Hotel Booking Platform

Overview

A hotel booking platform (like Booking.com or Airbnb) connects travelers with accommodation providers. The platform lists millions of properties, supports real-time availability and pricing, handles reservations with strong consistency, processes payments, and manages reviews and recommendations. The core design challenges include maintaining real-time inventory consistency across thousands of OTA (Online Travel Agency) channels, handling concurrent booking attempts for the same room on the same dates, and providing fast geographic search with filters.

Key Requirements

Functional

• Search hotels by location, dates, guests, price, amenities, rating
• View hotel details: photos, amenities, room types, policies
• Check real-time availability and pricing for specific dates
• Make reservations (select room, enter guest details, payment)
• Payment processing (credit card, PayPal, wallets)
• Booking management (modify, cancel, refund)
• Reviews and ratings system
• Recommendations (similar hotels, frequently booked together)
• Multi-currency and multi-language support

Non-Functional

Capacity Estimation

Monthly active users: 150M
Daily searches: 10M
Daily bookings: 1M
Properties listed: 5M
Room types per property: 5 (avg)
Booking QPS: 1M / 86400 ≈ 12/sec (avg), 100/sec (peak)

Storage (properties): 5M × 10KB = ~50 GB
Storage (bookings): 1M/day × 3KB × 365 = ~1.1 TB/year
Storage (reviews): 2M/day × 1KB × 365 = ~730 GB/year

Bandwidth (search): 50K/sec × 100KB (search results with images) = ~5 GB/s

High-Level Architecture

graph TB
    subgraph "Clients"
        Web[Web App]
        Mobile[iOS/Android]
        Partner[Partner API<br/>B2B]
    end

    subgraph "Edge"
        LB[Load Balancer]
        CDN[CDN / Images]
        GW[API Gateway]
    end

    subgraph "Core Services"
        SearchSvc[Search Service]
        InventorySvc[Inventory Service]
        PricingSvc[Pricing Service]
        BookingSvc[Booking Service]
        PaymentSvc[Payment Service]
        ReviewSvc[Review Service]
        RecSvc[Recommendation Service]
        NotifSvc[Notification Service]
    end

    subgraph "Data Stores"
        PropertyDB[(Property DB<br/>PostgreSQL)]
        InventoryDB[(Inventory Store<br/>Redis + PostgreSQL)]
        BookingDB[(Booking Store<br/>PostgreSQL)]
        SearchIdx[(Search Index<br/>Elasticsearch)]
        ReviewDB[(Review Store<br/>PostgreSQL)]
    end

    subgraph "External"
        PaymentGW[Payment Gateway<br/>Stripe/Adyen]
        ChannelMgr[Channel Manager<br/>OTA sync]
    end

    subgraph "Messaging"
        Kafka[Kafka Event Bus]
    end

    Web --> LB
    Mobile --> LB
    Partner --> GW
    LB --> GW
    GW --> SearchSvc
    GW --> InventorySvc
    GW --> BookingSvc
    GW --> PaymentSvc
    SearchSvc --> SearchIdx
    InventorySvc --> InventoryDB
    InventorySvc --> ChannelMgr
    BookingSvc --> BookingDB
    BookingSvc --> InventorySvc
    BookingSvc --> PaymentSvc
    PaymentSvc --> PaymentGW
    ReviewSvc --> ReviewDB
    RecSvc --> SearchIdx
    BookingSvc --> Kafka
    Kafka --> NotifSvc

Deep Dive: Search and Availability

Hotel search is geo-based with complex filters: dates, guest count, price range, amenities, star rating, distance from landmarks.

sequenceDiagram
    participant User
    participant SearchSvc
    participant ES[Elasticsearch]
    participant Inventory[Inventory Service]
    participant Pricing[Pricing Service]

    User->>SearchSvc: Search (city, dates, 2 guests)
    SearchSvc->>ES: Geo query + filters
    ES-->>SearchSvc: Matching properties (top 200)
    SearchSvc->>Inventory: Check availability for dates
    Inventory-->>SearchSvc: Available properties
    SearchSvc->>Pricing: Get prices for available rooms
    Pricing-->>SearchSvc: Prices per room type
    SearchSvc->>SearchSvc: Rank and sort
    SearchSvc-->>User: Search results

Search ranking factors:

• Relevance (location match, filter match)
• Price competitiveness (cheaper options rank higher for price-sensitive queries)
• Review score and quantity
• Booking conversion rate (historical)
• Partner commission (sponsored listings)
• Availability and urgency (few rooms left)

Geographic search: Elasticsearch geo_point with geo_distance query. Properties within the search radius are retrieved, then filtered by date availability.

Deep Dive: Inventory Management and Double Booking Prevention

Preventing double bookings when multiple users try to book the same room on the same dates is the most critical consistency requirement.

graph TB
    Book[Booking Request] --> Check{Inventory<br/>available?}
    Check -->|No| Reject[Reject: No availability]
    Check -->|Yes| Reserve[Reserve inventory<br/>Redis DECR]
    Reserve --> Payment[Process payment]
    Payment -->|Success| Confirm[Confirm booking<br/>+ DB write]
    Payment -->|Failure| Release[Release inventory<br/>Redis INCR]
    Confirm --> Kafka[Kafka]
    Kafka --> ChannelMgr[Update Channel Manager]

Double booking prevention strategy:

1. Redis inventory cache — for each (property_id, room_type, date), maintain an available count in Redis
2. Atomic decrement — DECR is atomic; if the result is < 0, immediately INCR back and reject
3. Database lock — use SELECT ... FOR UPDATE on the inventory row in PostgreSQL for strong consistency
4. Optimistic concurrency — version-based: read inventory with version, write only if version hasn’t changed
5. Compensation — if two bookings succeed (race condition), the later one is cancelled with an apology

Channel Manager sync: Hotels may be listed on multiple OTAs (Booking.com, Expedia, Airbnb). A Channel Manager synchronizes availability and pricing across all channels in near-real-time.

