System Design: Designing a Food Delivery App

A food delivery platform (like Uber Eats, DoorDash, or Grab) is more than just an e-commerce site. It is a Three-Sided Marketplace that must coordinate between Customers, Restaurants, and Couriers (Drivers) in real-time, all while managing a complex logistical window (the food is only hot for 30 minutes).

1. Core Requirements

• Order Management: Taking orders and managing the lifecycle (Created -> Accepted -> Prepared -> Picked Up -> Delivered).
• Inventory/Menu: Restaurants updating their menus and stock in real-time.
• Courier Dispatching: Efficiently matching an order with the best courier.
• Tracking: Customers seeing the courier's location in real-time.

2. The Order State Machine

The state of an order is the source of truth for all three parties. We use a State Machine to ensure valid transitions:

• PLACED -> ACCEPTED_BY_RESTAURANT -> PREPARING -> READY_FOR_PICKUP -> OUT_FOR_DELIVERY -> DELIVERED.
• Consistency: Use a relational database (PostgreSQL) for order state to ensure ACID compliance during transitions.

3. Courier Dispatching (The Core Challenge)

Matching an order with a courier is an optimization problem.

• The Process:
  1. An order is READY_FOR_PICKUP.
  2. The system identifies available couriers within a 3-5km radius using Geohashing (like in our Yelp/Uber articles).
  3. The Dispatch Service ranks couriers based on distance, vehicle type, and historical performance.
  4. It sends a "Gig Request" to the top-ranked courier. If they reject, it moves to the next.

4. Real-time Location Tracking

• Ingestion: Couriers send GPS pings every 5 seconds via WebSockets or gRPC.
• Processing: Apache Kafka buffers these pings.
• Storage: Only the "Latest Location" is stored in Redis for fast access by the tracking UI. Historical paths are stored in Cassandra.

5. Handling Flash Sales and Peak Hours (Lunch/Dinner)

Traffic spikes during meal times are massive.

• Adaptive Throttling: If a restaurant is overwhelmed, the system automatically increases their "Estimated Prep Time" or hides them from the search results to prevent a bad user experience.
• Batching: To improve efficiency, the system may assign two orders from the same restaurant to a single courier if the destinations are close.

6. Menu Management and Search

• Search: Use Elasticsearch to handle fuzzy search ("Burgir" -> "Burger") and filtering by category/rating.
• Cache: Restaurant menus are cached in Redis with a short TTL, as they don't change every second but need to be retrieved millions of times.

Summary

The engineering of food delivery is a logistical dance. By treating the order as a strict state machine and using geospatial indexing for courier dispatching, you can build a system that scales to millions of orders across thousands of cities.