Managed services versus self-hosting trade-offs

WHAT THIS TESTS This checks whether you evaluate total cost of ownership and operational responsibility rather than just the price tag, and whether you can match the choice to team context.

A GOOD ANSWER COVERS Managed services such as RDS or Lambda absorb undifferentiated heavy lifting: provisioning, patching, backups, replication, failover, and scaling are the provider's job. Benefits are faster delivery, less operational burden, built-in high availability, and a smaller on-call load. Costs are a higher per-unit price, less control over versions and tuning, and some vendor lock-in. Self-hosting MySQL on EC2 gives full control, any version, kernel and storage tuning, and possibly lower raw compute cost, but you own patching, HA setup, monitoring, backups, scaling, and the pager. The right call depends on team size and expertise, how much customization the workload truly needs, compliance constraints, and whether running the software is core differentiation or a distraction.

COMMON WRONG ANSWERS Comparing only the monthly invoice and concluding self-hosting is cheaper, ignoring engineer time and risk. Assuming managed always wins or always loses. Forgetting lock-in. Overstating how much custom tuning the workload actually requires.

LIKELY FOLLOW-UPS When is self-hosting genuinely justified? How do you mitigate lock-in? How do you price an engineer's time into the comparison? What about a hybrid approach?

ONE CONCRETE EXAMPLE A five-person startup runs its database on RDS Multi-AZ so nobody handles patching or failover and they ship features instead. A later workload needs a database extension and kernel tuning RDS does not allow, so for that specific case they self-host on EC2, accepting the added operational ownership in exchange for the control they genuinely need.