25,000 Orders/min Peak
Sustained peak request volume during high-traffic events.
GeospatialStaff Level
DoorDash / UberEats 3-Sided Marketplace
Orchestrate real-time food delivery between customers, restaurants, and couriers with batched routing and live order tracking.
Target Scale
Engineering Scale & Performance SLAs
Target production parameters expected in a senior or staff interview round.
25 Million Active Diners
Daily active users generating read and write operations.
100 GB/day order and telemetry logs
Projected data ingestion and replication storage capacity.
Driver dispatch assignment < 2s
Strict end-to-end percentile latency SLA constraint.
Stage 01
Functional & Non-Functional Requirements
Establish clear problem boundaries before proposing architectural components.
Core System Capabilities
- Diners search restaurants, customize menu items, and place checkout orders.
- Restaurants receive, confirm, prepare, and mark orders ready for pickup.
- Couriers receive optimized dispatch offers based on proximity and route batching.
- Live GPS order tracking showing courier journey and real-time updated ETA.
Reliability & Latency SLAs
- Strict order transaction consistency (payment capture synchronized with restaurant POS).
- High availability for restaurant kitchen display tablets (never miss an incoming order).
- Sub-2-second dispatch assignment optimization to minimize food wait time.
Stage 02
Capacity Estimation Math
Step-by-step arithmetic conversions for QPS, storage, and bandwidth.
| Dimension | Calculation Formula | Estimated Result |
|---|---|---|
| Peak Order Placement Volume | 25,000 orders/minute / 60 seconds = 416 orders/sec during dinner peak | 416 orders/second peak |
| Active Courier GPS Telemetry Ingestion | 200,000 active couriers transmitting GPS coordinates every 4 seconds | 50,000 GPS telemetry updates/second |
| Daily Dispatch Optimization Solver Cycles | 5 Million daily orders * 4 state transitions (Dispatch, Pickup, In-transit, Delivered) | 20 Million state machine lifecycle events/day |
Stage 03
Multi-Tier Architecture & Component Topology
How requests navigate ingress gateways, application logic, caching, and persistence.
GraphQL API Gateway · Restaurant Catalog Service · Elasticsearch Geo-Index
Serve restaurant menus with dynamic delivery radius bounding, pricing tiers, and dietary filters.
Dispatch Optimizer (Linear Programming / VRP Solver) · Courier Location Cache (Redis H3) · Offer Router
Batch multiple nearby orders, evaluate courier travel times, and issue timed 30-second acceptance offers.
Order Orchestrator (Temporal / Cadence) · Restaurant Notification Bridge · Kafka Event Bus
Durable workflow managing: Placed -> Accepted -> Cooking -> Ready -> Picked Up -> Completed.
Telemetry Ingestion Gateway · WebSocket Broadcast Cluster · Map Matching Engine
Snap raw courier GPS coordinates to road networks and stream smooth car icon movement to customer phones.
Stage 04
Database Schemas & Partitioning Strategy
Entity models, indexing, and primary key partitioning.
PK: order_id
- order_id (UUID PK)
- customer_id (UUID)
- restaurant_id (UUID)
- courier_id (UUID NULL)
- status (VARCHAR(32))
- total_cents (INT)
- created_at (TIMESTAMP)
Composite index on (restaurant_id, status) for kitchen display system order queues.
PK: item_id
- item_id (UUID PK)
- order_id (UUID)
- menu_item_id (UUID)
- customizations (JSONB)
- quantity (INT)
- price_cents (INT)
Foreign key reference on order_id.
PK: courier_id
- courier_id (UUID PK)
- h3_index (VARCHAR(15))
- lat (DOUBLE)
- lng (DOUBLE)
- updated_at (TIMESTAMP)
Redis Geospatial / H3 index for instant radius lookup of available delivery drivers.
Stage 05
Critical Architectural Trade-Offs
How to defend engineering compromises when challenged by interviewers.
Courier Dispatch: Greedy Nearest Neighbor vs Batched Optimization Window
Option A: Greedy Instant Dispatch (Assign closest driver immediately)
Option B: Batched Time-Window Optimization (Aggregate orders over 30-60s intervals)
Rationale: Greedy dispatch causes couriers to make separate trips for orders from the same restaurant. Batched optimization reduces total travel distance by 30% by bundling deliveries.
Order Workflow: Ad-hoc Database Flags vs Durable Workflow Engine (Temporal)
Option A: Database status column updates with cron pollers
Option B: Durable Workflow Orchestrator (Temporal / Cadence)
Rationale: Food delivery has dozens of timeout edges (restaurant declines, driver cancels, delivery delayed). Temporal provides bulletproof state machines with built-in retry timers without dangling orders.
Technical FAQ
Frequently Asked Questions: DoorDash / UberEats 3-Sided Marketplace
Key interview questions and conceptual defenses.
How do you handle a courier rejecting a delivery offer?
If a courier rejects or lets the 30-second offer countdown expire, the dispatch engine instantly evaluates the next best candidate in the active H3 geospatial cell.
How do you keep restaurant tablet order delivery reliable when kitchen Wi-Fi is spotty?
Tablets maintain persistent bidirectional connections with automatic local SMS/phone call failover alarms if an incoming order is not acknowledged within 3 minutes.
Simulate this architecture
Download ClawPad