dajac commented on a change in pull request #9114: URL: https://github.com/apache/kafka/pull/9114#discussion_r465997810
########## File path: clients/src/main/java/org/apache/kafka/common/metrics/stats/TokenBucket.java ########## @@ -0,0 +1,101 @@ +/* + * 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.common.metrics.stats; + +import java.util.concurrent.TimeUnit; +import org.apache.kafka.common.metrics.MeasurableStat; +import org.apache.kafka.common.metrics.MetricConfig; +import org.apache.kafka.common.metrics.Quota; + +import static org.apache.kafka.common.metrics.internals.MetricsUtils.convert; + +/** + * The {@link TokenBucket} is a {@link MeasurableStat} implementing a token bucket algorithm + * that is usable within a {@link org.apache.kafka.common.metrics.Sensor}. + * + * The {@link Quota#bound()} defined the refill rate of the bucket while the maximum burst or + * the maximum number of credits of the bucket is defined by + * {@link MetricConfig#samples() * MetricConfig#timeWindowMs() * Quota#bound()}. + * + * The quota is considered as exhausted when the amount of remaining credits in the bucket + * is below zero. The enforcement is done by the {@link org.apache.kafka.common.metrics.Sensor}. + * + * Token Bucket vs Rate based Quota: + * The current sampled rate based quota does not cope well with bursty workloads. The issue is + * that a unique and large sample can hold the average above the quota and this until it is + * discarded. Practically, when this happens, one must wait until the sample is expired to + * bring the rate below the quota even though less time would be theoretically required. As an + * examples, let's imagine that we have: + * - Quota (Q) = 5 + * - Samples (S) = 100 + * - Window (W) = 1s + * A burst of 560 brings the average rate (R) to 5.6 (560 / 100). The throttle time is computed as Review comment: Yes, that is correct. I have added "expected" to be clearer. ---------------------------------------------------------------- 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. For queries about this service, please contact Infrastructure at: us...@infra.apache.org
