New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Things more reliable than AWS
As most people on the internet know, AWS had another large scale outage this morning in their US-East availability zone.
Many popular websites, airlines and other companies were impacted.
This "feels" like an annual or twice annual thing with AWS now.
So, it poses the question, is your LowEnd VPS more, or less, reliable than AWS?
What a world...
Comments
Fully on board, Jon.
a solid business should always have multiple nodes from different providers in different locations.
silly cat don't be silly!
Some prviderers here on LET are quite reliable and powerful. Not all of them, but some of them based on strong foundations.
That's a big question.
If we were to go on uptime numbers, the likelihood is my "low end" services actually probably do have slightly better uptime. (At least over the us-east locations...) There was a brief us-east-1 blip for GCP a few months ago too.
However, as far as consistent reliability at scale, I have seen several LowEnd machines have significant service degradation (such as poor network, noisy neighbours, security issues, full VM loss, node reboots without much warning) - generally with hyperscalers like AWS, my box is on and working amazingly - or it's off/not reachable.
So, I'd still say hyperscalers are more reliable. Even with the ceremonial "us-east1 is broken" once or twice a year.
(but, you should never have all your eggs in one basket anyway, and if you require pretty much permanent availability, relying on a single region is foolish. especially since AWS & friends make it relatively trivial to have multiple availbility zones).
How many posts you have created for this?
2+1 for redundancy
its all @shallownorthdakota fault!
Add one for me
Honestly, its just more sales material, everytime big tech giants like AWS goes down.
If you cant get on the phone with someone that actually matters, within minutes, it isn't worth it...
Time to bring back hosting to personal and non corporate giants again.
Def a unique situation when an outage like this can create this effect.
Is there any big cloud with 100% uptime? I think all of them go down sooner or later, no?
https://i.ibb.co/yrG3zcz/1760958447101.jpg
Wait, where are the other two nice threads? Eaten up by AWS the big boss of the universe?
Just because we made some fun of LET search func?
We want those two back to see which one derails away from topic faster.
RBAC is the only reason they win. Think about your next feature!
My dedi is more reliable and a few of my VPS are as well. But I'm not sure that's actually the point re. 99.999% availability.
But relying on a single provider, no matter how supposedly great, is not?
Also, if "oh well, occasionally my server is down/unreachable" is good enough, why go with AWS in the first place?
Almost certainly a major problem, yes. And a reason to run away from AWS (and from the other giants as well, for the same reason).
I find Hetzner and OVH to be more reliable than AWS.
1984 Toyota Hilux with 22RE motor.
I use AWS at scale for work and it's honestly trash. EC2 servers often brick on restart, (especially the expensive ones), and then you have to migrate the disk to a new instance.
One particularly bad day I lost 3 of them in a row, (each hooked up with a GPU), and it cost about 6hrs and $10 for the wasted effort. I finally redeployed to a new region, performing exactly the same setup steps, and everything was fine so it can't have been something I was screwing up.
Then there's just the general performance. The 'burstable' EC2s have various arbitrary limits, like you can download ~10GB fairly rapidly but then you get harshly limited, so downloading large amounts of data is a nightmare, and you also get a solid 70% Steal once you run out of CPU credits, so you really have to bed applications in to see how they perform after the CPU limit kicks in.
The only good thing I can say about them is that they're consistent, (as in consistently shit). if they were a LET host I'd have binned them ages ago, but I use their services to pay the mortgage.
iHostArt VPS is more reliable than AWS in terms of uptime per dollar.
Gut feeling says you wong if the metric is
1 / (downtime * price). Which is betta metric 💪🏿Let's keep other attempts linear BTW, otherwise the methods will start to look "promising"...
The metric is: (yearly uptime in seconds) / (yearly price in dollars)
what's the point of counting the nines then
(as in five nines)
you're cheating
Of all the hosting, services, and cloud services I've used, DigitalOcean has been the most reliable, and any maintenance windows are notified well in advance.
I think AWS, GCP, and Azure are all fairly comparable in being... imperfect. After all, these are huge, complex things... your typical LowEndVPS is far simpler with many fewer moving parts, so to speak. Yes, there is less redundancy, but it's also just so much less that can go wrong.
nissan altima in memphis, tn is more reliable than aws us-east-1
Reasons to use hyperscalers are, not surprisingly, scale, perf, convenience, and billing granularity. You can spin up and down hundreds and thousands of instances across multiple regions with high-end GPUs and 400Gbps interconnects and only get charged for the seconds you're using. Reliability is never their forte despite marketing claims. AWS EC2 SLA for 100% refund is only 95% (1.5 days/month outage), anything above that is only up to 30% refund.
The internal backend services of these hyperscalers are incredibly complex and brittle, as they're owned by many siloed and competing (OKRs) teams, no one fully understand the entire lifecycle of various resources. A single harmless looking commit that passes CI/CD can lay down latent landmines that are only triggered way later by another harmless looking commit from a different team. Making these LSEs and OMGs hard to prevent, predict, and RCA. It will only get worse, as they're racing to build out more for AI.
Diverse complementary (regions) vendors are good for building more resilient HA control planes and monitoring services.
Things more reliable than AWS: a coin toss, your WiFi during a thunderstorm, and my $1 VPS running on goodwill and duct tape.
I've noticed a semi shift of companies going back off the cloud and to colo space.
If I'd compare to CC, definitely less.