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.
When do you recommend going with VDS instead of VPS?
FrankCastle
Member
in General
I am curious to hear what others think is a good threshold to switch from a VPS to a VDS solution? Game servers, crypto mining, AI, compiling the latest Linux kernel non-stop, looped YABS runs, what are some of the more popular reasons people go with a VDS? If you are currently looking for a VDS please share what you plan to use it for.
Thanked by 110thHouse
Comments
...professional idling?
jokes apart,
when something uses constant CPU(like production website/Windows Server/Desktops)
Crypto mining, what else
Hah, I should have included professional idling. That was my oversight.
With how cheap you can get a VPS for might it not make more sense to load balance or split the load on that production website that is so busy where you can fit it onto a couple VPS instances instead of having to buy a VDS for it? Unless you find a really good deal on a VDS anyway. Desktop usage is an interesting one, I had not thought about that. That's why I created this thread. Good to hear what people are using them for. Thanks!
I mentioned crypto mining already. You need to be more like @FatGrizzly and provide something unique.
Is it ok to use constant CPU up to 70% on a VPS?
Not likely, I think most VPS solutions prefer you keep it under 50% at the highest for extended periods but YMMV. Short burts to 70% and above is fine but constant is when I think it would make sense to switch to VDS. What are you doing to keep a constant CPU load of 70%?
video encoding or streaming site
Mine some unique crypto then
I am running Azuracast with 4 stations (for only personal and family use) It only uses 25% CPU. I just wonder how much more CPU I can use from it to run some more applications. It's not worth using that VPS only for Azuracast.
You should check with your VPS provider and see what they consider fair use for a VPS to be certain. Depending on their response you can decide how much more you are able to push it.
Har har Mr. Funny Man!
Yes, I thought it also, I have to check it with my VPS provider first, thanks.
People prefer VDS for stable performance, avoiding drops caused by busy neighbors on shared VPS resources.
Personally, when constantly high CPU usage is expected, I prefer physical servers. VDS while useful for similar high loads, are more expensive but offer nice things as live migration (no hassles with maintenance of hardware - your hosting handles it for you transparently), snapshots available in providers' dashboard (again - no hassles for you implementing and maintaining it yourself) and autoscaling.
Hi,
in our offers we currently define a threshold of 30% average permanent usage to be the limit for VPS usage. Beyond this the customer should take a VDS ( 100% permanent usage allowed ).
This 30% is the maximum average value that is permanently used within the last 7 days. You can still peak to 100%, even for multiple days and still wont suffer any CPU power limitations.
But as soon as you used at least 30% permanently in average over the last 7 days, you will be limited, until your average is under the 30% threshold.
The normal average VPS use permanently < 10% of its CPU power. So even you are using it as a gameserver to play in the evening for 3-4h with your friends using 100% CPU power, you wont run into any issues with this 30% limit.
So this 30% really just filter out the powerusers who likes to pay a small price but consume resources in a dedicated way and this way ruine the offer/experience for the 95% of normal users.
So in general: VDS you will need if you want to ensure 100% resources availbility for you 24/7/365
( No matter what for )
Wonderful explanation, thank you for the great detail.
I think some providers here on LET allow descent amount ~ 75% usage if it is a legitimate reason because that 100% is already a soft limit on a vps.
Hi,
yes of course. Every provider decide how much usage he/she wants to allow in a shared environment. Depending on how much brave the provider is and how much he/she will overbook the physical system things will work more or less good/bad or explode.
For us, since we start as low as 2,xx EUR per month ( yearly payment ) for a vps with 2 cores, allowing 75% permanent usage would mean that we would have a total income of ~80-90 EUR for a physical host of 80 Cores, 320 GB RAM, 3.2 TB redundant NVMe space and tons of traffic.
You dont need to have studied math to see that this wont do anything good economically to the provider.
So we dont allow 75% but 30% permanent usage to be able to maintain this very small price with a good bunch of resources while the company wont die economically and this way serve this offer reliably.
But thats of course all the decision of the provider how much money will be asked for how many resources.
Yes, I agree with you the price also should be reasonable, otherwise it's not profitable for providers. As a user I expect 70% constant usage should be allowed (for normal/personal things) when I buy a VPS for more than 4 EUR/m.
I would often take a dedicated server over a VDS because there is no guessing game on whether x provider actually gives you dedicated cores or not. To be honest, I’ve never had issues with a vps provider for using too much cpu.
With a dedicated server, you get resources all to yourself. There are plenty of cost effective dedicated servers from providers like Hetzner, Pebblehost, Oneprovider, Data Ideas, PulsedMedia, ReliableSite
For reference regarding VDS and dedicated cores:
https://lowendtalk.com/discussion/195101/avoid-avoro-eu-php-friends-and-dataforest-oversold-root-servers-and-fraudsters/p1
VDS / VPS.
interesting to read. @layer7 THX on the explanation, now I know that all of our products are VDS, as we allow a 75% average in 24H.
I personally do not agree on the 30% or 50% or anything below 70%, as in that case it is not a Virtual Private Server, it is a OVPS - Overselled Virtual Private Server and should be marketed and sold like that.
Hi,
i hope that if you talk about "average in 24H" you actually mean a permanent 75% usage. Or what do you do after the 24H? What do you do if a customer permanently consume 75% of the CPU ( which is in fact a dedicated usage ).
