Ask HN: Why are cheaper hosting services like OVH not very popular? | Hacker News Reader

Ask HN: Why are cheaper hosting services like OVH not very popular?

81 pointsrgun66 comments

We recently decided to move from AWS to OVH, primarily for saving money. We realized that the bulk of our AWS bill was made of EC2 and Data transfer costs. So, leaving S3, Cloudfront and SES, we moved everything away from AWS. The costs have become less than half.

My question is: Why is OVH not as popular as AWS, Google Cloud or Azure among startups when it is so much more cheaper?. I am aware of the following drawbacks of OVH:

  * No disk replication i.e unlike EBS or Google disk, if my disk fails, I am at the mercy of my backups (given no RAID) 
  * No private network for all the servers by default. Unlike AWS security groups or Google Cloud private networks (which are setup by default), I will have to pay for vRack to setup a private network and is more complicated.
  * Decreased flexibility i.e I may not be able to spin up instances instantaneously or shut down instances to save costs in off peak times.

These maybe big factors for large corporations but I think that a startup which starts with a small monolith (or a handful of services) and is focused on eventual profitability may benefit from using services like OVH.

Am I missing something here?

Also, please let me know of any better and cheaper services out there than OVH.

Edit: I would add that we have moved to OVH dedicated servers not VPS.

View on HN

OVH is massive and very popular in Europe because it's a French company with datacentres primarily in Europe. Meanwhile AWS is popular with startups from Silicon Valley who set the tone for this site and startup culture in general.

The other difference is that AWS has the first mover advantage in cloud services. OVH has always had the strongest presence in the dedicated server market and is only now catching up their OpenStack-based cloud offer.

I very much enjoy access to raw hardware and use OVH a lot.

Have you seen their new storage servers? These are basically hard drives attached to a calculator (ARM) with a NIC, amazing and I haven't seen anything like it elsewhere: https://www.soyoustart.com/en/server-storage/

They also have a top notch DDoS protection that is not based on your wallet outlasting the attackers.

> I am aware of the following drawbacks of OVH:

They are all false. OVH is a Public Cloud provider, just like AWS, GCP and Azure.

> * No disk replication i.e unlike EBS or Google disk, if my disk fails, I am at the mercy of my backups (given no RAID)

https://www.ovh.com/world/public-cloud/storage/additional-di...

> * No private network for all the servers by default. Unlike AWS security groups or Google Cloud private networks (which are setup by default), I will have to pay for vRack to setup a private network and is more complicated.

https://www.ovh.com/world/solutions/vrack/

> * Decreased flexibility i.e I may not be able to spin up instances instantaneously or shut down instances to save costs in off peak times.

https://www.ovh.com/world/public-cloud/instances/features/

You realize that you replied to a point saying "I will have to pay for vRack to setup a private network and is more complicated." with a link to vRack, right?

Right, because his claim is false. You can get up to 10 Gbps vRack for free, and it's included with Public Cloud.

True. I failed to mention that I was talking about dedicated servers. In which case, the servers with vRack enabled are costlier.

I did not check for their VPS offering.

That's true. But OVH provides a free private network with all public cloud instances, just like AWS and GCP. If we are comparing apples to apples.

I think it is because many people's bill is not primarily in EC2.

AWS, Google Cloud, and Azure's biggest value add come from managed services, like auto scaling databases and VMs. It makes it a lot easier for a small startup to build out their services without thinking about scaling challenges.

Snapchat, for example, was built on google app engine and to this day has billions of dollars in contracts with Google Cloud. They scaled their app to millions of users remarkably well with what I hear was just 2 guys.

No one does a startup because they expect it to stay small, so eating the upfront cost of using a managed service is a reasonable price to pay for peace of mind if things take off and the team can't keep up.

Hiring people to manage infrastructure and do it well is likely more expensive in many scenarios.

I'd be interested to see an analysis for a typical startup of when managing your own infrastructure becomes less expensive than managed services from cloud providers. I'd imagine it has to be pretty high considering most startups don't rush to leave AWS even after becoming massive.

