Looking for a small KVM VPS

ZuckZwing

As title suggests, I'm looking for a small KVM VPS for idling a side project.

Budget: Less than USD 20 / EUR 20 (yearly)

Spec:

• RAM: At least 2GB RAM with 1 GB Swap
• Disk: At least 15GB (SSD, NVMe)
• CPU: At least 2 shared cores (e5 or better)
• Net: At least 1TB BW, 1 IPv4, /64 IPv6
• Good to have: Non-US location, reliable provider, decent customer support

Thanks!

lapua

20€/month or year?

ZuckZwing

@lapua said:
20€/month or year?

Yearly. Thanks for noticing. Edited!

sh97

Take a look at https://lowend-deals.redbull.ink/ (some links are aff) for some deals under 15.
It would be hard to find 2 cores under 20$/yr.

If you can increase your budget to $25, you can get Greencloud VPS (Budget KVM-2)
Ticks all your requirements, and as a bonus some locations even have EPYC.

Edit: if US works - [RN cart link redacted]

ZuckZwing

@sh97 said:
Take a look at https://lowend-deals.redbull.ink/ (some links are aff) for some deals under 15.
It would be hard to find 2 cores under 20$/yr.

If you can increase your budget to $25, you can get Greencloud VPS (Budget KVM-2)
Ticks all your requirements, and as a bonus some locations even have EPYC.

Edit: if US works - [RN cart link redacted]

Just for curiosity, why is it 2 core is hard to find? Also it's not a hard requirement I can do 1 core as w

somik

@ZuckZwing Try this: https://www.serverhunter.com/#query=cpu_cores:>=2+memory_amount:>=2048+storage_capacity:>=15+traffic:>=1024

Customize it to your needs. Once you choose a provider, look up it's review over on this forum or just ask about it. Most people here will know most providers offering servers on that server hunter website. I also use it to find deals or plans.

crunchbits

@ZuckZwing said:

@sh97 said:
Take a look at https://lowend-deals.redbull.ink/ (some links are aff) for some deals under 15.
It would be hard to find 2 cores under 20$/yr.

If you can increase your budget to $25, you can get Greencloud VPS (Budget KVM-2)
Ticks all your requirements, and as a bonus some locations even have EPYC.

Edit: if US works - [RN cart link redacted]

Just for curiosity, why is it 2 core is hard to find? Also it's not a hard requirement I can do 1 core as w

Unless the provider is grossly overselling, that is usually the first resource that you'll (reasonably) exhaust. RAM, bandwidth, and to a degree disk size/performance can all scale a lot further than CPU cores can (per chassis). If you're getting more CPU cores it likely has restrictions that become hard to enforce/follow, and I think that most hosts don't want to bother with it at the lower price ranges. The admin time/support effort to diagnose and handle abuse eats away at the profit too much, and it is very possible a good customer was just unaware that they were over their 20% allotment on 2vCPU.

smart2host

Do you have a favorite location?

Patriarch

Maybe our homies can help you? @Kuroit @SpeedyPage @Hostaris @Labze @Calin

yoursunny

@crunchbits said:

@ZuckZwing said:
Just for curiosity, why is it 2 core is hard to find? Also it's not a hard requirement I can do 1 core as w

Unless the provider is grossly overselling, that is usually the first resource that you'll (reasonably) exhaust. RAM, bandwidth, and to a degree disk size/performance can all scale a lot further than CPU cores can (per chassis). If you're getting more CPU cores it likely has restrictions that become hard to enforce/follow, and I think that most hosts don't want to bother with it at the lower price ranges.

$11.69/year 2 vCPU, how grossly is the overselling?

The admin time/support effort to diagnose and handle abuse eats away at the profit too much, and it is very possible a good customer was just unaware that they were over their 20% allotment on 2vCPU.

Good customer like me proactively monitors CPU usage and sets proper limits in systemd or Docker.
A big prerequisite is that the provider must disclose the permissible CPU quota in terms of percentage, instead of insisting on "fair use" and "we will contact you when you are using too much".

FrankZ

@yoursunny said: $11.69/year 2 vCPU, how grossly is the overselling?

