syswork

$ syswork --explain

See how systems actually work. Then break them.

Animated simulations of real system internals: message brokers, load balancers, replication, transactions. Stall a consumer, cut a network link, crash a node, or pull a component out entirely. Then watch, step by step, what the system does and why.

Redis

In-memory data structure servertechnology

Kafka

Distributed, partitioned commit logtechnology

Produce & replicate

planned

Partitioner, leader and followers pulling (fetching) from the leader, ISR, high watermark, acks=0/1/all, min.insync.replicas, leader failover.

pull deliveryreplicationdurability

Consumer groups

planned

Pull-based consumers, lag, offset commits and rebalancing.

pull deliveryat-least-oncebackpressure

Load balancing

Spreading requests across serversconcept

Algorithms side by side

planned

Round robin, weighted, least connections, power of two choices, under uneven servers and slow requests.

load balancingtail latency

Consistent hashing

planned

The hash ring, virtual nodes, and how many keys move when a node joins or leaves.

partitioningload balancing

Consensus & replication

How replicas agree on a leader and a logconcept

Raft leader election

planned

Terms, randomized election timeouts, votes, split votes, log replication and commit index under partitions.

consensusreplication

MongoDB replica set

planned

Primary election, oplog replication, write concern majority, and rollback after a failover.

consensusreplicationdurability

Transactions

ACID, isolation levels, MVCCconcept

Isolation levels & anomalies

planned

Two transactions on parallel timelines: dirty reads, non-repeatable reads, lost updates and write skew at each isolation level.

ACIDMVCCisolation

WAL & crash recovery

planned

Why a committed transaction survives a crash: write-ahead log, fsync, checkpoints and replay.

ACIDdurability