I think if you need something more reliable than us-east-1 that you should be hosting on prem in facilities you own and operate.
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.
I like this a lot, this is a great comparison for hetzner american offerings since it's not big enough for them to even bother investing much into it so there's not that many complains about it. People just dumping it (me included) after discovering the amount of random issues it has probably also doesn't help.
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.
Hetzner does not have any "fra region". They have Helsinki, Falkstien and Nuremberg in Europe. None of them which has any issues as far as I know. They used to have some issues with the very old stuff in Falkstien.
A sound banker, alas, is not one who foresees danger and avoids it, but one who, when he is ruined, is ruined in a conventional and orthodox way along with his fellows, so that no one can really blame him. JM Keynes
Ass covering-wise, you are probably better off going down with everyone else on us-east-1. The not so fun alternative: being targeted during an RCA explaining why you chose some random zone no one ever heard of.
Places nobody's ever heard of like "Ohio" or "Oregon"?
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.
IIRC, some AWS services are solely deployed on and/or entirely dependent on us-east-1. I don't recall which ones, but I very distinctly remember this coming up once.
The Route53 control plane is in us-east-1, with an optional temporary auto-failover to us-west-2 during outages. The data plane for public zones is globally distributed and highly resilient, with a 100% SLA. It continues to serve DNS records during regular control plane outages in us-east-1, but access to make changes is lost during outages.
CloudFront CDN has a similar setup. The SSL certificate and key have to be hosted in us-east-1 for control plane operations but once deployed, the public data plane is globally or regionally dispersed. There is no auto failover for the cert dependency yet. The SLA is only three 9s. Also depends on Route53.
The elephant in the room for hyperscalers is the potential for rogue employees or a cyber attack on a control plane. Considering the high stakes and economic criticality of these platforms, both are inevitable and both have likely already happened.
I find it funny that we see complaints about why software quality has got worse alongside people advocating to choose objectively risky AWS regions for career risk and blame minimisation reasons.
They are for the same reason. How do customers react to either? If us-east-1 fails, nobody complains. If Microsoft uses a browser to render components on Windows and eats all of your RAM, nobody complains.
Oh, people complain. The companies responsible have just gotten to the point where they are so entrenched that they don't need to care at all about customer complaints.
The value now is not really money from customers, but a company's share price or valuation. That, together with the hard push for subscriptions from every single app and service, devaluated customer experience and feedback. Because not many will go through the hell of unsubscribing process even after the outage or serious issues like private data stolen.
There's just not much motivation left to do better systems.
Istr major resource unavailability in US-East-2 during one of the big US-East-1 outages because people were trying to fail over. Then a week later there was a US-East-2 outage that didn't make the news.
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.
This is an interesting point. As recently as mid-2023 us-east-2 was 3 campuses with a 5 building design capacity at each. I know they've expanded by multiples since, but us-east-1 would still dwarf them.
Makes one wonder, does us-west-2 have the capacity to take on this surge?
> being targeted during an RCA explaining why you chose some random zone no one ever heard of.
“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…”
how about following the well-architected framework and building something with a suitable level of 9s where you can justify your decisions during a blameless postmortem (please stamp your buzzword bingo card for a prize.)
I look forward to the eventual launch of a new and improved version of your app using electron.
What’s the point in having 64 Gb of DDR5 and 16 cores @ 4.2 GHz if not to be able to have a couple electron apps sitting at idle yet somehow still using the equivalent computational resources of the most powerful supercomputer on earth in the mid 1990s.
This to me was the real lesson of the outage. A us-east-1 outage is treated like bad weather. A regional outage can be blamed on the dev. us-east-1 is too big to get blamed, which is why it should be the region of choice for an employee.
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.”
It's about risk profile. The question isn't "which region goes down the least" but "how often will I be blamed for an outage."
If you never get blamed for a US east outage, that's better than us-east-2 if that could get you blamed 0.5% of the time when it goes down and us1 isn't down or etc
But ise1 is down 4x more than use2 (AWS closely guards the numbers and won’t release them, but that is what I’ve seen from 3rd party analysis). Don’t you want your customers to say, “wow, half the internet was down today but XYZ service was up with no issues! I love them.”
I can’t tell if it’s you thinking this way, or if your company is setup to incentivize this. But either way, I think it’s suboptimal.
That’s not about “risk profile” of the business or making the right decision for the customer, that’s about risk profile of saving your own tail in the organizational gamesmanship sense. Which is a shame, tbh. For both the customer and for people making tech decisions.
I fully appreciate that some companies may encourage this behavior, and we all need a job so we have to work somewhere, but this type of thinking objectively leads to worse technology decisions and I hope I never have to work for a company that encourages this.
