DelayBox

DelayBox is a High Throughput Distributed ONE-TIME Task Scheduler ⚡. It is an advanced system designed to manage and execute a vast number of tasks across a distributed network of servers with sub-second accuracy. Built on top of the Redis database, it leverages Redis's high-speed in-memory data store 🔥 for quick access and efficient task management.

Why DelayBox

DelayBox is a concept coming from the bigger domain of scheduling and handling delays in distributed systems, mainly towards the improvement of resilience, consistency, and fault tolerance of distributed applications. It is a middleware component that can be used to buffer messages or events with the intention of creating delays to manage distributed timing issues.

Read more

Here’s how DelayBox can help in designing a distributed system click for more.

1. Handling Eventual Consistency: click for more.
2. Mitigating Network Partitions: click for more.
3. Load Management and Throttling: click for more.
4. Failure Recovery and Redundancy: click for more.
5. Decoupling Components in Event-Driven Architecture: click for more.
6. Improving Reliability in Asynchronous Workflows: click for more.
7. Dealing with Temporal Anomalies: click for more.
8. Application in Geo-Distributed Systems: click for more.

General Idea

DelayBox consists of 2 topics (Scheduler and Executor)

Use-Case

🚀 This tool simplifies the workflow for niche task execution by removing the need to write code to handle Redis and Kafka development complexities. It manages these tasks for you through straightforward REST calls.

✅ It is ideal for applications requiring massive parallel processing capabilities, such as data processing pipelines, large-scale simulations, and real-time analytics.

Examples

Docker Compose

3 Worker Nodes with Kafka

clone the project

cd docker-multi-worker
docker-compose -f docker-compose.yml up -d

Wait until all worker nodes are up and running (you'll see in their console: "🚀 scheduler is ready!"). Then run the following command to create 1000 tasks which designed to execute in 10 second. (tasks just simply print the date in the worker nodes console)

sh ./create-task.sh

Note: your terminal console will only print {"message":"task created"} 1000 times, but the worker consoles shows the date. (in this example its docker logs)

You can try the docker compose example with 3 worker nodes example docker-compose example with kafka

3 Worker Nodes

Keep in mind that in this setup Delay-Box gets advantage og redis pub sub, so you are limited to single Redis node.

clone the project

cd docker-multi-worker-no-kafka
export COMPOSE_PROJECT_NAME=delay-box
docker-compose -p delay-box -f docker-compose.yml up -d

Wait until all worker nodes are up and running (you'll see in their console: "🚀 scheduler is ready!"). Then run the following command to create 1000 tasks which designed to execute in 10 second. (tasks just simply print the date in the worker nodes console)

sh ./create-task.sh

Note: your terminal console will only print {"message":"task created"} 1000 times, but the worker consoles shows the date. (in this example its docker logs)

You can try the docker compose example with 3 worker nodes example Docker-compose example without kafka

Local Example

Following task is type of SHELL_CMD which mean it will execute at any defined worker nodes on the given timestamp, (Payload is just the OS Date command)

curl -X POST http://localhost:8088/task  \
  -H "Content-Type: application/json" -d \
  '{"parameter":{"executionTimestamp":"1720672590913","taskType":"SHELL_CMD"},"pyload":"ZGF0ZQ=="}'

To Get list of pending tasks (First 100 tasks)

curl "http://localhost:8088/task"

pyload

This fild is stored as byte format and published in kafka topic taskExecutionTopic

parameter

parameter name	type	required	description
taskType	string	YES	Defines the task type Type of task
executionTimestamp	number	NO	With this parameter, the task is expected to be executed at the specified Unix epoch time (in milliseconds).
delay	number	NO	With this parameter, to delay task execution in millisecond.

Note: If neither executionTimestamp nor delay is provided, the task will be executed immediately.

taskType

name	description
PUB_SUB	This type is the basic schedule task, which you just publish payload in kafka topic schedulerTopic and the payload will be published kafka topic taskExecutionTopic when its scheduled to be executed. Note: schedulerTopic and taskExecutionTopic are already configured in the application config file.
SHELL_CMD	In this type, your payload, which is a Linux command, will be executed. Note: If you expect to run any application, it must be pre-installed on the worker machine prior to task execution.

General Architecture

Enable Kafka

To enable kafka support add the following config.

kafka:
  enabled: true
  brokers: localhost:9092
  group: scheduler
  schedulerTopic: Scheduler
  taskExecutionTopic: TaskExecution