Providers; do you factor in 'idlers' when planning your node's capacity?
It's no secret that 'we' love to hoard Low End Boxes and thus often have a collection of idlers. I was wondering if providers factor these in while planning their node's capacity?
Comments
Of course they do that, likely goes into how much can be oversell number, but I doubt that anyone here is going to say these public.
Idlers are a fucking money grab, you buy stuff and don't use it.
Free NAT KVM | Free NAT LXC | Bobr
ITS WEDNESDAY MY DUDES
Not really because usually the node limit is decided by the amount of available disk and ram, not cpu. And an idle vps is using both.
This.
even if the VPS is thin provisioned (not reserving the full amount of disk at deployment) there is a lot of factors to calculate.
For me, I estimate the number of packages that I can "fit in" on the amount of disk and RAM available.
Then this number is adjusted along the way as services are cancelled
“Technology is best when it brings people together.” – Matt Mullenweg
I'm sure SSDNodes has a solution to this.
I bench YABS 24/7/365 unless it's a leap year.
For OVZ NAT, ofc I factor idling, no. of vps is calculated based on available disk space & bandwidth.
For KVM, it is different cuz the idling ones too occupy ram...so not much 'idling' is taken into consideration here.
https://webhorizon.net
Any idling VM can become an active VM, so that'd be another thing to factor in.
I've had my share of idling servers that I decided to go use after a few months. So, yeah.
The answer is yes but they/we shouldn't.
Michael from DragonWebHost & OnePoundEmail
We need a special idler product.
Idle VPS
Active VPS
Collectors can buy the idler product, run YABS, and let it idle.
UptimeRobot would still think it's online, and it can even serve websites
via reverse proxy, but it consumes very little CPU and RAM for the reverse proxy.
Accepting submissions for IPv6 less than /64 Hall of Incompetence.
anynode already has 2GB RAM VPS for $15/yr.
I bench YABS 24/7/365 unless it's a leap year.
It used to be that 80% of users are basically idlers. This rule holds true still if your target audience is the general people.
But that changes when you enter lowend. These people demand a lot for a penny. Also avoid anyone who asks for dedicated cores/threads. They ain't worth the trouble.
♻ Amitz day is October 21.
♻ Join Nigh sect by adopting my avatar. Let us spread the joys of the end.
We don't count idlers as all our KVM VPS nodes have a hard limit set for RAM as well as disk space to match the server specs usually 90%-95% of RAM And Disk space as we have to even consider possibility of client asking for upgrade in future
So if a VPS node has 128 G RAM, we sell at the most VPS with total of 120 GB RAM, same for disk space
hostEONS.com - 9 Locations for VPS (US and EU).. Free Blesta License, Free Windows 2019 License and more. Shared DA Hosting. AS142036 .. Ryzen 7950X Based VDS in SLC, LA and Dallas
No sir. We don't. That's exactly why we publish our limits within the TOS. We do try our best to keep things at 90%, so you know, buffer room.
Nexus Bytes Ryzen Powered NVMe VPS | NYC|Miami|LA|London|Netherlands| Singapore|Tokyo
Storage VPS | LiteSpeed Powered Web Hosting + SSH access | Switcher Special |
Interesting, thanks for all the replies. I never knew the node's capacity would be limited by it's RAM and/or Storage. I always figured the CPU was the bottleneck, since 24 threads seems a lot of more restrictive than 256GB of RAM (fictive numbers of course).
LinuxFreek.com
Indeed, line 135
https://github.com/Ne00n/DealsLists/blob/main/data/lowendtalk-available-posts.json
Free NAT KVM | Free NAT LXC | Bobr
ITS WEDNESDAY MY DUDES
No. We only plan capacity based on allocated resources regardless if it's in use or not.
Also, KVM will use a little more RAM than your allocation, and control-plane software also requires RAM, so we usually only allocate < 90% of RAM to customers. For disk space, same reason :-)
Misaka.io | Blazing fast AnyCast DNS with 60+ POPs GeoDNS, AXFR, DNSSEC supported.
And Reliable high-performance virtual server | Ashburn, New York, Seattle, San Jose, Hong Kong, Tokyo, Singapore, São Paulo, Johannesburg
ping.sx | Ping any server from global locations in parallel
Surprised people are saying disk is a limiting factor. I would have thought that's the one thing you can scale fairly independently of the rest of the server as needed. Adding more CPU to a mobo could be challenging though
iirc the vm's disk size for the host is different than what the users are actually using. e.g. if someone does a luks encryption, it takes up all the available space. And if someone downloaded 100GB of files and deleted 90GBs of it, the disk itself would still use 100GB, unless you do some kind of garbage collection.
CPU will be a limiting factor when you sell dedicated cores.
In a fair share environment, the cpu usage is not that high.
“Technology is best when it brings people together.” – Matt Mullenweg
exactly.
https://webhorizon.net
Yeah I get that - can't be shared in the same way as mem/cpu from a technical perspective.
Just not what I would have expected from a financial analysis perspective if you're looking to maximize profits (Theory of constraints and throughput accounting and CVP analysis specifically).
Then again I've never run a VPS provider....
I have no idea
Disk can be added, disk speed cannot.
From my personal experience, disk io was often the bottleneck. Outdated second hand cpu and ram is pretty cheap, but disk must be new. In the age of nvme, I rarely see providers crappy in performance anymore. More often it is providers deadpool because their cannot paid their bill.
Action and Reaction in history
haha. I studied finance not IT so sometimes come at things from a different angle. @elliotc explanation makes sense though