System design interview question
Design a Distributed Graph Query Engine for a Social Knowledge Network (Like LinkedIn's Economic Graph at Scale)
Design a system that powers relationship and entity queries across a global social knowledge network serving 500M users with 50B relationships (follows, connections, endorsements, skill associations). The system must support low-latency (sub-200ms p99) graph traversal queries (e.g., "find all second-degree connections who work at company X and have skill Y"), handle 100K queries per second with geographic distribution across 3 continents, support real-time relationship mutations (5K writes/sec), maintain eventual consistency across replicas, and enable efficient caching of hot subgraphs while managing memory constraints. Address how you'd partition the graph, cache frequently accessed nodes and edges, handle write propagation, prevent cascade failures during graph explosions, and support ad-hoc queries without pre-computation.
Practice this question step by step with an AI interviewer that reviews your requirements, your architecture and your trade-offs, and scores the design.