tree: 9c352fc4ea50db278f052e37fcecae9dae4de831 [path history] [tgz]
  1. README.md
  2. pom.xml
  3. run_concurrent.sh
  4. run_concurrent_addmembers.sh
  5. run_concurrent_login.sh
  6. run_everyone_acl.sh
  7. run_writeacl.sh
  8. src/
oak-benchmarks/README.md

Oak Benchmark Jar

This jar is runnable and contains test related run modes.

The following runmodes are currently available:

* benchmark       : Run benchmark tests against different Oak repository fixtures.
* scalability     : Run scalability tests against different Oak repository fixtures.

See the subsections below for more details on how to use these modes.

Benchmark mode

The benchmark mode is used for executing various micro-benchmarks. It can be invoked like this:

$ java -jar oak-benchmarks-*.jar benchmark [options] [testcases] [fixtures]

The following benchmark options (with default values) are currently supported:

--azure                - Azure Connection String (default:
                           DefaultEndpointsProtocol=http;
                           AccountName=devstoreaccount1;
                           AccountKey=Eby8vdM02xNOcqFlqUwJPLlmEtlCDXJ1OUzFT50uSRZ6IFsuFq2UVErCz4I6tq/K1SZFPTOtr/KBHBeksoGMGw==;
                                       BlobEndpoint=http://127.0.0.1:
                                       10000/devstoreaccount1;)
--azureContainerName   - Azure container name (default: oak)
--azureRootPath        - Azure root path (default: /oak)
--host localhost       - MongoDB host
--port 27101           - MongoDB port
--db <name>            - MongoDB database (default is a generated name)
--mongouri             - MongoDB URI (takes precedence over host, port and db)
--dropDBAfterTest true - Whether to drop the MongoDB database after the test
--base target          - Path to the base file (Tar setup),
--mmap <64bit?>        - TarMK memory mapping (the default on 64 bit JVMs)
--cache 100            - cache size (in MB)
--wikipedia <file>     - Wikipedia dump
--runAsAdmin false     - Run test as admin session
--itemsToRead 1000     - Number of items to read
--report false         - Whether to output intermediate results
--csvFile <file>       - Optional csv file to report the benchmark results
--concurrency <levels> - Comma separated list of concurrency levels
--metrics false        - Enable metrics based stats collection
--rdbjdbcuri           - JDBC URL for RDB persistence (defaults to local file-based H2)
--rdbjdbcuser          - JDBC username (defaults to "")
--rdbjdbcpasswd        - JDBC password (defaults to "")
--rdbjdbctableprefix   - for RDB persistence: prefix for table names (defaults to "")
--vgcMaxAge            - Continuous DocumentNodeStore VersionGC max age in sec (RDB only)

Please run --help to list all options.

These options are passed to the test cases and repository fixtures that need them. For example the Wikipedia dump option is needed by the WikipediaImport test case and the MongoDB address information by the MongoMK and SegmentMK -based repository fixtures. The cache setting controls the bundle cache size in Jackrabbit, the NodeState cache size in MongoMK, and the segment cache size in SegmentMK.

The --concurrency levels can be specified as comma separated list of values, eg: --concurrency 1,4,8, which will execute the same test with the number of respective threads. Note that the beforeSuite() and afterSuite() are executed before and after the concurrency loop. eg. in the example above, the execution order is: beforeSuite(), 1x runTest(), 4x runTest(), 8x runTest(), afterSuite(). Tests that create their own background threads, should be executed with --concurrency 1 which is the default.

You can use extra JVM options like -Xmx settings to better control the benchmark environment. It‘s also possible to attach the JVM to a profiler to better understand benchmark results. For example, I’m using -agentlib:hprof=cpu=samples,depth=100 as a basic profiling tool, whose results can be processed with perl analyze-hprof.pl java.hprof.txt to produce a somewhat easier-to-read top-down and bottom-up summaries of how the execution time is distributed across the benchmarked codebase.

