Infrastructure & Operations › Cloud Computing

Managed vs Self-Hosted

Paying for convenience vs running it yourself.

Also known as: managed vs self-hosted, managed services, build vs buy

Managed means a provider runs a piece of infrastructure for you — a database, a queue, a Kubernetes control plane. Self-hosted means you run it yourself, on your own servers or compute. The decision is a make-or-buy trade: managed services trade money for reduced operational work; self-hosting trades work for control and, sometimes, lower marginal cost.

What you’re really comparing:

• Operational load. A managed database handles backups, patching, failover and scaling. Self-hosting means you own all of that, including the 3am pages (see toil).
• Control. Self-hosting lets you tune configuration, install extensions, and pick exact versions. Managed services expose a subset of settings and upgrade on their schedule.
• Cost. Managed usually has a premium; self-hosted can be cheaper per unit if you correctly count the engineering time to run it.
• Integration. Managed services slot into a provider’s networking, identity and monitoring — convenient, but part of why moving later is work (see vendor lock-in).

managed:  pay more, operate less, less control
self:     pay (in time) more, operate more, more control

The classic mistakes:

• Self-hosting to save money without counting ops. The license or instance may be cheaper, but the engineer-hours to run, patch, monitor and fix it usually aren’t. Self-hosting is often more expensive once labor is priced in.
• Choosing managed for something you must tune deeply. If you need a specific engine version, extension or low-level setting that the managed offering doesn’t expose, you’ll fight it. Self-host instead.
• Assuming managed means no operations. You still configure, monitor, back up at the application level, and pay. The boundary shifts; it doesn’t disappear (see shared responsibility).
• Not planning for exit. Even a managed choice should have an idea of how you’d leave, for cost or resilience reasons.

How to decide: default to managed for undifferentiated infrastructure — databases, queues, orchestration — where the provider genuinely removes work. Self-host when control, cost at scale, compliance, or a highly specific need justifies the operational burden. Frontend and data teams make the same call for their storage and processing; the reasoning is the same.