RETURN TO SYSTEM DESIGN VAULT (55)BLUEPRINT #22 // Easy
#22EasySearch, Autocomplete & Big Data
Top K Elements: App Store Rankings, Amazon Bestsellers, etc.
ASKED IN INTERVIEWS AT:AmazonBloombergMetaPinterest
EXECUTIVE SUMMARY & PROBLEM STATEMENT
Design a real-time leaderboard service displaying the top 100 best-selling items or trending apps updated continuously across millions of purchases.
SCALE ESTIMATION & TRAFFIC ENVELOPE
DAILY ACTIVE USERS / TPS100,000 purchases/second peak
READ / WRITE TRAFFIC RATIO1:10 read-to-write
5-YEAR STORAGE CAPACITYRedis ZSET with 1M products ~ 100 MB RAM
FUNCTIONAL REQUIREMENTS
- [1]Increment score when purchase occurs
- [2]Query top K items by category or platform in real time
NON-FUNCTIONAL REQUIREMENTS
- •Query response latency < 20ms
- •Support high write throughput without Redis CPU saturation
CORE ARCHITECTURAL COMPONENTS
Purchase Event Bus
Write Aggregation Buffer
Redis Master-Replica ZSET
Top-K Cache Snapshot Worker
DATA SCHEMA & PERSISTENCE
Redis ZSET `leaderboard:{category}:{date}` score: sales_count, member: product_id
DISTRIBUTED TRADE-OFFS & FAILURE MODES
- ⟡Batching score increments in application memory for 1 second reduces Redis commands by 90% at the cost of 1s latency.