FAMILY A · DATA ACCESSp99 < 15 ms
Read-Heavy Systems
Workloads where reads vastly outnumber writes. Strict tail-latency requirements, high cacheability of popular items, and susceptibility to cache stampedes.
Key PatternsMulti-tier caching, Read replicas, CDN edge distribution, Materialized views, Single-flight request coalescing.
Core EnginesRedis, Memcached, Varnish, CDN edge caches, B+ Tree primary indexes.
Critical TrapQuerying database directly for hot keys without stampede protection; cache invalidation race conditions.
Product CatalogueUser Profile ServiceContent PublishingDNS Resolution
FAMILY B · INGESTIONHigh Throughput (≥ 100k ops/s)
Write-Heavy & Ingestion Systems
Massive continuous write streams dominated by append-only operations. Immediate read consistency is rarely required, but sustained write throughput and low write amplification are paramount.
Key PatternsAppend-only logs, Write-Ahead Logging (WAL), Log-Structured Merge (LSM) Trees, Memory-mapped ring buffers, Backpressure propagation.
Core EnginesRocksDB, Apache Kafka, Apache Cassandra, ClickHouse, Apache Pulsar.
Critical TrapIn-place B+ Tree updates causing random disk I/O; synchronous unbuffered write transactions.
IoT Telemetry CollectorClickstream Analytics LakeApplication Event PipelineAudit Trails
FAMILY C · TRANSACTIONALp99 < 50 ms (100% ACID)
Mixed Transactional Systems (ACID)
Workloads requiring absolute data integrity, zero drift, and strict serializability or snapshot isolation. Double-entry ledgers, payments, and inventory allocation.
Key PatternsMulti-Version Concurrency Control (MVCC), Two-Phase Locking (2PL), Distributed Sagas, Transactional Outbox, Idempotency keys.
Core EnginesPostgreSQL, CockroachDB, TiDB / TiKV, FoundationDB, MySQL with InnoDB.
Critical TrapDistributed 2PC across microservices with long holding locks; missing idempotency tokens on retry paths.
Core Banking LedgerPayment GatewayInventory ReservationAirline Booking Engine
FAMILY D · TRAFFIC SHOCKS10x - 100x Spikes in seconds
Burst-Heavy & Spike Workloads
Normal baseline operation interrupted by unpredictable, extreme demand surges that threaten to saturate CPU, memory, and database connection pools simultaneously.
Key PatternsVirtual waiting rooms, Token Bucket / Leaky Bucket admission control, Adaptive load shedding, Priority queues, Feature degradation.
Core EnginesRedis Lua token buckets, Cloudflare Waiting Room, AWS SQS FIFO with backpressure.
Critical TrapAllowing unbounded requests directly into connection pool; uncoordinated client retry storms (dogpiling).
Flash-Sale Ticket DropsViral DropsTax Filing DeadlinesEmergency Alert Broadcasts
FAMILY E · SUB-MILLISECONDp99 < 1 - 2 ms
Low-Latency & Real-Time Systems
Systems where every microsecond, CPU context switch, and network hop directly impacts business outcomes. In-memory execution, non-blocking I/O, and lock-free concurrency.
Key PatternsCPU core pinning, In-memory state, Kernel bypass (io_uring, XDP, eBPF), Lock-free queues, Request hedging, Zero-copy serialization.
Core EnginesAerospike, Redis, In-memory C++/Rust state machines, Aeron IPC.
Critical TrapUnbounded Garbage Collection pauses; blocking I/O calls inside hot loop; inter-thread mutex lock contention.
Real-Time Bidding (RTB)High-Frequency TradingFraud Scoring EngineMultiplayer Game Loops
FAMILY F · STREAMINGSub-second Event Time
Streaming & Continuous Event Processing
Unbounded event streams processed incrementally with stateful aggregations, sliding time windows, watermarks, and exactly-once semantics.
Key PatternsEvent time vs processing time, Watermark alignment, RocksDB state backends, Asynchronous Chandy-Lamport barrier checkpointing.
Core EnginesApache Flink, Kafka Streams, Apache Spark Streaming, Apache Pulsar.
Critical TrapRelying on processing time for out-of-order events; synchronous external API lookups inside stream operators.
Live Fraud DetectionContinuous Billing AggregatorDynamic Surge PricingIoT Anomaly Detector
FAMILY G · OBSERVABILITYSeconds to Ingest, Sub-second Query
Logging & Observability Platforms
High-volume unstructured and semi-structured telemetry combining immense ingestion throughput, label-based filtering, and multi-tier retention economics.
Key PatternsChunked append-only compression, Inverted posting indexes vs label-only indexes (Loki model), Tiered storage (NVMe -> S3 -> Glacier).
Core EnginesGrafana Loki, OpenSearch, ClickHouse, Apache Kafka, Jaeger, VictoriaLogs.
Critical TrapFull-text indexing of high-cardinality payloads (UUIDs, timestamps) causing index explosion.
Enterprise Log LakeDistributed Tracing CollectorSIEM Security Event Store
FAMILY H · TELEMETRYSub-second Window Rollups
Time-Series & Metrics Systems
High-frequency timestamped numeric metrics arriving at regular intervals. Requires extreme compression and continuous automated rollups.
Key PatternsDelta-of-deltas timestamp compression, Gorilla XOR float compression, Continuous downsampling (raw -> 5m -> 1h), Time-based partitioning.
Core EnginesPrometheus TSDB, VictoriaMetrics, TimescaleDB, InfluxDB, ClickHouse.
Critical TrapHigh-cardinality label explosions creating millions of concurrent open metric series.
