Estimate System Capacity & Scale Requirements

A reference document of scale-benchmark numbers and estimation formulas for system-design capacity planning, named MONOPOLY. It covers quick estimation formulas for converting DAU to average and peak RPS, with peak multipliers by app type - 5-10x for social media, 10-20x for news/media breaking-news spikes, 2-3x for B2B SaaS business-hours spikes - storage estimation (requests per day times payload size times 365, times a typical 3x replication factor, reduced by cache-hit ratio) with common payload sizes from a 500B tweet to 150MB per minute for 1080p video, and bandwidth estimation with the Gbps-to-MB/s conversion. It tabulates known single-node scale limits for seven databases - PostgreSQL at roughly 5-20K writes/s, Cassandra at 200K-1M writes/s, DynamoDB effectively unlimited under managed provisioning - with recommended sharding or clustering triggers, five queue and stream technologies such as Kafka at 1M+ msgs/s per cluster and SQS FIFO capped at 3K msgs/s per queue with their retention windows, and three caching technologies including Redis at roughly 1TB RAM and 1M ops/s at sub-millisecond latency. It gives full capacity-planning profiles at five DAU tiers - 1K, 10K, 100K, 1M, and 10M - each with average and peak RPS ranges, yearly database size, the concrete infrastructure stack needed from a single RDS t3.medium up to multi-region microservices with a distributed database and dedicated SRE, and a monthly cost band from $50-200 up to $200K-2M+. It also tabulates the five standard SLO and availability tiers, 99% through 99.999%, with their allowed monthly downtime, noting that four-nines availability specifically requires multi-AZ deployment, automated failover, zero-downtime deploys, chaos engineering, and 24/7 on-call, plus latency-budget guidelines for perceived user latency, network latency by geographic distance, and database query time targets by query type.

Use it when estimating request rates, storage, bandwidth, database, queue, or cache technology limits, infrastructure sizing and cost, SLO targets, or latency budgets for a system-design or capacity-planning task at a given user scale. It is a reference document of typical, approximate figures, not a substitute for verifying the actual architecture, load-testing results, or vendor-specific limits before a production decision.

Input is a scale parameter - DAU, request volume, payload size, or a target SLO or latency tier. Output is an estimated RPS, storage, bandwidth, or cost figure derived from the documented formulas, for example:

Requests per second (avg) = DAU × avg_requests_per_user_per_day / 86400
Requests per second (peak) = avg_RPS × peak_multiplier

References named technologies across three categories: databases (PostgreSQL, MySQL, MongoDB, Cassandra, DynamoDB, Redis, Elasticsearch), queues and streams (Kafka, RabbitMQ, SQS Standard/FIFO, Redis Pub/Sub), and caches (Redis, Memcached, Caffeine/Guava), plus AWS instance sizing examples like RDS t3.medium and r5.xlarge.

System designers and engineers doing back-of-envelope capacity planning or interview-style system-design estimation who need concrete formulas and known technology limits rather than vague rules of thumb.