🏗️ System DesignID:
recsys-funnel-slaRecSys Multi-Stage Funnel & 50ms SLA
推荐多阶段漏斗与 50ms SLA
🎯Core Definition
The Multi-Stage Recommendation Funnel is a classic industrial cascading architecture that refines global item candidates () down to final user recommendations within a strict 50ms Latency SLA; the standard pipeline enforces a strict latency budget across 4 stages: 1) Multi-Channel Retrieval (10ms budget, filtering down to items via ANN/collaborative filtering); 2) Pre-Ranking (5ms budget, lightweight scoring reducing to items); 3) Heavy Ranking (30ms budget, deep multi-task models predicting CTR/CVR for items); 4) Re-Ranking (5ms budget, diversity DPP, fatigue rules, yielding Top 10).
💡Use Cases
E-commerce feeds (Amazon/Taobao), short-video feeds (TikTok), visual discovery feeds (Pinterest/Xiaohongshu), and streaming media homepages (Netflix).
⚡Key Problems Solved
Heavy multi-task ranking networks cannot forward-pass millions of items within a 50ms SLA; the cascading funnel balances computational depth against candidate size at every transition, maximizing relevance under hard latency constraints.
🎯5 High-Frequency Exam Points
1
Diagram the 4-stage funnel architecture and explain why candidate sizes exhibit exponential decay: ?
2
Break down the latency budget across Network RTT, Feature Store fetching, and Model forward passes under a 50ms SLA?
3
Why does the retrieval layer optimize for Recall@K while heavy ranking optimizes for fidelity metrics like GAUC and NDCG?
4
How to implement graceful load shedding (e.g., pruning pre-ranking output from 1000 to 300) during peak traffic spikes to preserve SLA?
5
Analyze the 'retrieval ceiling dictates ranking quality' dependency and how to monitor multi-channel cross-coverage ratios?
📖 In-depth Guide:📄 recommendation-system-design →
Updated 2026-08-14🎯
🚀 Start Card Practice ➔Test Your Knowledge: Practice Questions for "RecSys Multi-Stage Funnel & 50ms SLA"
Single choice pitfall questions with instant feedback and mistake tracking.