Once a startup has seed funding it can typically get a ton of free credits from AWS using their "activate" program. Depending on the company this can be anywhere from $15k to $250k in AWS credits (although $100k seems to be the most common).

https://aws.amazon.com/activate/

This. If people wanted to manage dumb boxes it would be a no-brainer, but there's a crap ton of value add on managed services

Interesting that you say this. In my experience EC2 instances have been major part of bill. Can't comment on Snapchat scaling with just 2 guys but I have seen people trip up with managed services as well for eg. DynamoDB (https://hackernoon.com/beware-of-dilution-of-dynamodb-throug...). Similarly for EMR - lot of issues where you still needed to understand how Hadoop was managed by EMR (Spark and HBase versions, jar mismatch, classpaths). So if you are getting into something complex you still need to hire people to manage infrastructure of course not at the same level as building your own DC but OP is talking about using a cheaper cloud provider.

My personal opinion is to use EC2 and then take managed services from cloud agnostic vendor (preferably Open Source at core providing services on all major clouds). Does not lock you and lets you move fast - ex. Managed Kafka/Hortonworks

Before moving to GCP we actually did experiment with putting some of our services on a VPS at OVH as well as other providers to see how well they would handle our traffic. The issue we had with OVH specifically was that whenever traffic got high they thought there was an attack in place and automatic firewalls went up that took us down for hours at a time. Upon contacting support we were told "After looking into the matter, it would seem that our VPS do not have a profile that would accomodate your traffic spikes"

We've never had an issue like this after moving to GCP.

You should have used OVH Public Cloud[1] (= AWS EC2), not OVH VPS (limited to 100 Mpbs).

[1] https://www.ovh.com/world/public-cloud/instances/prices/

You would think OVH would have said that if they wanted the business

The thing with OVH is, that it's a fully automated service with a very basic level of support. There is no hand-holding. They expect you to read the docs yourself.

Disclamer: working at OVH.

both right and wrong.

Yes we provide fully automated services and it's out goal.

We have three levels of service support. One that is free (hotline, email, social, ...) and one where you can subscribe VIP support, and one where you can have a dedicated technical account manager. (Maybe we're not good at communicating)

For this example, if there is an issue with a DDoS profile, we are able to adjust it given you're workload. Most common use cases are not causing issues. Keep in mind that a "VPS" is not really something you should use for business because it has shared resources and very small sizing. This is probably why your workload was identified as suspicious on the DDoS side. Instead you should pick a cloud instance : more sizing choices and dedicated resources.

> Keep in mind that a "VPS" is not really something you should use for business because it has shared resources and very small sizing.

That's exactly what I meant. The problem is, that people expect OVH VPS to be a direct alternative to Linode, DigitalOcean, and Vultr – when it's not. And the free support service doesn't always make this clear enough to the users. It only becomes clear, when you read the docs yourself.

There are all kinds of good and valid business reasons why, and all kinds of good and perhaps valid counterarguments to those reasons. None of which are the real answer. The real answer is that the CEO/CTO/CxO doesn't want to have to think about these things, or getting expertise in-house for these things. They want it to be like electricity or water; something they have to pay for, but don't have to think about. Whether or not that's what really happens is irrelevant; it's what they think will happen, and to be honest it is what often happens. Even if they're paying more $$, they paying less in mental bandwidth, and never have to worry about whether they have (retained or retrained) sufficient in-house talent to maintain it. That's the real reason.

This is effectively false because you just shift the burden of understanding infrastructure intricacies from a nominal infra team to a bunch of fractured developer teams, and they often have fight through idiot org-wide cloud policies to get anything done.

Whatever you think you save by not having to hire or retrain AWS expertise on a nominal infra team, you lose far more by now needing to hire web developers, database engineers, machine learning engineers, mobile developers, etc., that are not just very good in their particular application domain, but also competent to operate self-service AWS tooling in an optimal way, despite having to contort around and fight with org-wide cloud policies that inhibit them for no actual cost, security, uniformity, or other benefit.

What I mean specifically is that there is not even a reason to think that shifting to managed services with a cloud provider will allow you to cut costs in IT workforce or “not think about it” like you say.

It’s just marketing snake oil that CTOs continue to believe this sort of thing at all.

It is entirely possible that you are just shifting the problem around to a place where it is less visible. Answer to the original question would still be the same. But, fwiw, I think you may sometimes be correct, especially since Amazon's flotilla of services is ever-expanding, resulting in having to know just as much as before, just for a proprietary domain. One is reminded of how corporations get talked into putting their businesses into SAP-world, and then discover that they need to hire SAP specialists to run it.

