ulearn/systems

System architectures, taken apart live

Every topic below is a running simulation, not a page of prose. Pick one, switch its behaviour, break it on purpose, and watch what actually happens.

Traffic & Routingavailable

Load Balancer

Round robin, least connections, weighted, IP-hash, URL-hash and random routing — watch requests get distributed and rerouted live as backends fail.

open simulation →
Traffic & Routingavailable

Rate Limiter

Fixed window, sliding window, token bucket and leaky bucket throttling — hammer one client, then flood from thousands of spoofed IPs and watch per-client limits stop helping.

open simulation →
Async & Messagingavailable

Message Queue

Work queue vs. fan-out pub/sub, backpressure policies, and redelivery — kill a consumer mid-job and watch the visibility timeout, retries and dead-letter queue take over.

open simulation →
Async & Messagingavailable

RabbitMQ vs Kafka

Smart broker vs dumb log — watch the same event get pushed and deleted on one side while it sits in a log getting pulled by independent consumer groups on the other.

open simulation →
Compute & Concurrencyplanned

Concurrency vs Parallelism

Dealing with a lot of things at once vs. doing a lot of things at once — switch from Sequential to Parallel and watch the wall clock actually drop instead of just interleaving.

Resilienceplanned

Circuit Breaker

Open, half-open and closed states protecting a failing dependency.

Data & Cachingplanned

Consistent Hashing

Why adding one node shouldn't reshuffle your entire cache.

Consistency & Consensusplanned

Raft Consensus

Leader election, heartbeats and log replication — kill the leader mid-demo and watch the cluster vote in a new one before a single write is lost.

AI & Machine Learningplanned

Retrieval-Augmented Generation

Embed a question, pull the nearest chunks from a vector store, and stuff them into the prompt — turn retrieval off and watch the same model start hallucinating an answer instead.