15 Asynchronous Workflows and Message Queues

Created by

@DecisiveGreatWallOfChina1467

Study Notes

Asynchronous workflows speed up operations that are time-consuming by doing the work in advance
They reduce request times for expensive tasks
Examples of time-consuming tasks include periodic data aggregation

Message queues hold and deliver messages
Useful when an operation is too slow to perform in real-time
Workflow:
- Application publishes a job to the queue
- A worker picks up the job from the queue, processes it, then marks it as complete.
- The user isn't blocked while the job runs in the background.
- Users can optionally perform a small amount of processing related to a task to appear like it is in progress (like a tweet appearing on your timeline immediately, but the full delivery to other timelines takes time)

Redis: Simple broker, but messages may be lost
RabbitMQ: Popular, but requires adapting to the AMQP protocol and managing own nodes.
Amazon SQS: Hosted, but can have high latency, potential for message duplication.

Queues that grow significantly may exceed available memory.
This can lead to problems like cache misses and slower performance
Back pressure helps limit queue size, maintaining high throughput and response times
When the queue fills up, clients receive server-busy or HTTP 503 error codes, and can retry later using exponential backoff.

Introducing queues can add delays and complexity
Certain use cases like some calculations, or real-time tasks, might benefit more from synchronous operations.