I'm gonna say overly high expectations probably have an effect here. As you said, these things aren't necessary for a startup with a small amount of users, but... very few startups seem to realise that, or consider that they won't become the next Google or Facebook overnight.

So services like OVH aren't as popular as AWS because startups always buy on the assumption they'll suddenly hit it big, even if their needs are more modest.

It could also come down to branding too; AWS and Google Cloud and Azure have done a lot of marketing in order to get themselves out there in front of the startup crowd, with whole conferences and events on how to use them. Hence some people are likely using them because it's 'the thing startups do'.

AWS, GCE and Azure solve problems that OVH, Hetzner, Online.net etc do not. For many businesses there is a lot of value here. Some of it is because the businesses do not spend that much on server hosting, some because the problems that AWS solves are very valuable.

If you do want to use a bare metal provider like OVH, keep in mind the differences from a service like EC2. You have physical disks. So you either need to implement network storage, or raid, to improve reliability and uptime, or make your software tolerant of data loss. Restoring a bare metal server from a backup has some considerations that a virtual server will not. You also have physical memory, cpus, nics, but in practice these do not fail often. Other areas of the infrastructure are less redundant or reliable because they are a budget provider. You do not have nearly as many physical locations to choose from.

A lot of older businesses are used to buying servers and collocating them instead of renting the whole server and could be saving a lot of money. Or businesses that insist that everything is on prem and never consider the whole cost of power, cooling, security, and connectivity. There are a lot of newer businesses that never really considered something like OVH because they started in AWS, and have only ever considered services like EC2 and competing VPS services, and not bare metal providers like OVH. You also have businesses that know that OVH etc is cheaper, but want to use some new thing for other reasons. It could be to improve their resume, or because this is what investors expect, or because this is what they think the companies that will eventually acquire them will want.

OVH is not typically seen as a big provider in the US but has a nice euro footprint. So some of it is branding and marketing. I used to work with most of the American leadership at OVH and know for a fact that they view the next couple years as an opportunity to become one of the key cloud providers in the us (along with the big 3). They see themselves as every bit as technically strong, just need to market better.

Because EC2 is not the only service that you'll use. Having services such as SQS, RDS, EBS, and Route53 all available and integrated into the VPC and access controlled via IAM is incredibly valuable.

External tools such as Terraform and Kubernetes already supports AWS. For smaller hosting providers this is usually not the case.

The big cloud providers all give a massive amount of credits for startups to use. Massive meaning multiple hundreds of thousands of dollars. So it’s easy for startups to hop on for the free credits and a couple years later be too busy on product cycles to make time for a move. The “lock-in” is strong.

We did receive large amounts of AWS and Google cloud credits.

I did not mention in the question for brevity, but we moved to Google Cloud (GCP) after exhausting AWS credits to utilise GCP credits. We came to know of OVH a couple of months back and having burnt through our GCP credits, we are again making a switch.

How much did you receive, on the order of $10k or on the order of $100k?

If it's really that easy to switch for y'all that you can be doing it to save money in cloud expenses, then either you really are best off with something like OVS/renting servers in some facility.

In the tune of $10k.

We also came to the same conclusion. Moreover, being an Indian startup, these dollar bills are burning a huge hole in our pocket :D

Took a look at Provakil and it seems interesting.

May I ask how much a typical engineer startup salary is in India? I had heard they got pretty high with all the US companies opening offices there (like 50-60k USD+)

I host some (personal/hobby) stuff on Scaleway. It's dirt cheap (a couple bucks a month for a Tor node pushing ~100 Mbps) and I like them, but they often have stockouts (can't start any new servers because they're out of servers, IPs, or some other resource) or other reliability issues (e.g. a server taking forever to start/stop). I can highly recommend them for hobby stuff, but wouldn't dare to use them for something that will cost me my livelihood if it goes down for a week.

I've heard various horror stories about OVH in terms of internal network reliability, but that was 5+ years ago.

Other reasons:

* The managed services (starting with the various file storage solutions, but extending to databases etc.)

* The list prices you see on the web site may not be the prices bigger companies actually pay.

* It's easier to find people who know how to use them, or tools that work with them.

