Overview
Design a low-latency auto-complete query suggestion service that returns the top 5 most relevant and trending search completions as users type their query character by character.
Requirements
- 1Sub-50ms latency for prefix search queries as the user types each keystroke
- 2Support 100,000 queries per second (QPS) at peak search traffic
- 3Return top 5 suggestions ranked by historical search frequency and recency
- 4Real-time and batch pipelines to update query frequencies and incorporate trending terms
- 5High availability and fault tolerance across globally distributed edge nodes
Architecture Hints (4)
Target Concepts
#Trie#In-Memory#Prefix Search#MapReduce#Edge Caching