Each system below is a real architecture you can run, not just read. Hit play and watch a request flow through the components step by step, click any box to see why it's there, trigger a failure to see what breaks, then get quizzed on what you learned.
The smallest real system on Reqflow. Hit play, follow the request through cache and DB, then click every box.
Fan-out on write vs read, ranking, CDN.
Persistent connections, message queues, delivery.
Room-based pub/sub fanout across gateway instances.
Geo-indexing, dispatch, real-time location streams.
CDN-first, adaptive bitrate, recommendation.
Idempotency, fraud, async webhooks, ledger.
Chunking, content-addressed dedup, multi-device sync.
Multi-channel async delivery, user preferences, retries.
Video pipeline, ML-driven recommendation, infinite scroll.
Web crawl, inverted index, scatter-gather query, ranking.
Geohashing, nearby search, review writes, hot-query caching.
Seat holds, payment timeouts, virtual queue for hot events.
Hybrid fan-out: push for most users, pull for celebrities, merged at read.
Partitioned, replicated log. Brokers, ISR, consumer groups, leader failover.
Multi-AZ erasure coding, sharded metadata, strong read-after-write.
Operational Transforms, per-document op log, snapshot + delta replay.
Single-threaded per-symbol matching, in-memory order book, microsecond latency.
Geo-search, atomic seat-holds, authorize-then-capture payments, async fan-out.
Federated social: per-user PDS, global firehose Relay, plug-in AppViews and custom feeds.
64-bit time-sortable IDs: timestamp + worker id + sequence, minted locally with no per-request coordination.
Redis sorted sets for O(log n) rank queries; durable points in MySQL; top-N, your rank, players around you.
URL frontier, politeness delays, Bloom-filter dedup, distributed fetching at billions of pages.
Precomputed trie cache in Redis, debounced prefix queries, batch + streaming frequency pipeline.
Cron at scale: leader election, CAS dispatch, at-least-once delivery, exactly-once via dedup, catch-up execution.
Leaderless replication, vector clocks, sloppy quorums, hinted handoff, the architecture behind DynamoDB and Cassandra.
Time-series ingestion at millions of points/sec, Prometheus scraping, TSDB columnar storage, alerting pipeline.
Exactly-once stream aggregation, Lambda vs Kappa architecture, time-windowed counts, fraud detection.
Cross-wallet transfers, double-entry ledger, saga compensations, event sourcing + CQRS.