* Possibly better service (if you have the appropriate plan - running your prod on a cloud provider without a service plan is a good idea until shit hits the fan, and a bad and possibly company-ending idea once that happens).

Personally, the 'free tiers' offered by the big PAAS and cloud providers are good enough for small projects and MVPs-- can't beat free. If and when extra resources are needed, it's usually easier to just start paying at that point rather than to switch platforms entirely.

Free tiers are are extremely effective (albeit crafty) in a "the-first-one-is-on-the-house" kinda way.

> Edit: I would add that we have moved to OVH dedicated servers not VPS.

This invalidates your other points. Your question actually is why are dedicated servers providers not as popular as cloud providers among startups. And I'm sure you'll come with the answer yourself.

Can you please elaborate? (Apart from the points I have already mentioned)

For me, it's how the hosting provider handles stuff when things go wrong:

- How reliable are their datacenters?

- Do their servers use good hardware? e.g. Hetzner is known to replace HDs on their budget servers with used HDs that have been "retested." Whatever it is, they fail more often than HDs at other hosts.

- Do they have organized procedures I can bet my business on? (DDoS response, hardware failure, power failure, network failure)

- Do they have a dedicated security team?

- Do they have professional, knowledgeable support? Low cost providers are typically lacking, IMO. When shit hits the fan and I need top-notch, professional support at 2am, can I get it?

To me, the only budget host that seems on par with the bigger companies is DigitalOcean.

We are running our services with Hetzner for years now. The less critical things run on low budget servers, for which we don't care too much if they break (everything is redundant - however, we had one hdd crash in 6 years).

The real important things run on dedicated server hardware. While these servers are more costly, they also are more powerful and unbelievably cheaper than AWS and alike.

I admit we do have dedicated people to work with our servers, but I can't imagine that other companies running on AWS don't have sysadmins to take care of their business critical services.

As for the support: It's always a question on how much you are willing to pay. Cloud services provide almost nothing unless you sign up for a dedicated support plan. We do have this with Hetzner, too. In case shit hits the fan, we can call them 24/7 (luckily this was never needed).

their website is awful, when I last paid them they still didn't do recurring billing, so for each service (i.e. virtual machine) you have to remember pay them monthly for each VM, each month

or you pay annually... then if one of your VMs dies good luck getting their CS to do anything about it (or even ever finding out why)

First part is not true for their cloud offerings, I think it was only for their cheap brands like Kimsufi and SoYouStart.

If your VM dies you are better off spawning a new one and migrate to the new one.

The reality is that they solve very different things. Part of what you're getting with AWS, GCP, Azure (DO, etc) is API support, and a big set of features. I view the big ones as a PaaS, and 'smaller' players as IaaS. Ultimately IMHO selecting a provider is either a risk mitigation strategy.

OVH in particular is really just phenomenal for dedicated hardware at a very reasonable price. They deploy very quickly (anywhere from 3 minutes to 48 hours, depending), and I've always found them INCREDIBLE to deal with - the last 4 years. Prior to that, when they were the go-to host for lowendwebhosting/cheapwebhost/lowendbox/etc they were a 'best of luck' provider. And that caused grief for many a friend / startup / client.

There's really a risk/reward difference. With AWS and friends, I'm getting incredible scale. With OVH and friends I'm getting tremendous disk space, processing power, but the scaling can be at best a PITA.

Ultimately I'm a big fan of physical to cloud bursting - if that's your need. Monitor your environment from a third party and use that to orchestrate your cloud-bursting strategy if necessary. It's fairly easy to do, if not a tiny bit time consuming.

That's simply not true. OVH launched their cloud lineup years ago with locations everywhere around the globe. Sure, you get bare metal servers, but similar to EC2 you also get your "cloud" machines that are spawned within seconds via their API.

They have their instances, sure but that's different than being a PaaS. It's about more than just compute now. When it's not then commodity is commodity, like Atlantic and Viper - they both have Api supports.

You mentioned DO, how is DO a PaaS service?

Agreed - DO isn't a PaaS. That's was silly of me.

Did you use spot instances when possible?

Did you use autoscaling to keep your instance count as small as possible?

These could be reasons why your costs were so high.

