ACID Isolation Locking Lab
Simulate a checkout race to see how ACID isolation, row locks, table locks, MVCC snapshots, range locks, and serializable retries keep database state correct.
Browse all browser-based simulators for failure modes, production behavior, reliability, scaling, and backend system design.
Simulate a checkout race to see how ACID isolation, row locks, table locks, MVCC snapshots, range locks, and serializable retries keep database state correct.
Simulate Kubernetes Horizontal Pod Autoscaler scaling pod replicas based on CPU utilization. Watch pods spin up and down as traffic changes, experiment with scaling thresholds
Test Kubernetes HPA autoscaling as explicit runtime components with parent-owned input ports, child injection, routed connector messages, component state, element state, controls
Test database B-Tree indexing as explicit runtime components with parent-owned input ports, child injection, routed connector messages, buffer pool caching, and query planning.
Simulate different cache invalidation strategies side-by-side. Watch how write-through, write-behind, cache-aside, and TTL-based invalidation handle reads, writes
Analyze cascading failure propagation across microservices when latency or errors are injected, and test mitigation strategies using circuit breakers and retry limits.
Simulate a CDN with multiple edge PoPs, an origin shield, and an origin server. Watch cache hits, misses, SWR revalidations
Simulate the Circuit Breaker pattern with count-based and time-based sliding windows, slow-call rate thresholds, fallback components, and per-component details. Watch requests flow, timeout
Simulate a database connection pool managing connections between application threads and a database. Watch connections get checked out, used, and returned. Experiment with pool sizing
Simulate feature flag rollout strategies. Watch how percentage-based rollouts, user targeting, and kill switches control which users see new features.
Simulate Horizontal Pod Autoscaler (HPA) and Cluster Autoscaler (CA) behavior under traffic spikes. Experiment with metrics collection lag, pod cold start delays (image pull and app boot)
Experiment with HA failover, stale DNS/client caches, backend health checks, fail-open routing, overload, latency, and recovery when a load balancer crashes.
Experiment with token bucket rate limiting by changing request rate, bucket capacity, refill rate, and scenarios such as normal traffic, bursts, overload, and recovery.
Build a typed software architecture, change realistic capacity settings, and watch overload, backpressure, and Kubernetes recovery propagate through the system.
Simulate what happens when Redis crashes: cache timeouts, database fallback, cache stampede pressure, rising latency, 5xx errors, and recovery after Redis restarts cold.
Explore geospatial indexing (Geohash vs QuadTree), real-time driver coordinates streaming, surge pricing dynamic feedback loops, and transactional trip matching states.
Simulate SLOs and error budgets to see how traffic, failures, burn rates, alert thresholds, and feature freezes affect reliability decisions.
Explore LMAX Disruptor sequencing, pre-trade risk validations, FIFO Limit Order Book matching, UDP multicast publishers, and market circuit breakers.
Walk through a Redis crash from healthy reads to degraded fallback, DB protection, circuit breaking, L1 hot keys, replica failover, gradual warm-up, and cache-optional correctness.
Simulate TCP congestion control mechanisms. Watch packets flow and observe Slow Start, Congestion Avoidance, and Fast Retransmit phases. Experiment with packet loss, latency
Simulate tunable read/write consistency levels (One, Quorum, All) in a Dynamo-style replicated database ring. Slice network connections to create partitions, execute operations