RETURN TO SYSTEM DESIGN VAULT (55)BLUEPRINT #49 // Easy
#49EasyReal-Time & Stream Processing
Design a Feature to Show the Number of Users Viewing a Page
ASKED IN INTERVIEWS AT:Booking.com
EXECUTIVE SUMMARY & PROBLEM STATEMENT
Design Booking.com's live viewer counter ('14 people are currently looking at this hotel') using heartbeat beacons and approximate cardinality counters.
SCALE ESTIMATION & TRAFFIC ENVELOPE
DAILY ACTIVE USERS / TPS20 Million concurrent browsing sessions, heartbeats every 20 seconds
READ / WRITE TRAFFIC RATIO1:1 write heartbeat to read count
5-YEAR STORAGE CAPACITYRedis HyperLogLog keys: 1.5 KB per hotel listing * 1M hotels ~ 1.5 GB RAM
FUNCTIONAL REQUIREMENTS
- [1]Display real-time count of active users currently viewing a hotel listing
- [2]Keep count accurate as users navigate away or close tabs
NON-FUNCTIONAL REQUIREMENTS
- •Minimal client battery/network impact
- •Response latency < 100ms
- •Tolerate occasional missed heartbeats gracefully
CORE ARCHITECTURAL COMPONENTS
Client Navigator.sendBeacon() hook
Lightweight Edge Heartbeat Collector
Redis HyperLogLog Sliding Window
Viewer Count Query API
DATA SCHEMA & PERSISTENCE
Redis Key: `viewers:{hotel_id}:{minute}` -> HyperLogLog or Sorted Set
DISTRIBUTED TRADE-OFFS & FAILURE MODES
- ⟡HyperLogLog provides 98.5% accurate viewer counts with constant 1.5 KB RAM per page, ideal for social proof urgency indicators.