Another reason why companies use AWS is that it allows them to move faster. If you're afraid of lock-in and want to build everything yourself then yes, AWS will be damn expensive. But if you use the services they provide, you can build things pretty quickly without much workforce.

Managing your own servers instead of managed services is hard! It requires admin skills in addition to dev skills on your team, gets seriously hard with increasing levels of security, makes planning and spikes much harder, ... but can also be highly rewarding in terms of costs, mastering your environment, optimization and dev quality, ...

The choice is mostly about how much time and will you have or plan to have on your team for it. You'll have to handle files and database replications, you'll have to update the systems (with reboots), you'll have to fix very tricky issues, ... that would have been handled otherwise.

OVH (and some others like Online.net) have great products with amazing pricings. I have seen a few startups relying on many of their (rather small) servers and it was all great.

One thing to remember is that human time costs much higher than cloud time.

To add to many other valid points others have provided, OVH is a discount provider and they use second-hand / refurbished hardware at their data centers to keep costs low. I used to manage hundreds of servers on OVH and found that their hardware failed much more frequently than even us-east-1 on EC2. Most common issues were memory and disk related failures. A few times, their techs tripped over a power cord that took a few of our racks down, and they used to have frequent issues with network connectivity (I believe it's gotten better since).

On top of it, you will find that they cut costs on some of the less obvious things. For example, one day we found that their vRacks are not redundant, so a failure in one of them caused hours of downtime of our intranet.

As far as customer support - our account manager was always very helpful and understanding.

> they use second-hand / refurbished hardware at their data centers

OVH manufactures its own servers in France[1]. That's as "first-hand", as it can be.

[1] https://twitter.com/olesovhcom/status/1009539890005118976

OVH' stance is "we keep our customer support shitty to keep our prices low".

This is hit-and-miss marketing strategy that works for some, but not for others.

I use them but tbh I miss quality, responsive customer support when things gets bad.

AWS/GC/Azure have many different services well-known APIs/SKDs, which are like Lego to build products quickly. So if all you need is hosting for a webapp, you aren't taking full advantage of the offering. Then providers like DigitalOcean/Linode/OVH can definitely make sense. Heroku isn't quite like either of these, but another option if you want even less ops work.

Anyway, those legos also give you a certain future-proofing. Even with normal growth, it's easy to outgrow environments, and provider migrations/cross-provider apps are a pain.

I don't have answer to you question, just a few notes about another low cost proder - Hetzner: * They have RAID by default * They have private network included now (vSwitch)

Drawbacks from my expirience: * You have to monitor hardware by yourself (SMART, CPU and HDD temperature) and requests checks if you suspect hardware failure * Decreased flexibility (curretly Hetzner have limited cloud support, but it's very limited). Probably doesn't matter for most startups. * More manual work for administration (provided services are more low level)

AWS may be the most flexible option out there. And if you need flexibility, that matters. But If you are more established, without changing needs, and can define exactly what yhose needs are, other vendors often will be cheaper.

Startups therefore like AWS - they have too many unknowns. (And too many "don't know what they don't know" factors). More established, mature organizations have more of an option to lose flexibility in exchange for a lower cost.

I don't know about nowadays, but 4 years ago, God forbid you did anything against OVH rules (which apparently included running a mailing list), you'd get your whole account suspended and access to all your servers frozen. They also really took their time to reply (2 days with inaccessible websites? Insane).

Hardware problems? We feel bad for you son, wait a week maybe we'll look into it.

But it was really cheap.

Fast forward to today, never again.

i guess it depends on how you start developing.

our service was developed on local servers several years ago and then moved to OVH for hosting. eventually we managed to break up the server into multiple (LXC) containers. maybe in the future we may move some services to docker or to the OVH cloud.

since this is mainly a hardware business not a lot of work is being invested into software development. mainly maintenance and scale for more users (which is solved by getting beefier servers).

we also have 20TB of customer data that we need to store. dedicated OVH disk servers cost a lot less than a managed NAS. there may be other options, but i doubt we could take advantage of those without rewriting the applications.

a change like that would be very costly.

but now consider the other side: if you start developing on AWS, how likely is it going to be that you'll move to dedicated servers that you need to manage yourself?

it's very expensive to make that kind of switch and only worth it if AWS turns out to be inadequate in a serious way.

greetings, eMBee.

