Redis perf_test App - padogrid/padogrid GitHub Wiki

◀️ Redis CLASSPATH 🔗 Spark CLASSPATH ▶️ 🔗 Home ▶️

The perf_test app provides Redisson client programs to perform the following:

• Ingest mock data of any size
• Ingest transactional data of any size
• Ingest mock data with entity relationships (ER)
• Ingest mock data directly to databases
• Simulate complex appliation workflows that invoke Redisson operations without coding
• Measure Redisson latencies and throughputs in a multi-threaded user session environment

The perf_test app provides a pair of scripts to ingest and transact mock data for testing Redisson throughputs and latencies. It provides a quick way to run your performance tests by configuring a few properties such as the payload size, the number of objects (entries), and the number of worker threads. Out of the box, these properties have already been pre-configured in etc/ingestion.properties and etc/tx.properties, which you can modify as needed to meet your test criteria.

The perf_test app also includes the test_group script that allows you to configure one or more groups of Redisson data structure operations and execute them in sequence and/or in parallel. A group is analogous to a function that makes multiple data structure method calls in the order they are specified in the etc/group.properties file. The etc directory also contains the group-put.properties and group-get.properties files for demonstrating complex workflows that invoke 22 put calls and 22 get calls on 22 different maps. There are also several example group-*.properties files for each data structure. You can also configure the Near Cache in etc/redisson-client.yaml to measure the improved throughput.

The perf_test app can directly upsert mock data into your database of choice using Hibernate, which automatically creates tables as needed. This capability allows you to quickly synchrnize Redisson with your database and perform latency tests.

All of the transaction test cases are performed on three (3) distributed maps with a simple PBM (Pharmacy Benefit Management) data model that associates the client group number with group members. Simply put, all of the members that belong to a group are co-located in the same Redisson partition. This enables each Redisson member to complete transactions with their local datasets without encountering additional netowork hops.

The following table describes a list of preconfigured properties files in the etc/ directory.

You can introduce your own test criteria by modifying the properties the above files or supply another properties file by specifying the -prop option of the scripts described below.

The bin_sh/ directory contains the following scripts. By default, these scripts simply prints the configuration information obtained from the etc/perf.properties file. To run the test cases, you must specify the -run option.

./test_ingestion -?

Output:

Usage:
   test_ingestion [-run] [-prop <properties-file>] [-?]

   Displays or runs data ingestion test cases specified in the properties file.
   The default properties file is
      ../etc/ingestion.properties

       -run              Run test cases.

       <properties-file> Optional properties file path.

   To run the the test cases, specify the '-run' option. Upon run completion, the results
   will be outputted in the following directory:
      /Users/dpark/Padogrid/workspaces/rwe-redis/myws/apps/perf_test_pod/results

Notes:
   The 'perf_test' app requires you to first run the 'build_app' command to download Redisson.
   If you are behind a firewall then you can manually download 'redisson-<version>.jar' and
   place it in the workspace lib directory. Please see 'pom.xml' for Redisson artifact details.

./test_group -?

Output:

Usage:
   test_group [-run|-list] [-db|-delete] [-prop <properties-file>] [-?]

   Displays or runs group test cases specified in the properties file.
   A group represents a function that executes one or more Redis data
   structure operations. This program measures average latencies and throughputs
   of group (or function) executions.
   The default properties file is
      ../etc/group.properties

       -run              Runs test cases.

       -list             Lists data structures and their sizes.

       -db               Runs test cases on database instead of Redis. To use this
                         option, each test case must supply a data object factory class
                         by specifying the 'factory.class' property and Hibernate must
                         be configured by running the 'build_app' command.

       -delete           Deletes (destroys) all the data structures pertaining to the group
                         test cases that were created in the Redis cluster. If the '-run'
                         option is not specified, then it has the same effect as the '-list'
                         option. It only lists data strcutures and their without deleting them.

       <properties-file> Optional properties file path.

   To run the the test cases, specify the '-run' option. Upon run completion, the results
   will be outputted in the following directory:
      /Users/dpark/Padogrid/workspaces/rwe-redis/myws/apps/perf_test_pod/results

Notes:
   The 'perf_test' app requires you to first run the 'build_app' command to download Redisson.
   If you are behind a firewall then you can manually download 'redisson-<version>.jar' and
   place it in the workspace lib directory. Please see 'pom.xml' for Redisson artifact details

Upon successful run, the test results are outputted in the results/ directory. The following shows an example.

cat ../results/ingestion-profile-190630-151618_x.txt

Output:

******************************************
Data Ingestion Test
******************************************

                   Test Case: putall
                         Map: profile
           PutAll Batch Size: 100
              Test Run Count: 1
    Test Run Interval (msec): 0
   Total Entry Count Per Run: 10000
                Thread Count: 16
        Payload Size (bytes): 10240
                      Prefix: x
      Entry Count per Thread: 625

Start Time: Tue Jun 21 21:08:31 EDT 2022

Actual Total Entry (Put) Count: 10000

Time unit: msec
   Thread 1: 714
   Thread 2: 758
   Thread 3: 715
   Thread 4: 756
   Thread 5: 750
   Thread 6: 721
   Thread 7: 753
   Thread 8: 666
   Thread 9: 751
   Thread 10: 750
   Thread 11: 694
   Thread 12: 702
   Thread 13: 750
   Thread 14: 757
   Thread 15: 654
   Thread 16: 750

        Max time (msec): 758
   Throughput (msg/sec): 13192.612
  *Throughput (KiB/sec): 131926.121
  *Throughput (MiB/sec): 128.834
 Latency per put (msec): 0.076
   **Total Volume (MiB): 100000
   **Total Volume (MiB): 97.656
   **Total Volume (GiB): 0.095
   Payload Size (bytes): 10240

 * Throughput does not take the keys into account.
   The actual rate is higher.
** Total Volume do not take the keys into account.
   The actual volume is higher.

Stop Time: Tue Jun 21 21:08:31 EDT 2022

The group_test -db command directly loads mock data into database tables without connecting to Redisson. You can use this command to pre-populate the database before testing database synchronization tests in Redisson. This command is also useful for testing the CDC use case in which database changes are automatically ingested into Redisson via a CDC product such as Debezium ansd Striim.

# Edit setenv.sh to set the correct hibernate configuration file.
vi setenv.sh

By default, setenv.sh is configured with hibernate.cfg-mysql.xml. Change it to another if you are using a different database. Please see the etc directory for all the available database configuration files. If your database is not listed, then you can create one by copying one of the hibernate-* files and specifying that file name in the setenv.sh file.

Make sure to set the correct database user name and password in the Hibernate configuration file.

# Hibernate
JAVA_OPTS="$JAVA_OPTS -Dredisson-addon.hibernate.config=$APP_ETC_DIR/hibernate.cfg-mysql.xml"

Run test_group -db.

./test_group -db -run -prop ../etc/group-factory.properties

If you want to add entity relationships to your data, then you can implement DataObjectFactory or extend AbstractDataObjectFactory and pass the object key to the createEntry() method using the factory.er.operation property. The perf_test app includes an ER example that creates one-to-many ER between Customer and Order objects by setting Customer.customerId to Order.customerId while ingesting mock data. Please see org.redisson.demo.nw.impl.OrderFactoryImpl for details. You can run the example as follows:

./test_group -run -prop ../etc/group-factory-er.properties

The ER capbility provides you a quick way to ingest co-located data into Redisson and test server-side operations that take advatange of data affinity.

◀️ Redis CLASSPATH 🔗 Spark CLASSPATH ▶️ 🔗 Home ▶️