All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Overselling Isn't Necessarily a Bad Thing
Let me offer a different perspective on overselling with an analogy.
Imagine a provider has a host server that can normally run 100 virtual machines (VMs), which they sell for $2/month each. At this price, a potential customer might hesitate before buying.
Now, what if the same provider uses that same host to run 200 VMs with the same specs, but drops the price to $1/month? The provider's operational costs don't significantly increase (especially with NAT), but customers will now likely buy without any hesitation.
The result is a win-win: the provider gets double the customers, and twice as many customers get a service they are happy to pay for.
On a related note:
Some might argue that since the total revenue remains the same, the increased cost from handling more support tickets means the provider is actually losing money.
I think this view is a bit short-sighted. A large customer base is always incredibly important. It's far better to have customers who submit support tickets than to have no customers at all. Furthermore, if those tickets are handled well, it can significantly boost customer loyalty.
A larger customer base also increases the visibility and potential sales of a provider's other products. These are all valuable, long-term benefits.
Comments
Overselling isn't an issue as long as the customer doesn't notice it with (for example):
Providers of VPS servers are usually expected to share resources between customers, but they should clearly explain what is the acceptable use and try to do their best to not affect the overall performance of VMs.
That's just my 2 cents
The only thing that will come of this is a Geekbench score of 200 points and closure a month after they collect their annual subscriptions
That’s what also my mom keeps saying since my dad left. Sad times.
Is this your business plan?
I think this provider and those customers probably deserve each other, good luck to all involved and look forward to reading the related posts.
Overselling isn't an issue when done right. But can you really consider that overselling by the definition: selling more resources than technically available?
Overselling for me is having nodes with a bad client mix resulting in bad performance of the node.
I think you underestimate support costs btw. Qualified in-house support is fucking expensive.
OP doesn't understand overselling. You can oversell unused resources, you can't oversell the minimum utilized resources.
Having 100 VPS with 1 vCPU, 1GB RAM and 10GB of storage is far more busier than 50 2 vCPU, 2GB RAM and 20GB VPS.
When all servers idle, the 1GB servers utilize closer to 100% of resources allocated resources, the 2GB closer to 50%. It's the amount of unused resources that can be oversold.
So fixed pricing by resources is not smart to increase revenue with increasing packages, pricing should be encouraged to increase the total amount of resources but not using 100% of the time.
Ie, 1TB $50, 2TB $90, 4TB $160, etc.
So someone who knows they'll store 800GB might stick with 1TB since there's no price break for 2TB, but if there was, they'd go to 2TB sooner and utilize less than half those resources whereas he'd consume 80% of the allocated storage.
Tl;dr price savings for increased resources is how to oversell and profit.
I refuse to oversubscribe SSD or RAM. It gives us higher performance than most other hosts.
Even my previous venture didn't, but they used old HPE DL360 servers nearing EOL while we're using brand-new Ryzen 9000 servers with NVMe.
You have to do it right, if you don't, you will get tickets.
I had to submit 3 tickets, to a 15€/y VPS, due to continuous issues.
Yet they are still dealing with the aftermath, offering yet another migration.
It's perfectly possible to combine a reasonable level of overselling and still maintain the performance to great levels. You just need to know where to look and have some common sense on certain implementations, and on what to do and what not to do.
And of course: tests, tests, tests. On-hand, off-hand, continuous revision on a given node, decent hands-on monitorization in the process, and so on. If it's not satisfactory for your usage, it won't be for your customers' usage and can't be sold - and this has to be seen prior to any node starting commercial activity.
That is what we do, I'm even a night owl at that and oftentimes go up until 4AM doing errands on that topic, because I want the VPSes at this company to have the least steal possible and a nice performance.
And then, we have a very interesting effect, that I think it may possibly never have been talked about here: the timezoning issue.
Because, on nodes where Chinese/Asian customers are a relevant percentage, I found it very interesting to see this: in European and American timezones there's not a relevant steal, it's for the most part a clean node... but after midnight UTC and until 8/9 AM there's a huge increase on the use of the nodes, to the extent that nodes that commonly have 0-2% steal suddenly may turn into 15-30%+ steal - and because of what?