Edit: addressing blame when things go wrong… don’t you think it would be a better story to tell your boss that you did the right thing for the customer, rather than “I did this because everyone else does it, even though most of us agree it’s worse for the customer in general”. I would assume I’d get more blame for the 2nd decision than the 1st.
If my cloud provider goes down and my site is offline, my customers and my boss will be upset with me and demand I fix it as fast as possible. They will not care what caused it.
If my cloud provider goes down and also takes down Spotify, Snapchat, Venmo, Reddit, and a ton of other major services that my customers and my boss use daily, they will be much more understanding that there is a third party issue that we can more or less wait out.
Every provider has outages. US-east-2 will sometimes go down. If I'm not going to make a system that can fail over from one provider to another (which is a lot of work and can be expensive, and really won't be actively used often), it might be better to just use the popular one and go with the group.
us-east-2 goes down far, far less frequently than us-east-1. AWS doesn’t publicly release the outage numbers (they hold them very close to the chest) but some people have compiled the stats on their own if you poke around.
The regions provide the same functionality, so I see genuinely no downside or additional work to picking the 2 regions over the 2 regions.
It seems like one of those no brainer decisions to me. I take pride in being up when everyone else is down. 5 9s or bust, baby!
I’ve seen people go with IBM Cloud because their salespeople were willing to discount more heavily than AWS/GCP/Azure were. Tier 2 players can be hungrier for your business than tier 1 are. And here I’m talking about completely mainstream workloads (Linux, K8S, etc)
Separately from that, if you are trying to move certain types of non-mainstream IBM workloads to cloud (AIX, IBM i, z/OS) then IBM is tier 1 in that case
This story missed a glaring detail. There are simply more data centers in northern VA [0]. More than the rest of the US by a wide margin, or the entire EU+Asia. Things break here because it's where most things are.
Cackling while reading this visiting my family in Northern Virginia for the holidays. Despite it being a prominent place in the history of the web, it's still the least reliable AWS region (for now).
us-east-1 is often a lynchpin for services worldwide. Something hinky happening to dns or dynamodb in us-east-1 will probably wreck your day regardless of where you set up shop.
Yes, it's the least reliable. Thanks for summarizing the data here to illustrate the issue.
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.
I stopped deploying to a single region for production years ago, so I don’t really have a horse in this region comparison race. That said, I’ve seen network level issues in every region I use — nothing like the big outage, but issues that may disrupt a service. Designing for how the world is rather than how I wish it was makes a lot of sense to me.
I don't know if this is still true, or related, but that area used to be (Circa 10-30 years ago) very highly prone to power outages. The reason was lots of old trees near the lines that would inevitably fall; blackouts in local areas were common due to this.
That's an interesting data point, but I don't think it's relevant. The datacenters themselves are designed with a high level of power reliability and can island themselves if needed.
I think part of this is that Status Page updates require AWS engineers to post them. In the smaller Tokyo (ap-northeast-1) region, we've had several outages which didn't appear on the status page.
Us-east-1 is far far from least reliable. It’s one of the more reliable ones. Smaller regions tend to have more reliability issues affecting the entire AZ.
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.
I think if you need something more reliable than us-east-1 that you should be hosting on prem in facilities you own and operate.
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.
> worst case (so far)
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.
> A true cloud disaster would be something like a major S3 data loss event
So like the OVH data center fire back in 2021?
No, a major one.
(No shade on OVH, but they are ~1% market share player)
I mean, EBS went offline and people were ok to continue using AWS…
https://arstechnica.com/information-technology/2011/04/amazo...
There are only two kinds of cloud regions: the ones people complain about and the ones nobody uses
I like this a lot, this is a great comparison for hetzner american offerings since it's not big enough for them to even bother investing much into it so there's not that many complains about it. People just dumping it (me included) after discovering the amount of random issues it has probably also doesn't help.
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.
Hetzner does not have any "fra region". They have Helsinki, Falkstien and Nuremberg in Europe. None of them which has any issues as far as I know. They used to have some issues with the very old stuff in Falkstien.
sorry, fsn* I have them typod fra internally and keep messing it up since it's stuck in my head.
Yeah, I was often the single source of reporting Claude outages (or even missing support completely) on less commonly used Amazon Bedrock regions.
Which regions were you using ? ( Thought claude had global inference support that routed to all regions)
A sound banker, alas, is not one who foresees danger and avoids it, but one who, when he is ruined, is ruined in a conventional and orthodox way along with his fellows, so that no one can really blame him. JM Keynes
That is incredibly appropriate.
Eu-west-1 is miles better and is huge
Ass covering-wise, you are probably better off going down with everyone else on us-east-1. The not so fun alternative: being targeted during an RCA explaining why you chose some random zone no one ever heard of.
Places nobody's ever heard of like "Ohio" or "Oregon"?
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.