Datacenter TelemetryIndustrial IoT Sensor FleetApplication Performance Monitoring (APM)
FAMILY I · SPATIALp99 < 20 ms Radius Searches
Geospatial & Location-Based Systems
Two-dimensional or three-dimensional coordinates subject to constant mobile updates and low-latency radius or bounding-box proximity queries.
Key PatternsHierarchical spatial tessellation (Uber H3 hexagons, Google S2 curves, Geohash), R-Trees, QuadTrees, Cell-level sharding, GPS jitter filtering.
Core EnginesRedis GEO, PostGIS, Uber H3, Elasticsearch geospatial.
Critical TrapNaive Euclidean distance scanning over raw latitude/longitude columns without spatial index.
Ride-Sharing Driver DispatchFood Delivery TrackingGeofence Breach AlertsMaritime Fleet Tracking
FAMILY J · FULL-TEXTp99 < 50 ms Ranked Searches
Search & Information Retrieval
Natural language keyword and prefix queries over massive document corpora. Requires linguistic tokenization, BM25 scoring, and distributed score merging.
Key PatternsInverted indexes, Finite State Transducers (FST) for autocomplete, BM25 relevance scoring, Segment merge policies, Distributed query fan-out.
Core EnginesApache Lucene, OpenSearch, Elasticsearch, Vespa, Meilisearch.
Critical TrapUsing SQL LIKE %keyword% queries; synchronous single-node search index updates during burst traffic.
E-Commerce Product SearchEnterprise Document SearchTypeahead AutocompleteCode Search Engine
FAMILY K · GRAPHSub-100 ms Multi-Hop Queries
Graph & Relationship-Heavy Systems
Interconnected entity networks where relationships (edges) are first-class citizens and queries traverse multiple hops rather than performing tabular joins.
Key PatternsIndex-free adjacency, Bidirectional breadth-first / depth-first search, Graph partitioning (vertex cut vs edge cut), Supernode mitigations.
Core EnginesNeo4j, AWS Neptune, NebulaGraph, JanusGraph.
Critical TrapExecuting 5-way recursive SQL JOINs in relational tables; ignoring supernodes (celebrities with 10M edges).
Social Follower NetworksFinancial Fraud Ring AnalysisKnowledge GraphsDependency Trees
FAMILY L · COLUMNARSeconds to Minutes (Multi-GB/s)
Analytical Data Warehousing (OLAP)
Analytical aggregations (SUM, AVG, GROUP BY) across billions of historical rows. Columnar data layout, compression, and vectorized execution.
Key PatternsColumnar storage (Parquet, ORC), SIMD vectorized execution, Column pruning, Predicate pushdown, Run-length encoding (RLE).
Core EnginesClickHouse, Apache Pinot, Apache Druid, Snowflake, Trino, Apache Iceberg.
Critical TrapAttempting OLAP aggregate queries against an OLTP transactional database during business hours.
Business Intelligence Data WarehouseProduct Usage Analytics LakeFinancial Reconciliation
FAMILY M · BLOB STORAGEHigh Throughput / Wire Speed
Large-Object, File & Media Systems
Large payload sizes (10 MB to 50 GB) requiring petabyte-scale storage durability, Reed-Solomon erasure coding, and byte-range streaming.
Key PatternsContent-Addressable Storage (CAS), Multipart chunked uploads, Reed-Solomon Erasure Coding (M+N), Metadata / Data separation, CDN edge caching.
Core EnginesCeph (RADOS), MinIO, Apache Ozone, HDFS, SeaweedFS.
Critical TrapStoring large binary blobs directly inside database row columns; single-stream full file uploads without chunking.
Cloud Object Storage (S3)Video-on-Demand (VOD) PipelineContainer Image Registry
FAMILY N · CONCURRENCYLocal Instant + Background Sync
Collaborative & Conflict-Tolerant Systems
Multi-master concurrent edits on shared document state without central locking. Requires mathematical convergence guarantees.
Key PatternsConflict-Free Replicated Data Types (CRDTs), Operational Transformation (OT), Join-Semilattices (⊔), State vector synchronization.
Core EnginesYjs, Automerge, Redis Pub/Sub, WebSocket gateways.
Critical TrapLast-Write-Wins (LWW) timestamp overwriting in collaborative documents causing silent character loss.
Real-Time Collaborative Doc EditingShared Whiteboard CanvasesOffline-First Mobile Sync
FAMILY O · TENANT ISOLATIONPer-Tenant SLA Enforcement
Multi-Tenant SaaS & Shared Infrastructure
Shared cloud infrastructure serving thousands of independent enterprise tenants with strict isolation, fair-share scheduling, and encryption boundaries.
Key PatternsPool vs Silo architecture, Hierarchical token bucket rate limiting, Noisy neighbor throttling, Row-Level Security (RLS), Per-tenant encryption (BYOK).
Core EnginesPostgreSQL with RLS, Envoy tenant rate limiting, Kubernetes soft multi-tenancy.
Critical TrapShared unmetered database pools allowing one runaway tenant to starve all other enterprise customers.
B2B Enterprise SaaSManaged Message Broker CloudShared Analytics Cloud
FAMILY P · HIGH THROUGHPUTMinutes to Hours Completion Window
Batch & Offline Processing Systems
Large-scale offline computations scheduled during off-peak windows or on recurring intervals. Fault-tolerant task checkpointing and speculative execution.
Key PatternsMap/Shuffle/Reduce, DAG dependency execution, Speculative task execution, Idempotent backfilling, Distributed data lineage.
Core EnginesApache Spark, Apache Airflow, dbt, Apache Hadoop.
Critical TrapNon-idempotent batch mutations that duplicate or corrupt financial ledger state on retry after failure.
Nightly Financial SettlementBulk Data MigrationsML Training PreprocessingOffline Search Indexing