Your team runs a large self-managed streaming cluster. Two engineers hold most of the operational knowledge. A managed alternative exists. How do you decide?

What the interviewer is testing

Whether you can evaluate this on differentiation, capability and exit rather than on cost alone.

The three questions

Is operating this a differentiator? If your product is not messaging infrastructure, running it yourself is a cost centre unless scale or specificity makes the managed option unworkable. Engineering time spent on cluster operations is time not spent on the product.

Do you have the expertise, and will you in three years? Two engineers holding the knowledge is key-person risk, not a capability. If either leaves, the organisation has a critical system it cannot operate — and that risk is present today, unfunded and unrecorded.

What does the exit look like? This is the one that deserves weight. A self-managed open-source system can be moved between providers. A managed proprietary one is a dependency on that provider's pricing and continued existence. Portability of the data and the interface matters more than the running cost, and it should be assessed before adoption rather than during a renegotiation.

What Spotify's example illustrates

They migrated event delivery from self-managed Kafka to a managed pub/sub service, publicly reasoning that the operational burden was not delivering competitive advantage and engineering time was better spent closer to the product.

The honest accounting of what they gave up: control over the substrate's behaviour and upgrade timing, the ability to tune for their workload, portability, and a cost model that scales with usage rather than with hardware — which can be better or worse depending on volume.

The decision

At small scale, managed nearly always wins. At very large scale the economics can invert, because usage-based pricing at extreme volume can exceed the fully-loaded cost of operating it. Model it with your actual volumes rather than assuming either direction.

Given two engineers holding the knowledge, the key-person risk alone probably decides it — and if the decision is to stay self-managed, that risk must be addressed explicitly rather than left implicit.

What a strong answer adds

The middle option: managed distribution of the open-source system rather than a proprietary service, which reduces operational burden while preserving the interface and therefore the exit.

Common weak answers

Comparing the monthly invoice against instance cost, which omits engineering time and risk. Choosing self-managed for control without naming what control is being used for.