Microservices vs. Monolith: The Trade You're Actually Making

Part 11 of 17 · ~1 min

Deployment and Infrastructure Overhead

Four independently deployable services means four container images, four CI pipelines, and four entries in whatever's scheduling them — which, made concrete, is what "operational overhead" actually looks like day to day: four sets of infrastructure metrics someone has to actually watch, four things, any one of which can turn up its own urgent patch on a random Friday afternoon unrelated to what the other three are doing, and a genuine question of inter-service network policy — which service is allowed to call which — that a monolith never had to answer because everything could always just call everything else.

That overhead is real, but a fair amount of it is fixed cost rather than something that scales with every new service. Furrow's small platform effort — a shared Dockerfile template, one observability stack every service ships logs and traces to the same way, a service mesh handling retries and internal traffic encryption uniformly instead of every team reimplementing it — means the fifth service, whenever Furrow adds one, costs noticeably less to bring up than the first one did. That's the actual argument for a platform investment: not that operational overhead goes away, but that its marginal cost per service keeps shrinking instead of staying flat.