Define Performance & Scalability

Objective

Define expected load profile, caching tiers, async processing strategy, connection pooling, scaling approach, and capacity planning.

Decisions to Make

1. Expected Load Profile

Characterize the system's expected usage:
- Concurrent users: How many at peak?
- Requests/sec: Expected throughput per endpoint category
- Data volume: How much data at launch? At 1 year? At 5 years?
- Read/write ratio: What percentage of requests are reads vs writes?
- Burst patterns: Are there predictable spikes (daily, weekly, seasonal)?

2. Caching Tiers

Define caching at each level:
- Application cache: In-memory cache (Django cache framework, Redis)
- CDN: Static assets, pre-rendered pages
- DB query cache: Query result caching, materialized views
- Session cache: Session storage (DB, Redis, cookie-based)
- HTTP cache: Cache-Control headers, ETags, conditional requests

For each tier: what to cache, TTL, invalidation strategy.

3. Async Processing

• Task queues: Celery, RQ, Dramatiq — for background jobs
• Workers: How many? Auto-scaling?
• Use cases: Email sending, PDF generation, data import/export, AI inference
• Priority queues: Critical vs. best-effort tasks

4. Connection Pooling

• DB connections: Pool size, timeout, max overflow
• HTTP clients: Connection reuse for external API calls
• WebSocket connections (if applicable)

5. Scaling Strategy

• Horizontal: Add more instances behind load balancer
• Vertical: Increase instance resources (CPU, RAM)
• Auto-scaling: Triggers (CPU%, request count, queue depth)
• Database scaling: Read replicas, sharding, connection pooling

6. Capacity Planning

• Baseline resource requirements
• Growth projections (linear, exponential)
• Cost model per scaling tier
• When to re-evaluate architecture

7. Scan Skills

Query Playbook Skills where capability_domain in:
- PERF_CACHING
- PERF_ASYNC
- PERF_SCALING

Report coverage and gaps.

Deliverables

• ✅ Load profile characterized
• ✅ Caching strategy defined per tier
• ✅ Async processing approach chosen
• ✅ Connection pooling configured
• ✅ Scaling strategy defined
• ✅ Capacity plan with cost model
• ✅ Skill coverage assessed for this domain
• ✅ Decision recorded for inclusion in SAO.md (DTA-18)