I did not check the pricing or your size or how long you have been on the market. So i can not judge how much long term real world experience you might have or not.
But from my experience, you have always some customers who are abusing the offers ( the cheaper, the higher the chance of abusive usage ). And either you will somehow limit this customers or get somehow rid of them or move them all to a bad-guys-host orr simply loose money. And as soon as you have enough customers, this 5% will become bigger and bigger and you will loose more and more money or your customers will suffer from bad experience.
I want to avoid that normal customers will have to pay for this 5% abusive users or will have bad cpu experience. Thats why we define this limit that, in combination with a quiet low price usually not harm the normal users.
So the normal users can enjoy small prices and good performance.
That has actually nothing to do with overbooking. Its actually the opposite.
Yes... of course you need to work with providers who will actually deliver what they promise. Thats a usual and common problem unfortunately. Finding the right partner might take some trial and error's before you hit a good one.
Goes also for providers who rent in datacenters, buying hardware or buying traffic from carriers....
I love all this behind the scenes shop talk. As a customer of many different providers you see the impact each of their decisions on how they priortize balancing their servers impacts us (the customers). You sign up with some providers where cpu steal is high and you know they just cram as many customers on as they can to maximize profits, to hell with performance. You also might run across a provider (seems more rare) where performance seems much higher than their competitors. Those are the ones that REALLY know what they are doing and definitely stand out. Most that I've come across are in the middle. They are a business so of course they are interested in making money so they try to balance the amount of customers they can stack but they do keep a close eye on the overall performance so it doesn't drop too much.
So thank you @layer7 for going into such detail and being open and transparent with how you tune the dials. I wish all providers were as open. It would be awesome if a provider created a dashboard of the servers so that any customer could see how many other users and how the resources were divided up amongst everyone. Probably not worth the time since most customers probably don't care or aren't even interested in it as long as it works. But for those like me, that would be super cool.
From the provider side, how do they detect is the VPS is used for crypto mining or just because of high website traffic when CPU usage is always high?
Yes, I meant a 75% average in 24H. 100% in 24H will get flagged as abuse, and we contact the customer via a ticket and go from there.
Those the not comply or we cannot find a middle way ( either get an upgrade to more vCores or fix the issue in the VPS ) we usually part ways.
Bare in mind that we sell primary Storage Services, so the number of customers / node is low, by this all of them can get the dedicated vCore assignment and do their stuff.
If the customer consistently uses 50-75% of resources, we have no issue with it. Those purchasing 5-10TB or more of storage typically use it for backups, which often involve compression, consuming CPU as the process runs intensely. The same applies to video streaming platforms like PLEX or similar services, where CPU usage spikes during heavy tasks. We often notice these spikes reaching 100% usage for 3-5 hours per day, for example.
And yes, there’s always that 5% who will abuse the service. You see it right after the purchase—someone buys the service and within ~15 minutes, 100% CPU Usage + they’re sending 25K SSH requests per second to random IPs, in this case, we file a divorce on the spot.
Again, thank you for breaking down the logic in your first reply—it really helped us a lot.
Just check your CPU usage from the past 24 hours, compare it to the last week's 95th percentile usage, and bingo, it's that simple to identify patterns. - Yes is that simple, no one would mine crypto with 50% of CPU, but hey, I might be wrong
With regards to CPU usage... It depends on the provider... Some don't allow constant usage of 25% & above, some 50% & above... If you feel you will be breaching these levels set by provider, then either go for VDS or VPS with higer specs...
Precisely. Before going with a provider for higher CPU usage, it's a good idea to ask if that’s acceptable or to carefully read their Terms of Service to ensure compliance. - This approach saves a lot of headaches for both parties involved
Oracle says if I don't use their CPU above 10% they will block my free OCI account.
When I either expect frequent peaks for more than a short period of time - or - when I want peace of mind, typically with a business related server I use a VDS (otherwise a VPS).
Re VDS/VPS per se I see two major factors, (a) WTH is "normal usage"? With some it's 75% avg, with most I guess it's 50% avg [over whatever period of time. 24 hrs seems to be common] but 30% is more common than one would like to think, and even lower isn't exactly super rare.
(b) a VPS core that is, a HWT, is not necessarily 50% of a physical core! And btw. actually neither a VDS core necessarily is. A 'vCPU' is whatever the provider wants it to be. The actually relevant question is "how many vCPUs do you have per physical core?" - and rarely gets answered properly.
Explanation: A provider can (with most hypervisors) create pretty much any number of vCPUs they want. THAT is a, if not the, major ingredient for the "secret sauce" of a provider. And of course many (prob. most) providers are very happy with most users/customers assumption that 1 vCPU ~ 1 HWT.
In reality however it seems that one should be quite happy if one's VPS runs on a host with "only" 4 vCPUs per physical core and 1 vCPU ~ 1 HWT already is luxury territory while even 8 vCPUs per phys. core isn't exactly rare.
Simple rule of thumb (but by no means a reliable fact!): A VDS from a reputable provider, and not cheap, tends to be 1 vCPU ~ 1 HWT. An "expensive" (for LET standards) quality VPS tends to be 2 vCPUs ~ 1 HWT or thereabout. With a $25/yr VPS you probably are quite lucky if it's "only" 4 vCPUs ~ 1 HWT.
That said, I personally don't care too much as I don't run stuff like mining, video en/decoding and the likes.