Deep Dive: Booking State Machine

stateDiagram-v2
    [*] --> Searching: User searches
    Searching --> Selected: User selects room
    Selected --> PaymentPending: User submits booking
    PaymentPending --> Confirmed: Payment success
    PaymentPending --> Cancelled: Payment failed / user cancel
    Confirmed --> CheckedIn: Check-in date
    Confirmed --> Cancelled: User cancels
    Cancelled --> Refunded: Refund processed
    CheckedIn --> CheckedOut: Check-out date
    CheckedOut --> Reviewed: User leaves review
    Reviewed --> [*]
    Refunded --> [*]

Payment handling:

• Authorization at booking — hold the amount on the credit card
• Capture at check-in — charge the card when the guest arrives
• Refund on cancellation — full or partial refund based on cancellation policy
• No-show handling — charge the full amount if the guest doesn’t show

API Design

Data Model

CREATE TABLE properties (
    property_id  BIGSERIAL PRIMARY KEY,
    name         VARCHAR(300) NOT NULL,
    latitude     FLOAT NOT NULL,
    longitude    FLOAT NOT NULL,
    city         VARCHAR(100),
    country      VARCHAR(50),
    star_rating  SMALLINT,
    amenities    TEXT[],
    image_urls   TEXT[],
    created_at   TIMESTAMPTZ DEFAULT NOW()
);

CREATE TABLE room_types (
    room_id      BIGSERIAL PRIMARY KEY,
    property_id  BIGINT NOT NULL,
    name         VARCHAR(100) NOT NULL,
    max_guests   INT DEFAULT 2,
    base_price   DECIMAL(10,2) NOT NULL
);

CREATE TABLE inventory (
    property_id  BIGINT,
    room_id      BIGINT,
    date         DATE,
    total_rooms  INT NOT NULL,
    booked_rooms INT DEFAULT 0,
    price        DECIMAL(10,2),
    PRIMARY KEY (property_id, room_id, date)
);

CREATE TABLE bookings (
    booking_id   BIGSERIAL PRIMARY KEY,
    property_id  BIGINT NOT NULL,
    room_id      BIGINT NOT NULL,
    user_id      BIGINT NOT NULL,
    check_in     DATE NOT NULL,
    check_out    DATE NOT NULL,
    status       ENUM('pending','confirmed','checked_in','checked_out','cancelled') DEFAULT 'pending',
    total_price  DECIMAL(10,2) NOT NULL,
    created_at   TIMESTAMPTZ DEFAULT NOW()
);

Scalability

Trade-Offs

Interview Tips

1. Lead with the double booking problem — “The most critical challenge is preventing double bookings when multiple users compete for the same room.”
2. Explain the Redis + DB two-phase approach — Redis for fast availability checks, DB for strong consistency.
3. Discuss search — geo-distance query with complex filters (dates, guests, amenities, price).
4. Mention the Channel Manager — hotels are listed on multiple OTAs; availability must be synchronized.
5. Cover the booking state machine — pending → confirmed → checked-in → checked-out → reviewed.
6. Talk about pricing — dynamic pricing based on demand, season, and competitor rates.

Interview Questions

1. How would you prevent double bookings when multiple users try to reserve the same room simultaneously?
2. Design the hotel search system — geo queries, date availability, and ranking.
3. How would you handle inventory synchronization across multiple booking platforms (OTAs)?
4. Design the booking payment flow — authorization, capture, and refund.
5. How would you implement dynamic pricing for hotel rooms?
6. Design the review and rating system — how do you detect fake reviews?
7. How would you handle booking cancellations and refunds at scale?
8. Design a recommendation system for hotels based on user preferences and booking history.
9. How would you implement multi-currency support with real-time exchange rates?
10. Design the notification system — booking confirmation, cancellation, check-in reminders, review requests.

Key Takeaways

• Double booking prevention uses a two-phase approach: Redis atomic decrement for fast checks, PostgreSQL SELECT FOR UPDATE for strong consistency.
• Hotel search uses Elasticsearch geo-point queries with complex filters and ML-based ranking.
• The Channel Manager synchronizes availability and pricing across multiple OTAs via event-driven updates.
• Booking lifecycle is a state machine: pending → confirmed → checked-in → checked-out → reviewed.
• Payment uses authorization-at-booking, capture-at-check-in, and refund-on-cancellation patterns.

Cross-References

• Payment System — Authorization, capture, refund patterns
• Search Autocomplete — Search infrastructure
• Notification System — Booking notifications
• Distributed Lock — Inventory locking strategies

References

• Booking.com Engineering Blog: “How We Scale to 1.5M+ Room Nights Booked Daily”
• Airbnb Engineering: “Search and Pricing at Airbnb”
• Google Cloud: “Building a Hotel Booking System on Cloud Spanner”