Ten interactive simulations. Tune the dials, trigger failures, and watch the trade-offs play out in real time. No setup, no terminal — just you and the system.
Distribute traffic across a pool of servers. Round-robin, least-connections, and random — watch how each algorithm handles a failed node and a traffic spike.
Absorb reads before they hit the database. Tune hit rate, TTL, and availability — and watch what happens when the cache disappears.
Decouple producers from consumers. Find the breaking point where a stable queue tips into an overflowing backlog.
Survive a primary failure. See how read/write split and replication lag affect stale reads, availability, and failover.
Protect a service from being overwhelmed. Compare token-bucket, leaky-bucket, and fixed-window under a real traffic spike.
Tune edge hit rate and see origin load, latency, and regional fan-out move together.
Watch retries amplify an outage, then open the breaker and fail fast before the cascade grows.
Add and remove nodes from a ring and estimate how many keys need to move.
Turn traffic, peak multipliers, and payload size into QPS and bandwidth estimates.
Increase shard count and introduce key skew to see why hot partitions become bottlenecks.
Each lab is a deterministic, client-side simulation — no backend, no API calls. Move the sliders, tap servers, fail primaries. Every change updates the live visualization within a second. Use the "Try this" prompt at the bottom of each lab to walk through a specific scenario, then read the "Key takeaway" for the mental model that ties it together.