Some system properties are also used to control the benchmarks. For example:

-Dwarmup=5         - warmup time (in seconds)
-Druntime=60       - how long a single benchmark should run (in seconds)
-Dprofile=true     - to collect and print profiling data

The test case names like ReadPropertyTest, SmallFileReadTest and SmallFileWriteTest indicate the specific test case being run. You can specify one or more test cases in the benchmark command line, and oak-run will execute each benchmark in sequence. The benchmark code is located under org.apache.jackrabbit.oak.benchmark in the oak-run component. Each test case tries to exercise some tightly scoped aspect of the repository. You might remember many of these tests from the Jackrabbit benchmark reports like http://people.apache.org/~jukka/jackrabbit/report-2011-09-27/report.html that we used to produce earlier.

Finally the benchmark runner supports the following repository fixtures:

FixtureDescription
JackrabbitJackrabbit with the default embedded Derby bundle PM
Oak-MemoryOak with default in-memory storage
Oak-MemoryNSOak with default in-memory NodeStore
Oak-MongoOak with the default Mongo backend
Oak-Mongo-DSOak with the default Mongo backend and DataStore
Oak-MongoNSOak with the Mongo NodeStore
Oak-Segment-TarOak with the Segment Tar backend
Oak-Segment-Tar-DSOak with the Segment Tar backend and DataStore
Oak-Segment-AzureOak with the Azure Segment backend
Oak-RDBOak with the DocumentMK/RDB persistence
Oak-RDB-DSOak with the DocumentMK/RDB persistence and DataStore
Oak-Composite-StoreOak with the Composite Node store with Segment Tar backend
Oak-Composite-Memory-StoreOak with the Composite Node store with in-memory NodeStore
Oak-Composite-Mongo-StoreOak with the Composite Node store with Mongo backend

(Note that for Oak-RDB, the required JDBC drivers either need to be embedded into oak-run, or be specified separately in the class path. Furthermore, dropDBAfterTest is interpreted to drop the tables, not the database itself, if and only if they have been auto-created)

Once started, the benchmark runner will execute each listed test case against all the listed repository fixtures. After starting up the repository and preparing the test environment, the test case is first executed a few times to warm up caches before measurements are started. Then the test case is run repeatedly for one minute and the number of milliseconds used by each execution is recorded. Once done, the following statistics are computed and reported:

ColumnDescription
Cconcurrency level
minminimum time (in ms) taken by a test run
10%time (in ms) in which the fastest 10% of test runs
50%time (in ms) taken by the median test run
90%time (in ms) in which the fastest 90% of test runs
maxmaximum time (in ms) taken by a test run
Ntotal number of test runs in one minute (or more)

The most useful of these numbers is probably the 90% figure, as it shows the time under which the majority of test runs completed and thus what kind of performance could reasonably be expected in a normal usage scenario. However, the reason why all these different numbers are reported, instead of just the 90% one, is that often seeing the distribution of time across test runs can be helpful in identifying things like whether a bigger cache might help.

Finally, and most importantly, like in all benchmarking, the numbers produced by these tests should be taken with a large dose of salt. They DO NOT directly indicate the kind of application performance you could expect with (the current state of) Oak. Instead they are designed to isolate implementation-level bottlenecks and to help measure and profile the performance of specific, isolated features.

How to add a new benchmark

To add a new test case to this benchmark suite, you'll need to implement the Benchmark interface and add an instance of the new test to the allBenchmarks array in the BenchmarkRunner class in the org.apache.jackrabbit.oak.benchmark package.

The best way to implement the Benchmark interface is to extend the AbstractTest base class that takes care of most of the benchmarking details. The outline of such a benchmark is:

class MyTest extends AbstracTest {
    @Override
    protected void beforeSuite() throws Exception {
        // optional, run once before all the iterations,
        // not included in the performance measurements
    }
    @Override
    protected void beforeTest() throws Exception {
        // optional, run before runTest() on each iteration,
        // but not included in the performance measurements
    }
    @Override
    protected void runTest() throws Exception {
        // required, run repeatedly during the benchmark,
        // and the time of each iteration is measured.
        // The ideal execution time of this method is
        // from a few hundred to a few thousand milliseconds.
        // Use a loop if the operation you're hoping to measure
        // is faster than that.
    }
    @Override
    protected void afterTest() throws Exception {
        // optional, run after runTest() on each iteration,
        // but not included in the performance measurements
    }
    @Override
    protected void afterSuite() throws Exception {
        // optional, run once after all the iterations,
        // not included in the performance measurements
    }
}

