdk.jmh.annotations.Measurement; +import org.openjdk.jmh.annotations.Mode; +import org.openjdk.jmh.annotations.OutputTimeUnit; +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.Threads; +import org.openjdk.jmh.annotations.Warmup; + +import java.util.ArrayList; +import java.util.Collections; +import java.util.HashMap; +import java.util.HashSet; +import java.util.List; +import java.util.Map; +import java.util.Optional; +import java.util.Set; +import java.util.concurrent.TimeUnit; + +@State(Scope.Benchmark) +@Fork(value = 1) +@Warmup(iterations = 5) +@Measurement(iterations = 5) +@BenchmarkMode(Mode.AverageTime) +@OutputTimeUnit(TimeUnit.MILLISECONDS) +public class TargetAssignmentBuilderBenchmark { + + @Param({"100", "500", "1000", "5000", "10000"}) + private int memberCount; + + @Param({"5", "10", "50"}) + private int partitionsToMemberRatio; + + @Param({"10", "100", "1000"}) + private int topicCount; + + private static final String GROUP_ID = "benchmark-group"; + + private static final int GROUP_EPOCH = 0; + + private PartitionAssignor partitionAssignor; + + private Map<String, TopicMetadata> subscriptionMetadata = Collections.emptyMap(); + + private TargetAssignmentBuilder targetAssignmentBuilder; + + private AssignmentSpec assignmentSpec; + + private final List<String> allTopicNames = new ArrayList<>(topicCount); + + private final List<Uuid> allTopicIds = new ArrayList<>(topicCount); + + @Setup(Level.Trial) + public void setup() { + // For this benchmark we will use the Uniform Assignor + // and a group that has a homogeneous subscription model. + partitionAssignor = new UniformAssignor(); + + subscriptionMetadata = generateMockSubscriptionMetadata(); + Map<String, ConsumerGroupMember> members = generateMockMembers(); + Map<String, Assignment> existingTargetAssignment = generateMockInitialTargetAssignment(); + + // Add a new member to trigger a rebalance. + Set<String> subscribedTopics = new HashSet<>(subscriptionMetadata.keySet()); + + ConsumerGroupMember newMember = new ConsumerGroupMember.Builder("new-member") + .setSubscribedTopicNames(new ArrayList<>(subscribedTopics)) + .build(); + + targetAssignmentBuilder = new TargetAssignmentBuilder(GROUP_ID, GROUP_EPOCH, partitionAssignor) + .withMembers(members) + .withSubscriptionMetadata(subscriptionMetadata) + .withTargetAssignment(existingTargetAssignment) + .addOrUpdateMember(newMember.memberId(), newMember); + } + + private Map<String, ConsumerGroupMember> generateMockMembers() { + Map<String, ConsumerGroupMember> members = new HashMap<>(); + + for (int i = 0; i < memberCount - 1; i++) { + Set<String> subscribedTopics; + subscribedTopics = new HashSet<>(allTopicNames); + + ConsumerGroupMember member = new ConsumerGroupMember.Builder("member" + i) + .setSubscribedTopicNames(new ArrayList<>(subscribedTopics)) Review Comment: Could we also use `allTopicNames` directly here? ########## jmh-benchmarks/src/main/java/org/apache/kafka/jmh/assignor/TargetAssignmentBuilderBenchmark.java: ########## @@ -0,0 +1,202 @@ +/* + * 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.kafka.jmh.assignor; + +import org.apache.kafka.common.Uuid; +import org.apache.kafka.coordinator.group.assignor.AssignmentMemberSpec; +import org.apache.kafka.coordinator.group.assignor.AssignmentSpec; +import org.apache.kafka.coordinator.group.assignor.GroupAssignment; +import org.apache.kafka.coordinator.group.assignor.MemberAssignment; +import org.apache.kafka.coordinator.group.assignor.PartitionAssignor; +import org.apache.kafka.coordinator.group.assignor.UniformAssignor; +import org.apache.kafka.coordinator.group.consumer.Assignment; +import org.apache.kafka.coordinator.group.consumer.ConsumerGroupMember; +import org.apache.kafka.coordinator.group.consumer.SubscribedTopicMetadata; +import org.apache.kafka.coordinator.group.consumer.TargetAssignmentBuilder; +import org.apache.kafka.coordinator.group.consumer.TopicMetadata; +import org.apache.kafka.coordinator.group.consumer.VersionedMetadata; +import org.openjdk.jmh.annotations.Benchmark; +import org.openjdk.jmh.annotations.BenchmarkMode; +import org.openjdk.jmh.annotations.Fork; +import org.openjdk.jmh.annotations.Level; +import org.openjdk.jmh.annotations.Measurement; +import org.openjdk.jmh.annotations.Mode; +import org.openjdk.jmh.annotations.OutputTimeUnit; +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.Threads; +import org.openjdk.jmh.annotations.Warmup; + +import java.util.ArrayList; +import java.util.Collections; +import java.util.HashMap; +import java.util.HashSet; +import java.util.List; +import java.util.Map; +import java.util.Optional; +import java.util.Set; +import java.util.concurrent.TimeUnit; + +@State(Scope.Benchmark) +@Fork(value = 1) +@Warmup(iterations = 5) +@Measurement(iterations = 5) +@BenchmarkMode(Mode.AverageTime) +@OutputTimeUnit(TimeUnit.MILLISECONDS) +public class TargetAssignmentBuilderBenchmark { + + @Param({"100", "500", "1000", "5000", "10000"}) + private int memberCount; + + @Param({"5", "10", "50"}) + private int partitionsToMemberRatio; + + @Param({"10", "100", "1000"}) + private int topicCount; + + private static final String GROUP_ID = "benchmark-group"; + + private static final int GROUP_EPOCH = 0; + + private PartitionAssignor partitionAssignor; + + private Map<String, TopicMetadata> subscriptionMetadata = Collections.emptyMap(); + + private TargetAssignmentBuilder targetAssignmentBuilder; + + private AssignmentSpec assignmentSpec; + + private final List<String> allTopicNames = new ArrayList<>(topicCount); + + private final List<Uuid> allTopicIds = new ArrayList<>(topicCount); + + @Setup(Level.Trial) + public void setup() { + // For this benchmark we will use the Uniform Assignor + // and a group that has a homogeneous subscription model. + partitionAssignor = new UniformAssignor(); + + subscriptionMetadata = generateMockSubscriptionMetadata(); + Map<String, ConsumerGroupMember> members = generateMockMembers(); + Map<String, Assignment> existingTargetAssignment = generateMockInitialTargetAssignment(); + + // Add a new member to trigger a rebalance. + Set<String> subscribedTopics = new HashSet<>(subscriptionMetadata.keySet()); + + ConsumerGroupMember newMember = new ConsumerGroupMember.Builder("new-member") + .setSubscribedTopicNames(new ArrayList<>(subscribedTopics)) + .build(); + + targetAssignmentBuilder = new TargetAssignmentBuilder(GROUP_ID, GROUP_EPOCH, partitionAssignor) + .withMembers(members) + .withSubscriptionMetadata(subscriptionMetadata) + .withTargetAssignment(existingTargetAssignment) + .addOrUpdateMember(newMember.memberId(), newMember); + } + + private Map<String, ConsumerGroupMember> generateMockMembers() { + Map<String, ConsumerGroupMember> members = new HashMap<>(); + + for (int i = 0; i < memberCount - 1; i++) { + Set<String> subscribedTopics; + subscribedTopics = new HashSet<>(allTopicNames); Review Comment: nit: You can combine these two into one line. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: jira-unsubscr...@kafka.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
