Select optimal tech stack components for your architecture

MONOPOLY - Technology Decision Matrix is a reference skill for choosing infrastructure technology across ten categories: database, cache, message queue and event streaming, API protocol, search engine, object storage, container orchestration, load balancer, observability stack, and CDN, each with a comparison table, a default recommendation, and a decision framework.

Use this as a reference when choosing a technology for a new system component or evaluating whether an existing choice still fits, rather than researching each category from scratch. It's explicitly a reference document that may not cover every edge case, so a chosen architecture still needs verification before production.

Database selection splits into relational (PostgreSQL as the default for complex queries and JSONB, MySQL/MariaDB for read-heavy legacy apps, CockroachDB for global distributed SQL, PlanetScale for serverless MySQL-compatible branching, Aurora for AWS-native) and NoSQL (MongoDB for flexible schemas, DynamoDB for massive-scale key-value, Cassandra for write-heavy time-series at companies like Apple and Netflix, Redis for cache and sessions, Elasticsearch for search, InfluxDB for time-series metrics, Neo4j for graph data), plus a decision tree by access pattern. Cache defaults to Redis over Memcached unless the workload is CPU-bound multi-threaded caching with only string values. Messaging picks Kafka for replay and audit trails and high-throughput pipelines, RabbitMQ or SQS for simple task queues, Redis Pub/Sub or NATS for ephemeral real-time push, and SNS or Kafka consumer groups for fan-out. API protocol defaults to REST for public APIs, gRPC for internal service-to-service, and WebSocket or SSE for real-time features. Search recommends staying on PostgreSQL full-text search under roughly 1M documents, then moving to Typesense or Elasticsearch above that. Object storage highlights Cloudflare R2 for zero-egress-cost media delivery versus S3's roughly $0.09/GB egress. Container orchestration defaults startups to ECS+Fargate and moves to EKS or GKE once a team passes roughly 5 engineers or 10 services. Load balancers are compared by layer - ALB and GLB at L7, NLB at L4, Nginx, HAProxy, and Traefik self-hosted, Cloudflare bundling DDoS protection with load balancing.

Observability is broken into metrics (Prometheus+Grafana, Datadog, CloudWatch, New Relic), logging (ELK, Loki+Grafana, Splunk, CloudWatch Logs, Datadog Logs), and tracing (Jaeger, Zipkin, AWS X-Ray, Datadog APM, Honeycomb), with a recommended open-source stack of Prometheus, Grafana, Loki, and Jaeger integrated via OpenTelemetry, or Datadog/Grafana Cloud as a managed alternative. CDN compares Cloudflare (300+ edge locations, best free tier), CloudFront (450+, deep AWS integration), Akamai (4000+, enterprise-priced), Fastly, and Vercel/Netlify edge, defaulting to Cloudflare for most use cases. A closing scale-benchmarks table gives approximate write and read throughput per technology - for example Cassandra at roughly 1M writes/s on a 10-node cluster, Kafka at roughly 1M msgs/s per partition, and PostgreSQL at roughly 10K writes/s with connection pooling - noting all figures are approximate and hardware-dependent.

Engineers and architects choosing infrastructure technology for a new system or component who want a fast, comparative reference with concrete defaults instead of researching every option from scratch.