The rough outline of how the benchmark will be run is:

test.beforeSuite();
for (...) {
    test.beforeTest();
    recordStartTime();
    test.runTest();
    recordEndTime();
    test.afterTest();
}
test.afterSuite();

You can use the loginWriter() and loginReader() methods to create admin and anonymous sessions. There's no need to logout those sessions (unless doing so is relevant to the benchmark) as they will automatically be closed after the benchmark is completed and the afterSuite() method has been called.

Similarly, you can use the addBackgroundJob(Runnable) method to add background tasks that will be run concurrently while the main benchmark is executing. The relevant background thread works like this:

while (running) {
    runnable.run();
    Thread.yield();
}

As you can see, the run() method of the background task gets invoked repeatedly. Such threads will automatically close once all test iterations are done, before the afterSuite() method is called.

Scalability mode

The scalability mode is used for executing various scalability suites to test the performance of various associated tests. It can be invoked like this:

$ java -jar oak-benchmarks-*.jar scalability [options] [suites] [fixtures]

The following scalability options (with default values) are currently supported:

--host localhost       - MongoDB host
--port 27101           - MongoDB port
--db <name>            - MongoDB database (default is a generated name)
--dropDBAfterTest true - Whether to drop the MongoDB database after the test
--base target          - Path to the base file (Tar setup),
--mmap <64bit?>        - TarMK memory mapping (the default on 64 bit JVMs)
--cache 100            - cache size (in MB)
--csvFile <file>       - Optional csv file to report the benchmark results
--rdbjdbcuri           - JDBC URL for RDB persistence (defaults to local file-based H2)
--rdbjdbcuser          - JDBC username (defaults to "")
--rdbjdbcpasswd        - JDBC password (defaults to "")

These options are passed to the various suites and repository fixtures that need them. For example the the MongoDB address information by the MongoMK and SegmentMK -based repository fixtures. The cache setting controls the NodeState cache size in MongoMK, and the segment cache size in SegmentMK.

You can use extra JVM options like -Xmx settings to better control the scalability suite test environment. It‘s also possible to attach the JVM to a profiler to better understand benchmark results. For example, I’m using -agentlib:hprof=cpu=samples,depth=100 as a basic profiling tool, whose results can be processed with perl analyze-hprof.pl java.hprof.txt to produce a somewhat easier-to-read top-down and bottom-up summaries of how the execution time is distributed across the benchmarked codebase.

The scalability suite creates the relevant repository load before starting the tests. Each test case tries to benchmark and profile a specific aspect of the repository.

Each scalability suite is configured to run a number of related tests which require the same base load to be available in the repository. Either the entire suite can be executed or individual tests within the suite can be run. If the suite names are specified like ScalabilityBlobSearchSuite then all the tests configured for the suite are executed. To execute particular tests in the suite, suite names appended with tests of the form suite:test1,test2 must be specified like ScalabilityBlobSearchSuite:FormatSearcher,NodeTypeSearcher. You can specify one or more suites in the scalability command line, and oak-run will execute each suite in sequence.

Finally the scalability runner supports the following repository fixtures:

FixtureDescription
Oak-MemoryOak with default in-memory storage
Oak-MemoryNSOak with default in-memory NodeStore
Oak-MongoOak with the default Mongo backend
Oak-Mongo-DSOak with the default Mongo backend and DataStore
Oak-MongoNSOak with the Mongo NodeStore
Oak-Segment-TarOak with the Tar backend (aka Segment NodeStore)
Oak-Segment-Tar-DSOak with the Tar backend (aka Segment NodeStore) and DataStore
Oak-RDBOak with the DocumentMK/RDB persistence
Oak-RDB-DSOak with the DocumentMK/RDB persistence and DataStore

(Note that for Oak-RDB, the required JDBC drivers either need to be embedded into oak-run, or be specified separately in the class path.)

