https://www.youtube.com/watch?v=iJLL-KPqBpM

<aside> ⚠️ When you look at real-world distributed queue implementations like https://github.com/antirez/disque this setup may seem like overkill. Do you really need metadata service and backend server (queue messages service) to be separate? Disque handles these together.

</aside>

1. Listen to interviewer, to gather

Problem statement

Synchronous communication between producer and consumer is easy to implement and reason about, but sync comms makes it hard to deal with consumer failures.

• What to do with failed requests? Retry?
• Am I sending too many requests? Too little?

We need to build an intermediary component, a message queue. A producer sends data into the queue, and exactly one consumer receives and processes the message.

Ask clarifying questions

• Is at-least-once or exactly-once deliver required?
• How are retries to be handled?
• What’s the maximum write volume from the producer that a queue needs to handle?
• What’s the max poll volume from consumers?
• Ordering gaurantees?

Functional Requirements

• send_message(queue_name, msg_body)
• receive_message(queue_name)
• Optional:
  • create_queue(name)
  • delete_queue(name)

Non-functional requirements