Why you choose big hosting provider like AWS?
Hi guys, this is my first post on this forum and I hope I’m not breaking any rules here.
I’ve been thinking about something that just doesn't sit right with me and is why are people still throwing 3x or 4x the money at AWS, Google Cloud, or Azure?
I get it, they have a massive ecosystem, but let’s be real. If your use case is just a VM sitting there running a service, whether it’s a simple script or a complex stack,why pay that insane premium? What exactly does AWS give you that justifies that bill? Is it just the brand name, or is there something I’m missing?
On the other hand, I’m tired of the "cheapest is best" mindset.
We all know the deal: if a VPS costs $1/year, it’s worth exactly what you paid for ancient hardware, heavy overselling, and zero stability. Quality and actual support have a cost, period.
So, if we take the "bottom-dollar" price out of the equation, what actually makes you choose one host over another?
Hardware Transparency? Network Quality?Real Support?
Also, I’m curious: what are you guys actually running on your VMs these days? Is it all Docker, personal VPNs, monitoring nodes, or something else?
Comments
I think you’re mostly paying for reliability, ecosystem, and convenience with AWS / OCI / GCP / Azure, not just raw compute. Things like managed databases, autoscaling, global networking, IAM, backups, and integrations save a lot of time (and headaches) at scale. For businesses, that often justifies the cost more thæn the hardware itself.
For a simple always-on VM though, yeah, it’s usually overkill. Plenty of smaller providers give you better value if you don’t need all the extra features.
When choosing a host, I’d say the big ones are:
As for workloads, it’s mostly Docker these days, apps, small services, monitoring, sometimes game servers or personal tools. Pretty boring, but it works
If you want information, feign ignorance reply with the wrong answer. Internet people will correct you ASAP!
It’s OK if you disagree with me. I can’t force you to be right!
I run stuff on Azure, Oracle and AWS because of the reliability. Literally years of VPS uptime on Azure.
It seems that what makes a difference is (a) live migration - I'm sure my VPS is not running on the same node/hardware for years without a reboot, they're doing proactive live migrations when there's a hardware issue; and (b) the network - even the "best" LES provider is at the mercy of their upstream and I'm getting some type of brief (at least) network outage from ALL of them more often than I'm happy with.
As for whether the incremental benefit is worth it, the issues with outages are that they tend to occur at the most inopportune times, and the uncertainty about when they'll be resolved. Is it a 1 minute "blip", or is it going to last hours? Is it worth migrating stuff somewhere else, or just ride it out? I'm getting better at this and have scripts that can do a migration between providers in a few minutes, and am working on automating that. And for sure not everything needs 100% reliability.
Last decade, I think they guaranteed 6 to 9, perhaps 11 nines on data reliability. Moreover, managed database with global availability is great. But, it's just expensive
As for computing, high-frequency cores aren't the default.
Aside from ecosystem there is also compliance. Bigger companies have a whole laundry list of requirements so they're never going to go for a LES style provider.
But who does that? I know people use Oracle free tier to run simple scripts, but other than that it must be extremely rare that people are paying premium providers to run simple scripts.
I have several customers that run with the big dragons for a bunch of reasons. It could be geographical spread, SLA's, some certain service or product, automation, or even legal. When you run 1000 production critical vm's you don't just click "I accept" on some random providers EULA and hope for the best, you tell your legal department to call their legal department and write down a shitload of contracts and get them signed. Small providers do not have that kind of resources. Also, those kind of contracts sometimes contains fines and stuff in the millions in case of contract breach, and small providers could never afford to sign such a clause.
Yes.
There's a few factors for me. One is cost. I explicitly preference smaller companies. I also run OpenBSD whenever possible, and ease of running OpenBSD is a factor.
I have three external name servers that also provide monitoring services.
ARPNetworks: This one is the most expensive at $10/month. It's twice the cost of comparable offerings, but it's another point of presence, I've had no issues with it, and they have a great reputation in the BSD community. I could spend elsewhere, but it's not breaking the bank, so why?
IRCNow: $5/month, which is what I've normally considered market rate for a basic VPS. They run OpenBSD as a hypervisor.
OpenBSD.Amsterdam: Definitely the gold standard for OpenBSD on OpenBSD. A little more money, maybe $6/month or so. A lot of these funds go back to the OpenBSD project.
In total, I spend about $21/month for off-site servers that I'm quite content with. I still have a couple stragglers on Vultr that I want to move over to more core setup in Spokane, but almost all of my stuff is in-colo now, on my own gear.