Because these are the usual working hours in China and other countries. As it's natural everywhere. They purchased a service, they want to use it.
That led me to remove two VPSes I subscribed recently, for example, because the steal was so high overnight it actually managed to bring a secondary service down every 24h.
And then, sometimes, these Asia customers report to the provider that there is a loss of performance, or the steal is too high, or there is a slowness on doing things, and the provider replies like "oh, but we see everything okay" - because it is, but on the sysadmin's timezone. Not on the customer one.
So definetly something interesting to look out for.
The result is more like a win-lose, the provider gets a ton of sales and the customers get really bad performance.
More support tickets means more time spent answering them, and more employees needed to answer them, unless you want to have really long response times, which means dissatisfied customers.
If you think overselling is good, you should try a VPS from Contabo.
you forgot support. most providers oversell but the contention ratio has to be monitored to ensure stability with no extra load on support resources.
I take less compains and less support over potential profits
My time is valueable and i'd rather make less but chill my ass, then make more but have to actually do something
CPU overselling is pretty much normal in the industry, it's easy to manage and mostly provides performance advantages because most customers never use their VPS CPU at 100% all the time. Plus, it's easy to limit and monitor.
Memory overselling can work scale with tools like KSM, but it's best to try and avoid it as much as possible because it can easily lead to instability. Most customers would prefer a more stable service, even if it means paying a couple of dollars extra. It really only works if you have infrastructure and custom software that can automatically balance VMs around, or if it is extremely minor (e.g., 5-10%).
You can probably get away with a minor disk oversell because most customers don't use their disk space, but there would be disastrous consequences if a hypervisor ever hit 100% usage. This should be avoided unless you have proper monitoring or some sort of scalable storage system like Ceph where you can easily add more disks and scale.
Overselling isn’t necessarily a bad thing — yeah, just like how Titanic wasn’t really overcrowded… until it mattered.
If your dad went for milk we are starting a service very soon. MilkFinder™
In another news, this is what happens when you oversell.
https://lowendtalk.com/discussion/209030/unreal-deal-7-a-year-kvm-vps
If you don't oversell, you are in VDS business.
If you oversell, you're in VPS business.
If your company/product has "Cloud" in it, your minimum CPU steal is 70% to the point that you disable cpu steal metric(yea im talking about you, big red provider).
Now, what if the same provider uses that same host to run 200 VMs with the same specs, but drops the price to $1/month? The provider's operational costs don't significantly increase (especially with NAT), but customers will now likely buy without any hesitation.
That is fine if most of the customers are idling their VMs, but if plenty are active then you are going to find that things scale badly like that. IO contention may balloon wildly, even with high spec NVMe drives (which I'm guessing aren't the drive type in question at those prices), exacerbated by the host needing to swap as it struggles with RAM load. Also, your people's time for dealing with support tickets needs to be taken into consideration, unless you get extra people in but then you have to factor that into whether the “some fraction of 100 x $2” comparison to “a larger fraction of 200 x $1” really holds out.
I've not sold hosting for many years, but I assume this is still the same: at those prices you will pick up some of the most oddly entitled, angry, shouty, customers who expect the moon on a stick despite the low price, who will “loose thousands” and expect compensation or years worth of free service if your service is down for 3½ seconds, or you don't respond to their ticket before they've actually sent it, or you can't restore the file they deleted months ago and have no backups for elsewhere. At any price you will get people like that, but I found more of them rather than less as the price falls, and they are not worth the $1/month.
Ask contabo, they know how to do it.
Oversell to the max if price 7/year. Once this was a thing: https://www.overzold.com/
Blatant slap to the face - overzold! Level over 9000 confidence in professional overselling. Living on the edge of madness.
Welcome to LET
@contabo_m here is a potential customer of you hes called @kadidala
That isn't a sysadmin, tech support, or a professional company. Otherwise, they'd have 24/7 monitoring and would look over a 24 hour period as the first thing. Looking at the current moment in time is a data point but doesn't further issue investigation along but does point out the lack of training and monitoring by the provider.
So it would be a red flag to find a new provider.
what did I miss?