คิวงาน: การกระจายงานอย่างเป็นธรรม

Welcome to Work Queues! In distributed systems, you often have tasks that take time to complete, like processing an image or generating a report.

Work Queues are a pattern that helps distribute these time-consuming tasks among multiple workers (consumers) efficiently, preventing any single worker from getting overloaded.

Imagine you have many jobs to do, but only one employee. If all jobs go to that single employee, they'll get overwhelmed and tasks will pile up.

• Load Balancing: Work queues allow you to add more workers to share the load.
• Asynchronous Processing: The producer doesn't wait for a task to finish, it just adds it to the queue.
• Reliability: If one worker fails, others can pick up tasks.

The setup for a work queue is simple:

• One producer sends messages (tasks) to a single queue.
• Multiple consumers listen to this same queue.
• RabbitMQ ensures that each message is delivered to only one of the waiting consumers.

This way, tasks are never duplicated and are processed in parallel.

By default, RabbitMQ distributes messages to consumers using a round-robin strategy. This means messages are sent to consumers in a rotating fashion:

• Consumer 1 gets the first message.
• Consumer 2 gets the second message.
• Consumer 1 gets the third message, and so on.

This aims for an even distribution of tasks among all active consumers.

Let's create a producer that sends 10 tasks to our queue. Each task will be a simple string like 'Processing image 1'.

Run this code once to populate the queue with tasks.

What's happening in our producer code?

• We establish a connection and a channel to interact with RabbitMQ.
• channel.queueDeclare(QUEUE_NAME, false, false, false, null); ensures the queue exists. The false flags keep it non-durable and non-exclusive for simplicity here.
• A loop sends 10 messages to the task_queue. Each message represents a distinct task.

Now, let's create a consumer, which we'll call a 'worker'. This worker will listen for tasks from the task_queue.

We'll simulate a long-running task using Thread.sleep(). Run this code in one terminal window.

Let's look at the consumer's logic:

• Similar to the producer, it connects and declares the queue.
• A DeliverCallback defines the actions when a message arrives.
• Inside the callback, Thread.sleep(1000) simulates a 1-second task.
• channel.basicConsume(QUEUE_NAME, true, deliverCallback, ...) starts consuming. The true means messages are automatically acknowledged after delivery.

Here's the core demonstration of work queues:

1. Run the TaskConsumer code in two separate terminal windows (or instances).

2. Then, run the TaskProducer code once.

You will observe that the 10 tasks are divided between your two worker instances, each processing roughly 5 tasks due to RabbitMQ's round-robin dispatch.

You've successfully implemented Work Queues, a fundamental pattern for distributing tasks across multiple consumers!

• Work queues enable asynchronous processing and prevent single points of failure.
• RabbitMQ's default round-robin dispatch strategy ensures tasks are distributed fairly.
• By running multiple consumer instances, you can easily scale your task processing capacity.

Next, we'll dive into making your message handling even more robust with acknowledgements and message durability!