Hacker News Reader

I understand the cost and complexity of doing so but I think that production loads should be engineered across two or more service providers. That way any incident on a given provider is recoverable without that providers cooperation and you can mitigate against Oracle style lock-in.

View on HN

I somewhat agree in principle. At the same time, the (implied) promise of the big providers like AWS and GCP is to provide everything needed for a reliable service while saving the time for setting it all up yourself. If you need multiple platform providers with fallback logic etc., it is hard to see a scenario in which setting up your own infrastructure would not be more effective.

Perhaps the latter is the more practical solution indeed, though.

If you think that production loads should be designed as multi-cloud as redundancy, then you don't understand the cost of complexity and doing it.

Reply on news.ycombinator.com