diff options
Diffstat (limited to 'infra')
4 files changed, 181 insertions, 17 deletions
diff --git a/infra/it/benchmark/pom.xml b/infra/it/benchmark/pom.xml index 54ed6e1b8..8f9eabc56 100644 --- a/infra/it/benchmark/pom.xml +++ b/infra/it/benchmark/pom.xml @@ -59,6 +59,12 @@ <artifactId>honeycomb-test-model</artifactId> <version>${project.version}</version> </dependency> + <dependency> + <groupId>org.apache.commons</groupId> + <artifactId>commons-csv</artifactId> + <version>1.4</version> + </dependency> + </dependencies> <build> @@ -129,6 +135,19 @@ </arguments> </configuration> </execution> + <execution> + <id>benchmark-post-processing</id> + <phase>test</phase> + <goals> + <goal>java</goal> + </goals> + <configuration> + <mainClass>io.fd.honeycomb.benchmark.format.BenchmarkOutputFormatter</mainClass> + <arguments> + <argument>${project.build.directory}/benchmark.csv</argument> + </arguments> + </configuration> + </execution> </executions> </plugin> </plugins> diff --git a/infra/it/benchmark/src/main/java/io/fd/honeycomb/benchmark/data/DataBrokerConfigWriteBenchmark.java b/infra/it/benchmark/src/main/java/io/fd/honeycomb/benchmark/data/DataBrokerConfigWriteBenchmark.java index 145cd2bcb..6d76ad2d3 100644 --- a/infra/it/benchmark/src/main/java/io/fd/honeycomb/benchmark/data/DataBrokerConfigWriteBenchmark.java +++ b/infra/it/benchmark/src/main/java/io/fd/honeycomb/benchmark/data/DataBrokerConfigWriteBenchmark.java @@ -92,17 +92,6 @@ public class DataBrokerConfigWriteBenchmark extends AbstractModule implements Fi private String data; private DataProvider dataProvider; - /* - * TODO HONEYCOMB-288 Visualization notes: - * - visualize as 3 graphs, 1 for each data - * - each graph should show 4 lines. for the combinations of parameters: submitFrequency and persistence - * (if that's too much split or reduce submitFrequecy values that are shown in graph) - * - * TODO data need to be prepared for such visualization. Maybe if each benchmark class exposed a method to prepare - * that data from aggregated results... it might be easy - * (just maven exec plugin + main that invokes some method on all benchmark classes) - */ - // Infra modules to load private final Module[] modules = new Module[] { new io.fd.honeycomb.infra.distro.schema.YangBindingProviderModule(), diff --git a/infra/it/benchmark/src/main/java/io/fd/honeycomb/benchmark/data/DataBrokerOperReadBenchmark.java b/infra/it/benchmark/src/main/java/io/fd/honeycomb/benchmark/data/DataBrokerOperReadBenchmark.java index 2fd9fa239..86ded0f66 100644 --- a/infra/it/benchmark/src/main/java/io/fd/honeycomb/benchmark/data/DataBrokerOperReadBenchmark.java +++ b/infra/it/benchmark/src/main/java/io/fd/honeycomb/benchmark/data/DataBrokerOperReadBenchmark.java @@ -85,12 +85,6 @@ public class DataBrokerOperReadBenchmark extends AbstractModule implements FileM private String data; private DataProvider dataProvider; - /* - * TODO HONEYCOMB-288 Visualization notes: - * - visualize as 1 graph - * - just 3 lines - */ - // Infra modules to load private final Module[] modules = new Module[] { new io.fd.honeycomb.infra.distro.schema.YangBindingProviderModule(), diff --git a/infra/it/benchmark/src/main/java/io/fd/honeycomb/benchmark/format/BenchmarkOutputFormatter.java b/infra/it/benchmark/src/main/java/io/fd/honeycomb/benchmark/format/BenchmarkOutputFormatter.java new file mode 100644 index 000000000..dc6552d53 --- /dev/null +++ b/infra/it/benchmark/src/main/java/io/fd/honeycomb/benchmark/format/BenchmarkOutputFormatter.java @@ -0,0 +1,162 @@ +/* + * Copyright (c) 2016 Cisco and/or its affiliates. + * + * 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 io.fd.honeycomb.benchmark.format; + +import com.google.common.base.Charsets; +import io.fd.honeycomb.benchmark.util.DataProvider; +import java.io.File; +import java.io.FileInputStream; +import java.io.FileNotFoundException; +import java.io.FileOutputStream; +import java.io.IOException; +import java.io.InputStreamReader; +import java.io.Reader; +import java.util.ArrayList; +import java.util.List; +import java.util.stream.Collectors; +import org.apache.commons.csv.CSVFormat; +import org.apache.commons.csv.CSVPrinter; +import org.apache.commons.csv.CSVRecord; +import org.opendaylight.controller.md.sal.common.api.data.LogicalDatastoreType; +import org.slf4j.Logger; +import org.slf4j.LoggerFactory; + +/** + * Processes JMH CSV file to format accepted by Jenkins plot plugin (header + one data line per CSV file). + */ +public final class BenchmarkOutputFormatter { + /* + Input format of the JMH CSV file: + + "Benchmark","Mode","Threads","Samples","Score","Score Error (99,9%)","Unit","Param: data","Param: dsType","Param: operation","Param: persistence","Param: submitFrequency" + "io.fd.honeycomb.benchmark.data.DataBrokerConfigWriteBenchmark.write","thrpt",1,1,"3099,563546",NaN,"ops/s",simple-container,CONFIGURATION,put,true,1 + ... + + */ + private static final int SCORE_POSITION = 4; + private static final int DATA_TYPE_POSITION = 7; + private static final int DS_TYPE_POSITION = 8; + private static final int PERSISTENCE_POSITION = 10; + private static final int SUBMIT_FREQUENCY_POSITION = 11; + + private static final Logger LOG = LoggerFactory.getLogger(BenchmarkOutputFormatter.class); + + private BenchmarkOutputFormatter() { + } + + /** + * Produces 4 CSV files (simple-container, list-in-container, complex-list-in-container write + one plot for read). + */ + public static void main(final String[] args) throws Exception { + final File csvFile = new File(args[0]); + if (!csvFile.exists()) { + throw new FileNotFoundException(args[0]); + } + final String path = csvFile.getParent(); + LOG.info("Preparing benchmarking plot data from: {}", args[0]); + + final Reader in = new InputStreamReader(new FileInputStream(csvFile), Charsets.UTF_8); + final List<CSVRecord> records = CSVFormat.RFC4180.parse(in).getRecords(); + writeStatistics(processSimpleContainer(records), path, "simple-container.csv"); + writeStatistics(processListContainer(records), path, "list-in-container.csv"); + writeStatistics(processComplexListContainer(records), path, "complex-list-in-container.csv"); + writeStatistics(processReadStatistics(records), path, "operational-read.csv"); + LOG.info("Finished benchmarking plot data preparation"); + } + + private static boolean isConfiguration(CSVRecord record) { + return LogicalDatastoreType.CONFIGURATION.toString().equals(record.get(DS_TYPE_POSITION)); + } + + private static List<DataEntry> processSimpleContainer(final List<CSVRecord> list) { + return list.stream().filter( + record -> DataProvider.SIMPLE_CONTAINER.equals(record.get(DATA_TYPE_POSITION)) && isConfiguration(record)) + .map(DataEntry::parseWriteData).collect(Collectors.toList()); + } + + private static List<DataEntry> processListContainer(final List<CSVRecord> list) { + return list.stream().filter( + record -> DataProvider.LIST_IN_CONTAINER.equals(record.get(DATA_TYPE_POSITION)) && isConfiguration(record)) + .map(DataEntry::parseWriteData) + .collect(Collectors.toList()); + } + + private static List<DataEntry> processComplexListContainer(final List<CSVRecord> list) { + return list.stream().filter( + record -> DataProvider.COMPLEX_LIST_IN_CONTAINER.equals(record.get(DATA_TYPE_POSITION)) + && isConfiguration(record)) + .map(DataEntry::parseWriteData) + .collect(Collectors.toList()); + } + + private static List<DataEntry> processReadStatistics(final List<CSVRecord> list) { + return list.stream() + .filter(record -> LogicalDatastoreType.OPERATIONAL.toString().equals(record.get(DS_TYPE_POSITION))) + .map(DataEntry::parseReadData) + .collect(Collectors.toList()); + } + + private static void writeStatistics(final List<DataEntry> data, final String path, final String fileName) + throws IOException { + final String absolutePath = path + '/' + fileName; + LOG.debug("Writing benchmark statistics to file {}", absolutePath); + final List<String> keys = new ArrayList<>(); + final List<String> scores = new ArrayList<>(); + data.stream().forEach(entry -> { + keys.add(entry.key); + scores.add(entry.score); + }); + LOG.debug("header: {}", keys); + LOG.debug("values: {}", scores); + + final StringBuilder buffer = new StringBuilder(); + final CSVPrinter csv = new CSVPrinter(buffer, CSVFormat.RFC4180); + csv.printRecord(keys); + csv.printRecord(scores); + csv.close(); + + try (final FileOutputStream out = new FileOutputStream(absolutePath)) { + out.write(buffer.toString().getBytes(Charsets.UTF_8)); + out.close(); + LOG.debug("Statistics written successfully"); + } + } + + private static final class DataEntry { + private final String key; + private final String score; + + private DataEntry(final String key, final String score) { + this.key = key; + this.score = score; + } + + static DataEntry parseWriteData(final CSVRecord record) { + return new DataEntry( + "persistence=" + record.get(PERSISTENCE_POSITION) + " freq=" + record.get(SUBMIT_FREQUENCY_POSITION), + record.get(SCORE_POSITION) + ); + } + + static DataEntry parseReadData(final CSVRecord record) { + return new DataEntry( + record.get(DATA_TYPE_POSITION), + record.get(SCORE_POSITION) + ); + } + } +} |