IIRC, some AWS services are solely deployed on and/or entirely dependent on us-east-1. I don't recall which ones, but I very distinctly remember this coming up once.
AWS IAM has caused multiple cross-region outages.
CloudFront certificates
Everything new basically, like the AI services.
IAM and Route53 have dependencies on us-east-1.
AWS Organizations/Account management is us-east-1.
And if you want a CDN with a custom hostname and want TLS…you have to use us-east-1.
The Route53 control plane is in us-east-1, with an optional temporary auto-failover to us-west-2 during outages. The data plane for public zones is globally distributed and highly resilient, with a 100% SLA. It continues to serve DNS records during regular control plane outages in us-east-1, but access to make changes is lost during outages.
CloudFront CDN has a similar setup. The SSL certificate and key have to be hosted in us-east-1 for control plane operations but once deployed, the public data plane is globally or regionally dispersed. There is no auto failover for the cert dependency yet. The SLA is only three 9s. Also depends on Route53.
The elephant in the room for hyperscalers is the potential for rogue employees or a cyber attack on a control plane. Considering the high stakes and economic criticality of these platforms, both are inevitable and both have likely already happened.
IAM
I find it funny that we see complaints about why software quality has got worse alongside people advocating to choose objectively risky AWS regions for career risk and blame minimisation reasons.
This was always the case. The OG saying was “no one got fired for buying IBM”. Then it was changed to Microsoft. And so on..
They are for the same reason. How do customers react to either? If us-east-1 fails, nobody complains. If Microsoft uses a browser to render components on Windows and eats all of your RAM, nobody complains.
Oh, people complain. The companies responsible have just gotten to the point where they are so entrenched that they don't need to care at all about customer complaints.
The value now is not really money from customers, but a company's share price or valuation. That, together with the hard push for subscriptions from every single app and service, devaluated customer experience and feedback. Because not many will go through the hell of unsubscribing process even after the outage or serious issues like private data stolen.
There's just not much motivation left to do better systems.
It all sticks with the 'monopoly' scent.
Istr major resource unavailability in US-East-2 during one of the big US-East-1 outages because people were trying to fail over. Then a week later there was a US-East-2 outage that didn't make the news.
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.
This is an interesting point. As recently as mid-2023 us-east-2 was 3 campuses with a 5 building design capacity at each. I know they've expanded by multiples since, but us-east-1 would still dwarf them.
Makes one wonder, does us-west-2 have the capacity to take on this surge?
us-west-2 is indeed very large, but will still not be able to take a full failover from us-east-1
> explaining why you chose some random zone no one ever heard of
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.
> being targeted during an RCA explaining why you chose some random zone no one ever heard of.
“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
how about following the well-architected framework and building something with a suitable level of 9s where you can justify your decisions during a blameless postmortem (please stamp your buzzword bingo card for a prize.)
We vibe code everything in flavor of the month node frameworks, tyvm, because elixir is too hard to hire for (or some equally inane excuse)
I look forward to the eventual launch of a new and improved version of your app using electron.
What’s the point in having 64 Gb of DDR5 and 16 cores @ 4.2 GHz if not to be able to have a couple electron apps sitting at idle yet somehow still using the equivalent computational resources of the most powerful supercomputer on earth in the mid 1990s.
I agree with your post conceptually.
However: Don’t underestimate community support (in the areas you’re likely to want it) when comparing development stacks.
This to me was the real lesson of the outage. A us-east-1 outage is treated like bad weather. A regional outage can be blamed on the dev. us-east-1 is too big to get blamed, which is why it should be the region of choice for an employee.
Bizarre way of making decisions.
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.”
It's about risk profile. The question isn't "which region goes down the least" but "how often will I be blamed for an outage."
If you never get blamed for a US east outage, that's better than us-east-2 if that could get you blamed 0.5% of the time when it goes down and us1 isn't down or etc
But ise1 is down 4x more than use2 (AWS closely guards the numbers and won’t release them, but that is what I’ve seen from 3rd party analysis). Don’t you want your customers to say, “wow, half the internet was down today but XYZ service was up with no issues! I love them.”
I can’t tell if it’s you thinking this way, or if your company is setup to incentivize this. But either way, I think it’s suboptimal.
That’s not about “risk profile” of the business or making the right decision for the customer, that’s about risk profile of saving your own tail in the organizational gamesmanship sense. Which is a shame, tbh. For both the customer and for people making tech decisions.
I fully appreciate that some companies may encourage this behavior, and we all need a job so we have to work somewhere, but this type of thinking objectively leads to worse technology decisions and I hope I never have to work for a company that encourages this.
Edit: addressing blame when things go wrong… don’t you think it would be a better story to tell your boss that you did the right thing for the customer, rather than “I did this because everyone else does it, even though most of us agree it’s worse for the customer in general”. I would assume I’d get more blame for the 2nd decision than the 1st.