I left OVH years ago after several outages (network, electricity and sometimes hardware failures) Also the OVH support was between horrible to nonexistent. All our servers went to AWS, cost more money for the hosting but a lot less in staff and no more downtimes since years! If you have serious workload and critical systems to run OVH is a joke

Why not Hetzner or Leaseweb or Rackspace? There are lots of choice among big EU-based bare metal hosting services.

Disclamer : working at OVH ;)

Few points come to my mind :

* Price versus popularity

Being cheaper is not our drive value. If we wanted to be the cheapest, we would miss our mission to provide a world class infrastructure that you can rely on to build your own business. We wouldn't have one of the best anti ddos solution and we wouldn't be recognized as one of the most performant Cloud provider. ( https://cloudspectator.com/reports/2017-Cloud-Spectator-EU-R... ). I'd rather say that our services are more "fair prices" than "cheap prices".

Fair prices, being lower than US market, many customers tend to think that the value isn't the same (see cloud spectator link above). If your price isn't considered credible, you won't be at the table not because of your price, but because of the market range that customer are looking for. Range is defining both min and max, and as hard as it can be, there is a minimum value that customers want to put money into.

Popularity is also something that you work on the long run. OVH is very famous in EU, because... it's will be ~20years (next year) that we've been here starting from web hosting and baremetal, then accompany our customers in their growing needs. This year, we've just opened our 2 first US datacenters, and launched a beta offering with baremetal, Hosted private cloud (VMWare), and Public Cloud (OpenStack).

Popularity will come with more people testing our services :)

* IaaS / PaaS / SaaS in an open world

We also provide managed services like databases, Observability, Object Storage, K8S, etc. Some may not be available in US for now but it will be. We work with many startups, and we also have a Startup program called Digital Launch Pad : https://www.ovh.com/world/dlp/ At least in EU, many startups now know this program. I guess it's popular :)

A very big difference is how we provide the service and the lock-in policy. We want our customer to use our service for the value we provide and not because they're too tied to go elsewhere. Aside from a fair pricing, it's fair business. It means there are no hidden cost, bandwidth is included with the service, and PaaS/SaaS services offer well known (or standard if possible) APIs. For example, our observability solution provide protocols abstraction with many popular API : OpenTSDB, Warp10, Prometheus, InfluxDB, Graphite, ... Customers can push datapoints with a protocol and query with another one. This is this kind of openness we want to create. Repeat the same with everything else : Compute (Nova, EC2, ...), Object Storage (Swift, S3), Pubsub (HTTP, Kafka, ...).

We're seeing more and more companies that, like you, are fed-up with lock-in solutions or extravagant pricing policies, and are moving to providers like us. Thank you for your comment and let's make (profitable ;) ) business together!

IAM and the various sdks/aws cli are worth the price alone.

Latency to Europe is a non-starter for US companies.

They got data centers in North America you know. Canada and US.

AWS and GCE and the like aren't selling vpses, or dedicated servers, or even cloud services.

They are selling 9s of uptime.

And for the amount of 9s you get, it is the cheapest way to get it.

I use both OVH and AWS extensively, and have done so for quite a while. OVH is actually quite popular; they're big enough to have their own TLD, which says something about the quantity of money they have just sitting around.

There are a few big issues that I have with OVH, all of which keep me from moving my most critical infrastructure there:

* Their support is very hit-or-miss. Sometimes they're great; other times, you'll be stuck with a dead disk that they're not willing to replace, for whatever reason. Sometimes they swap out the wrong disk. Sometimes they just screw up your server for seemingly no good reason.

* Behind-the-scenes, they operate mostly in French. I don't speak French. Not many of the people I work with are fluent in French. But whenever you hit an edge case, or some unusual part of the interface, or really, anything off the beaten trail, the message is going to be in French. Error messages? French. Recovery prompts? French. Important status announcements? Quite often French.

* Not a single US datacenter--which, being a US company, is a bit annoying.

