Monolithic vs Polyrepo Pipelines

One pipeline or many.

Monolithic pipeline pattern

Single CI pipeline runs every test, every check, every build for the whole repo.

Simple to understand; brutal at scale. A 1000-service monorepo cannot run every test on every commit.

Works up to ~20 services or modest test counts.

Polyrepo pipelines

Each service has its own pipeline. Independent CI for independent code.

Scales naturally; each pipeline is fast.

Cost: drift in tooling, duplicated config, harder cross-service refactors.

Selective pipelines (the modern monorepo pattern)

In a monorepo, only run tests for affected services. Bazel, Nx, Turborepo, Pants compute the affected set from the change set.

Best of both worlds: monorepo simplicity, scaled CI cost.

Requires good dependency tracking. Without it, the affected set is wrong and tests are skipped that shouldn't be.

Scaling each pattern

Monolithic: parallelize within the pipeline (matrix builds).

Polyrepo: shared CI templates, internal tooling team to keep them in sync.

Selective: invest in build graph correctness; the entire pattern depends on it.

How to pick

Single repo, under 20 services: monolithic pipeline.

Multiple repos: polyrepo with shared templates.

Monorepo above 20 services: selective pipelines with Bazel, Nx, or similar.