Once started, the scalability runner will execute each listed suite against all the listed repository fixtures. After starting up the repository and preparing the test environment, the scalability suite executes all the configured tests to warm up caches before measurements are started. Then each configured test within the suite are run and the number of milliseconds used by each execution is recorded. Once done, the following statistics are computed and reported:

ColumnDescription
minminimum time (in ms) taken by a test run
10%time (in ms) in which the fastest 10% of test runs
50%time (in ms) taken by the median test run
90%time (in ms) in which the fastest 90% of test runs
maxmaximum time (in ms) taken by a test run
Ntotal number of test runs in one minute (or more)

Also, for each test, the execution times are reported for each iteration/load configured.

ColumnDescription
Loadtime (in ms) taken by a test run

The latter is more useful of these numbers as it shows how the individual execution times are scaling for each load.

How to add a new scalability suite

The scalability code is located under org.apache.jackrabbit.oak.scalabiity in the oak-run component.

To add a new scalability suite, you'll need to implement the ScalabilitySuite interface and add an instance of the new suite to the allSuites array in the ScalabilityRunner class, along with the test benchmarks, in the org.apache.jackrabbit.oak.scalability package. To implement the test benchmarks, it is required to extend the ScalabilityBenchmark abstract class and implement the execute() method. In addition, the methods beforeExecute() and afterExecute() can overridden to do processing before and after the benchmark executes.

The best way to implement the ScalabilitySuite interface is to extend the ScalabilityAbstractSuite base class that takes care of most of the benchmarking details. The outline of such a suite is:

class MyTestSuite extends ScalabilityAbstractSuite {
    @Override
    protected void beforeSuite() throws Exception {
        // optional, run once before all the iterations,
        // not included in the performance measurements
    }
    @Override
    protected void beforeIteration(ExecutionContext) throws Exception {
        // optional, Typically, this can be configured to create additional
        // loads for each iteration.
        // This method will be called before each test iteration begins
    }

    @Override
    protected void executeBenchmark(ScalabilityBenchmark benchmark,
        ExecutionContext context) throws Exception {
        // required, executes the specified benchmark
    }

    @Override
    protected void afterIteration() throws Exception {
        // optional, executed after runIteration(),
        // but not included in the performance measurements
    }
    @Override
    protected void afterSuite() throws Exception {
        // optional, run once after all the iterations are complete,
        // not included in the performance measurements
    }
}

The rough outline of how the individual suite will be run is:

test.beforeSuite();
for (iteration...) {
    test.beforeIteration();
    for (benchmarks...) {
          recordStartTime();
          test.executeBenchmark();
          recordEndTime();
    }
    test.afterIteration();
}
test.afterSuite();

You can specify any context information to the test benchmarks using the ExecutionContext object passed as parameter to the beforeIteration() and the executeBenchmark() methods. ExecutionBenchmark exposes two methods getMap() and setMap() which can be used to pass context information.

You can use the loginWriter() and loginReader() methods to create admin and anonymous sessions. There's no need to logout those sessions (unless doing so is relevant to the test) as they will automatically be closed after the suite is complete and the afterSuite() method has been called.

Similarly, you can use the addBackgroundJob(Runnable) method to add background tasks that will be run concurrently while the test benchmark is executing. The relevant background thread works like this:

while (running) {
    runnable.run();
    Thread.yield();
}

As you can see, the run() method of the background task gets invoked repeatedly. Such threads will automatically close once all test iterations are done, before the afterSuite() method is called.

ScalabilityAbstractSuite defines some system properties which are used to control the suites extending from it :

-Dincrements=10,100,1000,1000     - defines the varying loads for each test iteration
-Dprofile=true                    - to collect and print profiling data
-Ddebug=true                      - to output any intermediate results during the suite
                                    run

License

(see the top-level LICENSE.txt for full license details)

Collective work: Copyright 2012 The Apache Software Foundation.

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.