* The Canada datacenter, the only one close to our physical location, has serious network issues far too often. Once every couple years, their network will just... die. There's always a good reason, but excuses don't keep my customers happy. Maybe construction was taking place nearby and someone busted a fiber line, while the other line was down for maintenance. Whatever the reason, it ends up down for unacceptably long periods of time. I still have nightmares about the last time it was down for what felt like a whole day, but was probably more like 12 hours (so much better, I know). They keep insisting they're improving their peering. Apparently they have a lot of trouble getting permits to run lines across the US - CA border. In the time it took them to get those permits, they could've built a whole new datacenter in the US to completely mitigate the issue. It's not like they don't have the money. But they keep saying it's getting better. It's not going to happen again. Everything's fixed. More peering, more connections. But it Keeps. Happening. I've seen more reliable hosting providers that are an order of magnitude smaller--it's beyond the point of ridicule.

* Until very recently, their billing was a joke. There was no way to automatically pay invoices; each month, you had to pay everything manually--which wouldn't have been so bad if the interface were less buggy and confusing. They added automated payments but didn't fix the interface, so everything works great until it doesn't.

But they're cheap so we plop some failover infrastructure there and pray we never have to use it.

Edit: I should probably note that if we were building our primary infrastructure on OVH, it would probably actually be more expensive because of the increase in man-hours we'd have to spend managing it ourselves, even if we used their cloud services. It's been a long time since we crunched the numbers on this one, but last we checked, AWS won hands down, no contest.

I've tried to some extent or another most of the cloud providers. What keeps me going back to AWS for my own needs and when I'm working with startups or larger companies.

* Free tier - you can't beat free for very small companies, and you can scale a t2.micro surprisingly well.

* Credits - it's pretty easy for startups to get $5-20k in credits for a 1-2 years. This is a great time to fire up a couple of EC2 instances, some RDS instances, CDN, auto backups, monitoring, etc etc. You don't have to worry about costs so you can always have duplicate dev, staging, prod environments. If you're still worried about budget in two years when the credits run out, you can scale down pretty easily -- hopefully at that time you're ready to scale up rather, and if you need the extra functionality it's right there instead of migrating clouds.

* SDK support - AWS command line tools, Python (boto), Java, and pretty much any language you need can integrate really easily with most of AWS.

* Documentation - it's not great, but there's a lot of it. You can find the official documentation on how to integrate the various services together, and best practices for security and scaling them, plus a gazillion third party tutorials that you can follow line by line or download third-party libraries if you need to do something in a hurry.

* Support - even on the free tier where they don't officially offer it, the giants have incentive to provide really good technical support to help you spend more money with them.

* Experience - you can find more developers and other technical and non-technical people who are used to the AWS console and terminology.

* Region support - even here in South Africa I can have an edge location for my CDN. Hoping for an actual data centre here too in the near future.

* Versatility - whether I want to scale a web app, or fire up a quick P2 GPU instance to try out some image classification, I can use the same interface and toolset. If I suddenly want to view my Apache logs in the cloud, I can send them to the same dashboard as my heartbeat alarms in <1h of set up time.

* Performance - I won't go find them now, but there are a lot of benchmark blog posts out there showing how the smaller providers sometimes cut costs by using more shared hardware than they make obvious from their marketing - not all cores are equal and for sustained usage, AWS does a pretty good job of being predictable.

* Security - IAM is useful. S3 has sane defaults (now) which make it very hard to accidentally make data public. EC2 instances come with all the ports closed and you have to manually open the ones you want.

Maybe I sound like I'm trying to shill AWS (I did work for them a few years ago), but it's honestly one of the products I am happy to pay for. They've done a great job. GCE is really fantastic to use too, and Azure to a lesser extent. I still keep around a couple of Digital Ocean instances because I'm too lazy to migrate them anywhere and I love the simplicity of the platform and a Scaleway one because it's EUR2.99/month and it's useful to have a box that's just there to play with when I need it.

AWS is the new "nobody has ever been fired for buying"...

Further, given that having AWS experience is valuable, both the non-technical and the technical people are incentivized to use it.

Never thought about it like that, but you are exactly right.

OVH support is atrocious. If you have a problem, fasten your seatbelt.

OVH is probably a bad example, they targeted mostly French and East Europeans willing to setup file sharing among friends.

They are still rebooting your servers if you happen to block ICMP requests, in an attempt to bring them back to live.

Yes, they diversified, but the spirit stayed.

Hetzner is much more professional, they targeted big file sharers and the porn industry from the inception.

You can disable the ovh monitoring from your manager.

Reply on news.ycombinator.com