Slow Servers IPv6-native VPSs hosted on OpenBSD's VMM in Spokane, WA, USA. (I racked these.)
SporeStack Resold Vultr VPS/baremetal, DigitalOcean, and a whitelabeled brand in Europe. KYC-free, simple to launch. (I didn't rack these.)
Neither are dirt cheap!
what is security for you?
I know a large providers wont snoop through my servers and files randomly cause they are bored.
If you want information, feign ignorance reply with the wrong answer. Internet people will correct you ASAP!
It’s OK if you disagree with me. I can’t force you to be right!
But they might train their LLM on them .....
Nope, AWS/OCI/Azure cannot use their paying client data to train AI models unless client explicitly allows the use of their server data to train AI models. This is against their TOS (under privacy), and it's much cheaper to train the AI models on the "free" or "trial" accounts.
If you want information, feign ignorance reply with the wrong answer. Internet people will correct you ASAP!
It’s OK if you disagree with me. I can’t force you to be right!
This is ridiculous...
If I find out that my provider is going to start poking around on my server, I’ll report them immediately.
Even if I don’t have anything illegal or anything like that, why do you have to stick your nose in?
For now ........
I remember there was one provider here on LES (or was it LET) that admitted to look through their client's files and data to make sure they are not hosting anything illegal. He was caught when one of his client reported his server being terminated and the provider admitted to look through the files.
Long ago, I had a shared hosting provider who did the same and "recommended" me to change the database password to something more random. I am grateful for the recommendation, but I do not expect my provider to look through my config files to check what the database password is...
I doubt they will ever do it for the major clients. It's not worth it as those clients value privacy over anything else. If they start using their data for LLM training, they will just move out and file a lawsuite which will cost millions of dollars. So it's best to just do it for the free services they offer.
Example, microsoft using github free accounts code to train their github copilot AI model.
If you want information, feign ignorance reply with the wrong answer. Internet people will correct you ASAP!
It’s OK if you disagree with me. I can’t force you to be right!
@somik it was host.charity
You don’t. Some of the worst performance metrics and environmental waste there
Insert signature here, $5 tip required
Ah yes, you are right!
If you want information, feign ignorance reply with the wrong answer. Internet people will correct you ASAP!
It’s OK if you disagree with me. I can’t force you to be right!
>
If a client on my hosting site uses the service illegally, I only find out through an abuse report or a police inspection. In any case, I would never take it upon myself to access someone else's server just because they are using it illicitly...
That's because you are not a piece of sh*t provider. It's because you respect clients privacy.
If the client has his server compromised or was using the server illegally, you have the right to shut it down or remove it. If the client wants the data back, you can also ask the client before accessing the server. If the client asks for help with his server and provides server access details, you can login and fix as needed. All of these are behavior exhibited by good hosts.
But how many "small" providers follow these basic guidelines? Many small hosts do access the clients files and rummage through them cause they are free or want to find something incriminating or just snooping around.
So if you dont want to take the risk, go with a big hosting provider like AWS/OCI/Azure etc.
If you want information, feign ignorance reply with the wrong answer. Internet people will correct you ASAP!
It’s OK if you disagree with me. I can’t force you to be right!
Welcome to the forum! You touched on a very interesting point. In my experience as a sysadmin, the 'AWS Tax' is often paid for two reasons: Fear and Scalability APIs. Managers choose AWS so they don't get fired if things go down ('No one ever got fired for buying IBM/AWS'), even if they only need a simple VPS.
To answer your question on what makes me choose a host over another (excluding the $1/year trash):
Hardware & Virtualization Transparency: I want to know if I'm on a cramped E5 v2 or a modern EPYC/Xeon Gold. Also, KVM is a must; I stay away from heavily oversold OpenVZ containers.
Network Resilience: It’s not just about 1Gbps vs 10Gbps, but about the Peering. I look for providers who have solid uplinks to major IXPs.
The 'Privacy' Factor: Lately, I've seen a shift towards providers that don't demand a passport scan just to spin up a Linux instance. Sovereignty and data location are becoming as important as IOPS.
Currently, I’m mostly running Dockerized stacks and self-hosted monitoring nodes (Uptime Kuma/Zabbix). Moving away from the 'Big Three' clouds to lean, high-performance providers is the best move for ROI. Looking forward to seeing more discussions like this here!