Đang mở hậu trường...
Chọn đáp án để xem tỉ lệ cộng đồng và giải thích. Câu hỏi được chọn tự động — mọi người cùng làm 1 câu mỗi ngày.
A product-listing API backed by Redis serves 80k req/s across 200 app instances. A single "homepage_feed" key is read by nearly every request and is refreshed with a 60s TTL. During a traffic spike, each time that key expires the origin database sees a burst of ~15k concurrent identical queries, latency spikes, and some instances time out and retry — amplifying the burst further. You must fix this without weakening freshness guarantees much. Which approach best addresses the failure mode at this scale?