There aren't that many businesses that truly can't handle the worst case (so far) AWS outage. Payment processing is the strongest example I can come up with that is incompatible with the SLA that a typical cloud provider can offer. Visa going down globally for even a few minutes might be worse than a small town losing its power grid for an entire week.
It's a hell of a lot easier to just go down with everyone else, apologize on Twitter, and enjoy a forced snow day. Don't let it frustrate you. Stay focused on the business and customer experience. It's not ideal to be down, but there are usually much bigger problems to solve. Chasing an extra x% of uptime per year is usually not worth a multicloud/region clusterfuck. These tend to be even less resilient on average.
It’s kind of amazing that after nearly 20 years of “cloud”, the worst case so far still hasn’t been all that bad. Outages are the mildest type of incident. A true cloud disaster would be something like a major S3 data loss event, or a compromise of the IAM control plane. That’s what it would take for people to take multi-region/multi-cloud seriously.
So like the OVH data center fire back in 2021?
You mean like stealing the master keys for Azure? Oh wait a minute...
https://arstechnica.com/information-technology/2011/04/amazo...
You forget things like emergency services. If we were to rely on AWS (even with a backup/DR zone in another region), and were to go down with everyone else and twiddle our fingers, houses burn down, people die, and our company has to pay abatements to the govt.
if you are using hetzner: avoid everything other than fra region, ideally pray that you are placed in the newer part of the datacenter since it has the upgraded switching spine I haven't seen the old one in a bit so they might have deprecated it entirely.
Yeah, I'm not worried about being targeted in an RCA and pointedly asked why I chose a region with way better uptime than `us-tirefire-1`.
What _is_ worth considering is whether your more carefully considered region will perform better during an actual outage where some critical AWS resource goes down in Virginia, taking my region with it anyway.
So if you tried to be "smart" and set up in Ohio you got crushed by the thundering herd coming out of Virginia and then bit again because aws barely cares about you region and neither does anyone else.
The truth is Amazon doesn't have any real backup for Virginia. They don't have the capacity anywhere else and the whole geographic distribution scheme is a chimera.
Makes one wonder, does us-west-2 have the capacity to take on this surge?
“Duh, because there’s an AZ in us-east-1 where you can’t configure EBS volumes for attachment to fargate launch type ECS tasks, of course. Everybody knows that…”
:p
Is this from real experience of something that actually happened, or just imagined?
The only things that matter in a decision are:
* Services that are available in the region
* (if relevant and critical) Latency to other services
* SLAs for the region
Everything else is irrelevant.
If you think AWS is so bad that their SLAs are not trustworthy, that's a different problem to solve.
us-east-2 is objectively a better region to pick if you want US east, yet you feel safer picking use1 because “I’m safer making a worse decision that everyone understands is worse, as long as everyone else does it as well.”
[0]: https://www.datacenters.com/providers/amazon-aws/data-center...
At this point my garage is tied for reliability with us-east-1 largely because it got flooded 8 month ago.
It's often seen as the "standard" or "default" region to use when spinning up new US-based AWS services, is the oldest AWS center, has the most interconnected systems, and likely has the highest average load.
It makes sense that us-east-1 has reliability problems, but I wish Amazon was a little more upfront about some of the risks when choosing that zone.
- Is X region and its services covered by a suitable SLA? https://aws.amazon.com/legal/service-level-agreements/
- Does X region have all the explicit services you need? (note things like certs and iam are "global" so often implicitly US-East-1)
- What are your PoP latency requirements?
- Do you have concerns about sovereign data: hosting, ingress, and egress? https://pages.awscloud.com/rs/112-TZM-766/images/AWS_Public_...
We've started to see some rather interesting consequences for grid reliability: https://blog.gridstatus.io/byte-blackouts-large-data-center-...
This analysis is skewed due to the major incident in 2025. What was the data for 2024 and over the last, say, 5 years? So the proclamation of least reliable of us-east-1 is based on 1 year of data, and it’s probably fair to say that at least last 3 years if not 5 are a better predictor of reliability.
us-east-1 also hosts some special things, so it will have more services to lose.
Big fail.
I have said for years, never ascribe to terrorism what can be attributed to some backhoe operator in Ashburn, Virginia.
We got a lotta backhoes in northern Virginia.