I do agree with you that good customers proactively monitor or restrict their usage to be within acceptable limits.

EDIT: Checking my stats shows an average 0.012 steal & 0.00 wait state, without any spiking, over the last 48 hours, on 3 minute checks. My early conclusion would be that the $11.69/year VM I bought from CrunchBits does not look oversold. Just saying you can't make assumptions based on price alone.

crunchbits

@yoursunny said:

@crunchbits said:

@ZuckZwing said:
Just for curiosity, why is it 2 core is hard to find? Also it's not a hard requirement I can do 1 core as w

Unless the provider is grossly overselling, that is usually the first resource that you'll (reasonably) exhaust. RAM, bandwidth, and to a degree disk size/performance can all scale a lot further than CPU cores can (per chassis). If you're getting more CPU cores it likely has restrictions that become hard to enforce/follow, and I think that most hosts don't want to bother with it at the lower price ranges.

$11.69/year 2 vCPU, how grossly is the overselling?

The admin time/support effort to diagnose and handle abuse eats away at the profit too much, and it is very possible a good customer was just unaware that they were over their 20% allotment on 2vCPU.

Good customer like me proactively monitors CPU usage and sets proper limits in systemd or Docker.
A big prerequisite is that the provider must disclose the permissible CPU quota in terms of percentage, instead of insisting on "fair use" and "we will contact you when you are using too much".

The post you are replying to from me was before the $11.69 one-off, one-time sale. CPU usage was further clarified and outlined in OGF original thread specifically for this one product. Not a standard product. Not our standard deployment. However, with people being reasonable no single yearly hypervisor has maintained >15% average load. Especially once sign-ups were restricted to avoid the outsiders. I'd say the overselling is not gross, but it's still not our standard operating procedure. I have to pick and choose our battles as system and network admin time is not unlimited. Worrying about and managing resource usage from our standard product lineup is not something I find an acceptable use of time so we simply just don't fill those hypervisors up to a point where it is even generally possible to become a problem. Customers are happy and we're not completely drowning in busy work. Lower margins per watt specific to those products, but our deployment envelope is not comparable to a lot of other hosts.

In our AUP physical system resources are treated relatively broadly for a reason, and outside of the 1 accidental case* on these yearly hypervisors we've never once had an issue regarding physical resource consumption and a suspension with standard product lineups. I could add language to further arbitrarily limit them, but that seems like it would be a step towards positioning and forcing us into the 'grossly over-sold' side of the meter which is not my goal. I understand the other side of it though: I've seen hosts hide behind broad language to justify questionable actions regarding policy implementation.

Because the stats are usually kind of fun to share (I'm guessing it will look similar on a lot of other hosts here too) this is a snapshot into some of our Ryzen 5950X machines: busiest / average / lowest

All 3 are completely full/sold out and have been for months.

* = Only 1 account has been flagged over the course of about 2 weeks for CPU use at an unreasonable amount, and it was just an error on their part and not intentional abuse. Enough resources were available on the hypervisor to not affect anyone else in the interim.

crunchbits

@FrankZ said:

@yoursunny said: $11.69/year 2 vCPU, how grossly is the overselling?

I do agree with you that good customers proactively monitor or restrict their usage to be within acceptable limits.

EDIT: Checking my stats shows an average 0.012 steal & 0.00 wait state, without any spiking, over the last 48 hours, on 3 minute checks. My early conclusion would be that the $11.69/year VM I bought from CrunchBits does not look oversold. Just saying you can't make assumptions based on price alone.

Generally speaking, it shouldn't be an oversold experience. Given that there are more people on those hypervisors than normal there is the potential for some issues and maybe re-balancing. There is a reason a completely empty hot-spare node exists: everyone can check their hypervisor tags, nobody has 47-DN-U08.

Aside from it not being a standardized product, there's a reason the ones who got their yearly's are the only ones. Not something we'll continue to sell. It ended when the LES thread ended. I tried to differentiate it as best I could and explain it was a one-off, one-time special product / offer back to the communities which I'm glad exist. We've had almost exclusively positive feedback, and I think people are very realistic and understand what the product is at that price-point.