sys-d — Interactive Distributed Systems Visualizer

Rate Limiting

Fixed Window Counter

Divide time into fixed windows with a counter per window. Simple but has the boundary burst problem.

Sliding Window Log

Store every request timestamp and filter old ones. 100% accurate but memory intensive.

Sliding Window Counter

Hybrid approach: uses weighted counters from current + previous windows. Best of both worlds.

Token Bucket

Tokens refill at a fixed rate. Each request consumes one. Allows controlled bursting up to bucket capacity.

Leaky Bucket

Requests fill a bucket that drains at a constant rate. Smooths bursty traffic into a steady flow.

Resilience

Circuit Breaker

Prevent cascading failures by switching between CLOSED, OPEN, and HALF_OPEN states based on failure thresholds.

Retry with Backoff

Automatically retry failed operations with exponentially increasing delays and optional jitter to avoid thundering herds.

Timeout Wrapper

Wrap operations with a timeout to prevent them from hanging indefinitely. Races the operation against a timer.

Bulkhead

Isolate concurrent workloads with execution slots and a bounded queue — prevent one overwhelmed resource from starving others.

Hedged Requests

Fire duplicate requests after a delay and use whichever responds first — tame tail latency by racing redundant calls.

Fallback Strategy

Chain alternative providers to gracefully degrade — try primary, then secondary, cache, and finally a static default.

Caching

LRU Cache

Evict the least recently used item when capacity is reached. O(1) get/put with a hash map + doubly linked list.

LFU Cache

Evict the least frequently used item. Tracks access counts to keep the hottest data in memory.

TTL Cache

Entries expire after a time-to-live. Ensures stale data is automatically purged without manual invalidation.

Cache Aside

Application checks cache first, loads from DB on miss, then populates cache. Most common caching strategy.

Write-Through/Back

Write-through writes to cache and DB synchronously. Write-back buffers writes for eventual persistence.

Cache Stampede

Prevent thundering herd when a popular key expires. Use locking or probabilistic early expiration.

Bloom Filter

Space-efficient probabilistic structure to test set membership. May have false positives but never false negatives.

Load Balancing

Round Robin

Distribute requests sequentially across servers in a circular order. Simple and fair for uniform workloads.

Weighted Round Robin

Assign weights to servers based on capacity. Higher-weight servers receive proportionally more traffic.

Least Connections

Route to the server with the fewest active connections. Adapts to varying request durations.

IP Hashing

Hash client IP to deterministically select a server. Ensures session affinity without shared state.

Consistent Hashing

Map keys to a ring of nodes. Adding/removing nodes only redistributes a fraction of keys.

Messaging

In-Memory Queue

FIFO queue for decoupling producers and consumers. Simple async communication within a single process.

Pub/Sub Broker

Publishers emit events to topics, subscribers receive matching messages. Decouples producers from consumers.

At-Most-Once

Fire and forget delivery. Messages may be lost but are never duplicated. Lowest latency guarantee.

At-Least-Once

Retry until acknowledged. Messages are never lost but may be delivered multiple times.

Dead Letter Queue

Route failed messages to a separate queue for inspection. Prevents poison messages from blocking processing.

Consumer Groups

Distribute partitions across consumers in a group. Enables parallel processing with load balancing.

Partitioned Log

Kafka-style append-only log split into partitions. Ordered within partition, parallel across partitions.

Coordination

Distributed Lock

Ensure mutual exclusion across distributed nodes. Only one process can hold the lock at a time.

Leader Election

Elect a single leader among distributed nodes to coordinate work. Handles failover automatically.

Heartbeat Monitor

Periodic health signals detect node failures. Missing heartbeats trigger failover or deregistration.

Service Registry

Central catalog where services register their endpoints. Enables dynamic discovery of available instances.

Service Discovery

Clients query the registry to find healthy service instances. Supports client-side or server-side patterns.

Consistency

Primary-Replica

One primary handles writes and replicates to read replicas. Scales reads but introduces replication lag.

Read/Write Quorum

Require W writes and R reads where W+R > N for strong consistency. Tunable consistency-availability trade-off.

Vector Clocks

Track causal ordering of events across nodes. Detect concurrent updates and resolve conflicts.

CRDT Counters

Conflict-free replicated data types that converge without coordination. Increment/decrement across replicas.

Consensus

Two Phase Commit

Coordinator asks all participants to prepare, then commit. Atomic but blocking if coordinator fails.

Three Phase Commit

Adds a pre-commit phase to reduce blocking. Participants can safely abort if coordinator is unreachable.

Simplified Raft

Leader-based consensus: elect a leader, replicate logs, commit when majority acknowledges. Understandable by design.

Storage

Key-Value Store

Simple get/put/delete interface backed by a hash map. Foundation of most distributed storage systems.

Append-Only Log

Immutable sequential writes. Fast writes, natural audit trail, basis for event sourcing and WAL.

Write Ahead Log

Log mutations before applying them. Enables crash recovery by replaying the log after restart.

SSTable

Sorted String Table: immutable, sorted key-value file with an index for fast lookups.

LSM Tree

Log-Structured Merge Tree: buffer writes in memory, flush to sorted files, merge in background.

Observability

Metrics Collector

Collect counters, gauges, and rates from your services. Aggregate and expose for monitoring dashboards.

Histogram

Track value distributions with configurable buckets. Compute percentiles (p50, p95, p99) from observations.

Structured Logger

Emit JSON log lines with consistent fields. Enables machine-parseable, searchable, and filterable logs.

Trace ID Propagation

Propagate a unique trace ID across service boundaries. Correlate logs and spans for distributed tracing.

sys-d Visualizer