role-architect:scalability-analysis

Scalability Analysis

When to use

• Choosing between vertical and horizontal scaling for a workload
• Designing a sharding strategy for a database that is approaching its single-node limit
• Implementing read/write splitting with replication-lag awareness
• Deciding whether CQRS is justified for a given read/write pattern divergence
• Performing back-of-envelope QPS, storage, bandwidth, or memory estimates
• Designing caching layers with appropriate invalidation strategies

Core principles

1. Scale vertically first — distributed coordination has a real cost; delay it until necessary
2. Shard key determines your fate — cardinality and query pattern alignment before anything else
3. Replication lag is not zero — design read/write splitting with explicit lag tolerance per flow
4. CQRS only when patterns diverge — premature CQRS is just complexity you'll regret
5. Estimate before architecting — back-of-envelope math prevents over- and under-engineering

Reference Files

• references/scaling-strategies-and-sharding.md — vertical vs horizontal scaling, hybrid approach, hash/range/geo-based sharding, shard key selection, read/write splitting with replication lag handling, eventual consistency patterns (compensating transactions, idempotency), and CQRS write/read model separation
• references/capacity-math-and-caching.md — QPS/storage/bandwidth/memory estimation formulas, application-level and distributed cache patterns, database query cache and materialized views, CDN and edge cache, TTL/event-based/write-through/cache-aside invalidation strategies