AWS Data Loss After Iranian Strike Is a Real Wake-Up Call
AWS confirmed it cannot restore some data from its Middle East facilities that were struck in the conflict. The HN thread immediately went after the obvious question: AWS S3 promises eleven nines of durability. How does a physical strike cause permanent data loss? The answer people arrived at is that the SLA assumes normal failure modes, not geopolitical ones that take out multiple availability zones in the same region simultaneously.
This is a different failure mode than the usual outage or bug. It is a reminder that cloud durability guarantees are probabilistic models built on assumptions about independent failure events. When an external force hits multiple zones at once, those assumptions break.
The thread also surfaced the older 2017 S3 outage story about whether S3 was even down during that incident. There is a recurring pattern of AWS's status page lagging reality. Founders who treat cloud SLAs as guarantees rather than probabilistic models under specific assumptions are building on a shaky mental model.
So what?
If your data is irreplaceable and your business operates in or near geopolitically sensitive regions, a single-provider, single-region strategy is not a backup strategy. The boring answer is still the right one: geographic distribution across providers, not just availability zones within one provider's footprint. This incident makes that conversation easier to have with your board or CTO.