> Don’t you want your customers to say, “wow, half the internet was down today but XYZ service was up with no issues! I love them.”
See any companies getting credit for it in the last AWS outage? I didn't. My employers didn't reward vendors who stayed up during it.
We got credit for it.
Shame about your employer, though.
If my cloud provider goes down and my site is offline, my customers and my boss will be upset with me and demand I fix it as fast as possible. They will not care what caused it.
If my cloud provider goes down and also takes down Spotify, Snapchat, Venmo, Reddit, and a ton of other major services that my customers and my boss use daily, they will be much more understanding that there is a third party issue that we can more or less wait out.
Every provider has outages. US-east-2 will sometimes go down. If I'm not going to make a system that can fail over from one provider to another (which is a lot of work and can be expensive, and really won't be actively used often), it might be better to just use the popular one and go with the group.
us-east-2 goes down far, far less frequently than us-east-1. AWS doesn’t publicly release the outage numbers (they hold them very close to the chest) but some people have compiled the stats on their own if you poke around.
The regions provide the same functionality, so I see genuinely no downside or additional work to picking the 2 regions over the 2 regions.
It seems like one of those no brainer decisions to me. I take pride in being up when everyone else is down. 5 9s or bust, baby!
I also don’t understand this.
US-East-2 staying up isn’t my responsibility. If I need my own failover, I’m going to select a different region anyway.
And it’s not like US-East-2 isn’t already huge and growing. It’s effectively becoming another US-East-1.
> US-East-2 staying up isn’t my responsibility.
No, but you can be blamed if other things are up and yours is not. If everyone's stuff is down, it is just a natural disaster.
Why aren't you using IBM cloud?
If IBM still had a good reputation, I probably would.
I’ve seen people go with IBM Cloud because their salespeople were willing to discount more heavily than AWS/GCP/Azure were. Tier 2 players can be hungrier for your business than tier 1 are. And here I’m talking about completely mainstream workloads (Linux, K8S, etc)
Separately from that, if you are trying to move certain types of non-mainstream IBM workloads to cloud (AIX, IBM i, z/OS) then IBM is tier 1 in that case
Bandwidth cost is also another major reason.
sort by duration on that page is broken
This story missed a glaring detail. There are simply more data centers in northern VA [0]. More than the rest of the US by a wide margin, or the entire EU+Asia. Things break here because it's where most things are.
[0]: https://www.datacenters.com/providers/amazon-aws/data-center...
At 34 hours of downtime that's two nines of uptime
At this point my garage is tied for reliability with us-east-1 largely because it got flooded 8 month ago.
Cackling while reading this visiting my family in Northern Virginia for the holidays. Despite it being a prominent place in the history of the web, it's still the least reliable AWS region (for now).
Its nice to know that where I grew up is Too Big to Fail lol.
I intentionally avoid using us-east-1 for anything, since I’ve seen so many outages.
us-east-1 is often a lynchpin for services worldwide. Something hinky happening to dns or dynamodb in us-east-1 will probably wreck your day regardless of where you set up shop.
The sorting for the "Duration" column appears to be lexicographical, not numeric.
Yes, it's the least reliable. Thanks for summarizing the data here to illustrate the issue.
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.
Nobody ever got fired for connecting to us-east-1
Answer these questions:
- 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_...
I stopped deploying to a single region for production years ago, so I don’t really have a horse in this region comparison race. That said, I’ve seen network level issues in every region I use — nothing like the big outage, but issues that may disrupt a service. Designing for how the world is rather than how I wish it was makes a lot of sense to me.
Of course it is, all of the NSA men in the middle add a lot of overhead that can interfere with regular operations.
I don't know if this is still true, or related, but that area used to be (Circa 10-30 years ago) very highly prone to power outages. The reason was lots of old trees near the lines that would inevitably fall; blackouts in local areas were common due to this.
That's an interesting data point, but I don't think it's relevant. The datacenters themselves are designed with a high level of power reliability and can island themselves if needed.
We've started to see some rather interesting consequences for grid reliability: https://blog.gridstatus.io/byte-blackouts-large-data-center-...
Glad to use us-west-2 for reasons.
I think part of this is that Status Page updates require AWS engineers to post them. In the smaller Tokyo (ap-northeast-1) region, we've had several outages which didn't appear on the status page.
The test environment is deployed on us-east-1, whereas the production environment deployed on us-west-2 on our side.
We get constant resource issues in GCP’s us-east4 region
Us-east-1 is far far from least reliable. It’s one of the more reliable ones. Smaller regions tend to have more reliability issues affecting the entire AZ.
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.
Yes
I searched for it, and did not find, the word "backhoe."
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.