Yardstick Hazelcast Benchmarks
Yardstick Hazelcast is a set of Hazelcast Data Grid benchmarks written on top of 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.
- Create a local clone of Yardstick Hazelcast repository
- Import Yardstick Hazelcast POM file into your project
The following benchmarks are provided:
HazelcastGetBenchmark- benchmarks atomic distributed cache get operation
HazelcastPutBenchmark- benchmarks atomic distributed cache put operation
HazelcastPutGetBenchmark- benchmarks atomic distributed cache put and get operations together
HazelcastPutTxBenchmark- benchmarks transactional distributed cache put operation
HazelcastPutGetTxBenchmark- benchmarks transactional distributed cache put and get operations together
HazelcastSqlQueryBenchmark- benchmarks distributed SQL query over cached data
HazelcastSqlQueryPutBenchmark- benchmarks distributed SQL query with simultaneous cache updates
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
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
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:
--nodeNumber <num>- Number of nodes (automatically set in
benchmark.properties), used to wait for the specified number of nodes to start
--backups <num>- Number of backups for every key
--hzConfig <path>- Path to Hazelcast configuration file
--hzClientConfig <path>- Path to Hazelcast client configuration file
--syncBackups- Flag indicating whether synchronous backups are used, asynchronous is a default
--clientMode- Flag indicating whether Hazelcast client is used
--range <num>- Range of keys that are randomly generated for cache operations
--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
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. You can easy run benchmark by using yardstick-docker extension, but it might have an influence on performance.
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.
For example for Ubuntu: # apt-get install java # apt-get install mvn # apt-get install git
- Clone this repository (on all nodes path should be the same) and build project.
git clone https://github.com/gridgain/yardstick-hazelcast mvn clean package
SERVER_HOSTSis comma-separated list of IP addresses where servers should be started, one server per host.
DRIVER_HOSTSis 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 optsand etc. More details there Properties And Command Line Arguments
config/hazelcast-ec2.xmlconfig which using
AWS discoveryor update IP addresses in network section from
config/hazelcast-client-config.xmlfiles. 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> ...
./bin/benchmark-run-all.shscript. For more details about running scripts see Running Yardstick Benchmarks.
- After execution the script in
resultfolder will be saved to results of benchmarks. For visualisation of results can be used
bin/jfreechart-graph-gen.shscript. For more details about the script see JFreeChart Graphs.
Use GitHub issues to file bugs.
Yardstick Hazelcast is available under Apache 2.0 Open Source license.