Katkat
Back

Design a scalable coffee ordering notification system

SalesforceSoftware Engineer · Onsite
MediumSoftware EngineerOnsiteSystem Design

Last updated

This problem tests your ability to design a distributed, asynchronous notification system capable of handling high-throughput order status updates and ensuring reliable message delivery to end users.

Design a Scalable Coffee Ordering Notification System

Design a backend service that notifies customers about their coffee order status (e.g., 'Order Received', 'Preparing', 'Ready for Pickup').

Functional Requirements

  • The system must track the state transition of an order.
  • Customers must receive real-time notifications via push notifications (mobile) and email.
  • The system must support at least 5,000 orders per minute during peak hours.

Non-Functional Requirements

  • Reliability: Notifications should not be lost. We must ensure 'at-least-once' delivery.
  • Latency: End-to-end latency from status update to user notification should be under 2 seconds.
  • Scalability: The system should handle sudden bursts of traffic.

Constraints

  • Assume an existing Order Service that publishes order events to a message bus.
  • The notification service must handle potential downtime of third-party notification providers (e.g., Firebase, SendGrid).

Deliverables

  1. High-level architecture diagram (described in text).
  2. Data model for storing notification preferences and message logs.
  3. Strategy for handling message retries and idempotency.
  4. Design for handling high-volume traffic bursts.

Hints

Hint 1

Use an asynchronous message queue (like Kafka or RabbitMQ) between the Order Service and the Notification Service to decouple event production from processing.

Hint 2

Implement a state machine within a database to track the lifecycle of each notification and prevent duplicate deliveries using idempotency keys.

Hint 3

Consider a 'Worker' pattern where multiple instances pull messages from a queue to scale processing horizontally based on load.

Follow-up questions

  • How would you handle a scenario where a third-party notification provider experiences a 30-minute outage?
  • How would you ensure that a customer doesn't receive the same 'Order Ready' notification twice if the network glitches?
  • What metrics would you monitor to identify bottlenecks in the notification delivery pipeline?