Highload Software Architecture 8 Lesson 13 Homework

Queues: Redis vs Beanstalkd

Test project setup

The demo is written in Kotlin/Quarkus and uses testcontainers to launch docker containers in various modes.

The com/example/Demo.kt/runDemo runs different queues with different load and writes a report in a Markdown format. The run report is located in the REPORT.md file.

Queue configurations and load profiles

This demo applies load to 6 queue configurations (see Docker.kt):

Redis NoP - Redis Pub-Sub queue with no persistence, --save "" --appendonly no
Redis AOF - Redis Pub-Sub queue with AOF persistence, --save "" --appendonly yes
Redis RDB - Redis Pub-Sub queue with RDB persistence, --save "5 1000" --save "1 100" --appendonly no
Beanstalk NoP - Beanstalkd queue with no persistence, -F
Beanstalk 0s - Beanstalkd queue with immediate save to file, -f 0
Beanstalk 1s - Beanstalkd queue with save every second, -f 1

Redis is tested with Pub-Sub and Lpush/Rpop queues.

Queues are tested with different number of concurrent producers and consumers:

10 producers % consumers
50 producers & consumers
100 producers & consumers

Producer emits as many messages as possible during 2 minutes, then consumer consumes all messages. Resulting time is used to calculate throughput in Operations per second.

Tests also use different message sizes:

1-2Kb
10-20Kb
50-100Kb
500Kb-1Mb

All queues run with default settings, so timeouts sometimes happen on large concurrency levels and large message sizes. These timeouts are also reported in the report as a percentage of total operations.

To make the testing process less boring and more informative, the demo also prints outputs to console, signalling about messages being sent and received:

. - message generated
+ - message sent to queue
- - message received from queue
' - message deleted (applies to Lpush/Rpop and Beanstalkd queues)

How to run

Build and run demo application (Requires Java 17+)

./gradlew build && \
java -jar build/quarkus-app/quarkus-run.jar

You can also run application in dev mode that enables live coding using:

./gradlew quarkusDev

Final results

The data from table with final results from the REPORT.md can also be found in the Google Sheets: