Interactive System Design Playground
Architecture is learned by observing system behavior under stress. Interact with these 8 deterministic simulators to develop genuine intuition for network latency, cache evictions, consistent hashing, rate limiting, and capacity sizing.
HTTP Request, Timeout & Retry Simulator
Simulate network latency, backend processing delays, timeout boundaries, and retries. Observe how naive retries amplify system strain when a server is already degrading.
Cache Topology, Eviction & Hit-Rate Simulator
Tune cache capacity, eviction strategy (LRU vs LFU), and traffic access distributions (Zipfian 80/20 vs Uniform). See how cache hit rate directly protects database throughput.
Under real-world Zipfian traffic (where 20% of items generate 80% of reads), an in-memory cache holding just 1,000 keys deflects nearly 90% of requests away from relational disk tables.
Consistent Hashing Ring & Rebalancing Simulator
Explore the circular hash ring ($0$ to $2^32-1$). See how Virtual Nodes (vnodes) prevent severe hash skew, and how adding or removing a database node only re-maps $K/n$ keys instead of flushing the entire cluster.
Rate Limiting Algorithm & Throttling Playground
Compare how Token Bucket, Fixed Window, Sliding Window Log, and Leaky Bucket handle sudden traffic bursts. Discover why Fixed Window permits $2\times$ boundary spikes and how Token Bucket accommodates temporary bursts without starvation.
Token Bucket has 10 initial tokens. 10 requests consume all tokens immediately and succeed. The remaining 6 requests are rejected with HTTP 429 until the bucket refills at 10 tokens/sec.
Message Queue Lag & Backpressure Simulator
Tune producer publish rates against consumer group processing capacity. Observe how consumer lag accumulates, when queue buffers overflow into Dead-Letter Queues (DLQ), and how upstream backpressure protects asynchronous pipelines from catastrophic memory exhaustion.
Producers generate 800 msg/sec while 3 workers process 600 msg/sec. The queue accumulates 200 msg/sec of consumer lag. To stabilize, scale to at least 4 consumer pods or throttle upstream intake.
LRU Cache: Doubly-Linked List & Hash Map Playground
Interactive visualization of how an LRU Cache achieves guaranteed $O(1)$ key lookup and $O(1)$ node promotion. Click keys below to trigger cache reads and watch node pointers splice in real time.
Most Recent
Eviction Target
Cache initialized with 4 warm nodes. Click any key above to inspect pointer updates.
Capacity, Bandwidth & Little's Law Infrastructure Sizing
Derive infrastructure requirements from user metrics. Calculate Read/Write QPS, 5-year persistent storage, bandwidth throughput, memory caching footprint, and minimum application server instances via Little's Law ($L = \lambda \times W$).
L = λ × W → Concurrency = 11,574 req/sec × 0.050 sec = ~579 concurrent in-flight requests. At 50 concurrent requests per pod → 12 server pods required. Architecture Trade-Off & PACELC Decision Matrix
Architecture is not about memorizing components—it is about balancing trade-offs under real-world constraints. Pick a scenario below, select an architecture strategy, and evaluate its impact across Consistency, Latency, Availability, and Cost.