Cloud architecture and cost

The chapter the source document did not have, added after a coverage audit found it missing entirely (see DECISIONS.md, D16). It matters because at staff and lead level a large share of architecture questions are really cost or operational-ownership questions wearing an architecture costume: "should we run our own Kafka", "why is the bill growing faster than traffic", "do we need multi-cloud".

The unprompted signals that land well here: knowing which cost lever is largest before proposing any of them, being able to say what a managed service actually removes versus what it only moves, and treating egress and data gravity as architectural constraints rather than line items.

What this chapter covers

• [done] Managed vs self-hosted
• [done] Cloud cost architecture
• [todo] The Well-Architected pillars, and which ones actually bind
• [todo] Infrastructure as code: Terraform state, drift, and the module boundary
• [todo] Multi-cloud and portability: what it costs and when it is real
• [todo] Reserved capacity, spot and commitment planning
• [todo] Landing zones, accounts and blast-radius boundaries
• [todo] Egress, data gravity and why the network bill surprises people

Source: none. This chapter is a gap-fill rather than an expansion.