Yardstick Hazelcast Benchmarks

Yardstick Hazelcast is a set of Hazelcast Data Grid benchmarks written on top of Yardstick framework.

Yardstick Framework

Visit Yardstick Repository for detailed information on how to run Yardstick benchmarks and how to generate graphs.

The documentation below describes configuration parameters in addition to standard Yardstick parameters.

Installation

1. Create a local clone of Yardstick Hazelcast repository
2. Import Yardstick Hazelcast POM file into your project
3. Run mvn package command

Provided Benchmarks

The following benchmarks are provided:

1. HazelcastGetBenchmark - benchmarks atomic distributed cache get operation
2. HazelcastPutBenchmark - benchmarks atomic distributed cache put operation
3. HazelcastSetBenchmark - benchmarks atomic distributed cache set operation
4. HazelcastPutGetBenchmark - benchmarks atomic distributed cache put and get operations together
5. HazelcastSetGetBenchmark - benchmarks atomic distributed cache set and get operations together
6. HazelcastPutGetBatchBenchmark - benchmarks atomic distributed cache set on many keys and getAll operations together
7. HazelcastPutTxBenchmark - benchmarks transactional distributed cache put operation
8. HazelcastSetTxBenchmark - benchmarks transactional distributed cache set operation
9. HazelcastSetGetTxOptimisticBenchmark - benchmarks transactional distributed cache set and get operations together
10. HazelcastSetGetTxPessimisticBenchmark - benchmarks transactional distributed cache set and getForUpdate operations together
11. HazelcastSqlQueryBenchmark - benchmarks distributed SQL query over cached data
12. HazelcastSqlQueryPutBenchmark - benchmarks distributed SQL query with simultaneous cache updates
13. HazelcastSqlQuerySetBenchmark - benchmarks distributed SQL query with simultaneous cache sets
14. HazelcastPutAllBenchmark - benchmarks atomic distributed cache putAll operation
15. HazelcastSetAllTxBenchmark - benchmarks transactional distributed cache set operations in one transaction
16. HazelcastGetAllSetAllTxOptimisticBenchmark - benchmarks transactional distributed cache set and get operations together
17. HazelcastGetAllSetAllTxPessimisticBenchmark - benchmarks transactional distributed cache set and getForUpdate operations together
18. HazelcastAffinityCallBenchmark - benchmarks affinity calls

Writing Hazelcast Benchmarks

All benchmarks extend HazelcastAbstractBenchmark class. A new benchmark should also extend this abstract class and implement test method. This is the method that is actually benchmarked.

Running Hazelcast Benchmarks

Before running Hazelcast benchmarks, run mvn package command. This command will compile the project and also will unpack scripts from yardstick-resources.zip file to bin directory.

Properties And Command Line Arguments

Note that this section only describes configuration parameters specific to Hazelcast benchmarks, and not for Yardstick framework. To run Hazelcast benchmarks and generate graphs, you will need to run them using Yardstick framework scripts in bin folder.

Refer to Yardstick Documentation for common Yardstick properties and command line arguments for running Yardstick scripts.

The following Hazelcast benchmark properties can be defined in the benchmark configuration:

• -nn <num> or --nodeNumber <num> - Number of nodes (automatically set in benchmark.properties), used to wait for the specified number of nodes to start
• -b <num> or --backups <num> - Number of backups for every key
• -hzcfg <path> or --hzConfig <path> - Path to Hazelcast configuration file
• -hzclicfg <path> or --hzClientConfig <path> - Path to Hazelcast client configuration file
• -sb or --syncBackups - Flag indicating whether synchronous backups are used, asynchronous is a default
• -nt or --nodeType - Type of driver nodes. One of: SERVER, CLIENT, LITE_MEMBER
• -r <num> or --range <num> - Range of keys that are randomly generated for cache operations
• -rb or --readBackups - Flag indicating whether backup reads are enabled or not, disabled by default

For example if we need to run 2 HazelcastNode servers on localhost with HazelcastPutBenchmark benchmark on localhost, with number of backups set to 1, backups are synchronous, then the following configuration should be specified in benchmark.properties file:

SERVER_HOSTS=localhost,localhost

# Note that -dn and -sn, which stand for data node and server node, are
# native Yardstick parameters and are documented in Yardstick framework.
CONFIGS="-b 1 -sb -dn HazelcastPutBenchmark -sn HazelcastNode"

Running on Amazon

This repo contains all necessary scripts and properties files for a comparison Hazelcast with other products.

For running on Amazon EC2 need to perform the following steps:

• Run Amazon EC2 instances. Choose number of instances and hardware according to your requirements.

The following actions need to perform on all instances:

• Install Java, Maven and Git on all instances.

For example for Ubuntu:

# apt-get install java
# apt-get install mvn
# apt-get install git

• Clone this repository and build project on one of the Amazon instances. Yardstick will copy all the built binaries on all nodes automatically.

git clone https://github.com/gridgain/yardstick-hazelcast

mvn clean package

• Change SERVER_HOSTS and DRIVER_HOSTS properties in config/benchmark.properties file. SERVER_HOSTS is comma-separated list of IP addresses where servers should be started, one server per host. DRIVER_HOSTS is comma-separated list of IP addresses where drivers should be started, one driver per host, if the property is not defined then the driver will be run on localhost. Property file contains many useful information about benchmarks such as list of benchmarks, JVM opts and etc. More details there Properties And Command Line Arguments

• Update IP addresses in network section from config/hazelcast-config.xml and config/hazelcast-client-config.xml files. For example:

config/hazelcast-client-config.xml

...
  <network>
    <cluster-members>
      <address>XXX.XXX.XXX.1:57500</address>
      <address>XXX.XXX.XXX.2:57500</address>
      <address>XXX.XXX.XXX.3:57500</address>
    </cluster-members>
    <connection-timeout>10000</connection-timeout>
    <connection-attempt-limit>50</connection-attempt-limit>
  </network>
...

config/hazelcast-config.xml

...
  <network>
    <port auto-increment="true">57500</port>
    <join>
      <multicast enabled="false"/>
      <tcp-ip enabled="true">
        <member>XXX.XXX.XXX.1:57500</member>
        <member>XXX.XXX.XXX.2:57500</member>
        <member>XXX.XXX.XXX.3:57500</member>
      </tcp-ip>
    </join>
  </network>
...

• This repository contains two main set benchmark for sync and async backups (see more details about it there). By default used set of benchmark for async backups. If you want to run benchmarks for sync backup then need to use benchmark-sync.properties file. Perform ./bin/benchmark-run-all.sh script for async backups or ./bin/benchmark-run-all.sh ./config/benchmark-sync.properties for sync backups. For more details about running scripts see Running Yardstick Benchmarks.
• After execution the script in result folder will be saved to results of benchmarks. For visualisation of results can be used bin/jfreechart-graph-gen.sh script. For more details about the script see JFreeChart Graphs.

Issues

Use GitHub issues to file bugs.

License

Yardstick Hazelcast is available under Apache 2.0 Open Source license.