Amazon: Some data won't be recovered after cloud outage
Post mortem wait — Amazon says that about 0.07 per cent of the EBS storage volumes in the East Region of its infrastructure cloud are not “fully recoverable” following the extended outage that hit the service last Thursday.
Context & Ripple Effects
Amazon's disclosure that roughly 0.07% of East Region EBS volumes are not fully recoverable closes one loop on the extended outage that took down customer sites last Thursday, and opens another: the company's full post-mortem is still pending. It is also a familiar shape for anyone tracking the platform — the 2008 S3 failure was the earlier proof that Amazon's cloud could fail in ways that took customers down with it.
The story has travelled beyond infrastructure trade press: Slate's pickup frames it as a referendum on whether repeated server failures will slow the rise of cloud computing itself, which is exactly the question enterprise buyers were already asking.
First-order effects
- Customers whose volumes fall in that 0.07% have suffered permanent data loss with no recovery path through Amazon — backup and disaster-recovery practices outside the affected region become their only recourse.
- Amazon faces immediate credibility pressure until its promised post-mortem explains how replication within EBS failed to preserve data it marketed as durable.
Second-order effects
- Enterprise buyers evaluating AWS contracts will press harder on multi-region architectures and off-cloud backups, shifting spending toward redundancy tools and rival clouds positioned on reliability guarantees.
- Competitors and managed-service providers gain a sales argument: single-region dependence on any one provider's block storage is now demonstrably a business-continuity risk, not just a theoretical one.
Third-order effects
- If recoverability disclosures like this become routine after major outages, durability claims move from marketing material to procurement checklists — buyers will demand stated loss percentages, regional isolation guarantees, and contractual remedies for unrecoverable data.
- Recurring single-region failures point toward structural change in how cloud applications are designed: multi-region and multi-provider resilience shifts from optional best practice to default engineering requirement for anything business-critical.
The trend: Cloud outages are converting data durability from a vendor promise into a measurable procurement criterion, pushing customers toward designs that assume any single region can lose data.