An interactive study companion for Martin Kleppmann's classic on reliable, scalable, and maintainable systems.
Virality shifts traffic from long-tail to hot objects. Design caching and storage tiers for both.
A video with 10 views and one with 10M views share the same architecture but wildly different cache hit ratios. Popularity-aware prefetch and tiered storage prevent paying CDN prices for cold content.
Netflix Open Connect caches hot titles on ISP appliances. YouTube's long tail lives in cheaper storage classes; only trending manifests get aggressive edge warming.
Search ranking and YouTube recommendations share DNA: offline batch features + online low-latency scoring. The serving path never scans raw watch history per request.
// Prevent thundering herd on trending metadata
const inflight = new Map<string, Promise<TrendingPage>>();
async function getTrending(region: string): Promise<TrendingPage> {
const cached = await redis.get(`trending:${region}`);
if (cached) return JSON.parse(cached);
if (!inflight.has(region)) {
inflight.set(region, loadTrendingFromDb(region).finally(() => inflight.delete(region)));
}
return inflight.get(region)!;
}