07 · Journal · Cloud ComputingVol. 10 · Q2 2026kleiotechnology.com

A good cloud architecture is often smaller than expected.

Many teams do not need more services. They need fewer boxes, clearer ownership, and cloud infrastructure that can still be understood before lunch.

Luke 14:28

For which of you, intending to build a tower, sitteth not down first, and counteth the cost, whether he have sufficient to finish it?

§ I — Cover concept

The context behind the article.

Journal 013
5 min
Image direction

Cloud Computing
5 min
Article

Many teams do not need more services. They need fewer boxes, clearer ownership, and cloud infrastructure that can still be understood before lunch.

Why it belongs in the journal

This entry exists to make the operating logic visible: not just the system we would build, but the constraint, tradeoff, or failure mode that forced the architecture to matter in the first place.

§ II — Article

A good cloud architecture is often smaller than expected.

Smaller is often better

The default instinct in cloud architecture is to split: more services, more containers, more independent deployments. But splitting has costs that are easy to underestimate and hard to reverse.

A well-structured monolith — or a small set of well-bounded services — is often more reliable, easier to operate, and cheaper to run than a distributed system that was split too early.

When to stay together

  • One team owns the code: splitting adds network calls without organizational benefit
  • Data is tightly coupled: if two services need a shared transaction, they probably should not be separate services
  • The team is small: distributed systems require distributed expertise

The modular monolith

The best of both worlds is often a modular monolith: a single deployable unit with clear internal boundaries. Clean code boundaries without network overhead. Simple deployment without orchestration complexity.


Do not split your architecture because you might need to scale. Split it because you have proven that you need to scale different parts independently.

§ III — Reading note

What the article is really about.

Operating tension

Many teams do not need more services. They need fewer boxes, clearer ownership, and cloud infrastructure that can still be understood before lunch. In practice, the hard part is usually not implementation syntax but aligning delivery, controls, and operator trust so the thing can survive contact with a real team.

Kleio view

We treat these articles as public design memos: short, opinionated, and anchored in systems that have to be bought, operated, and defended long after launch week.

§ III — Continue reading

Three adjacent articles.

Season