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.
How well are resources governed in VPSes?
raindog308
Administrator, Veteran
The recent thread about a customer who generated a load of 65 made me wonder how well resources are governed in a VPS. I'm thinking of common technologies like OvZ, Xen, KVM.
For disk - I don't think there's anything to guarantee or cap I/O, is there?
For memory - it's either there or it's not, right? You can oversell but ultimately if you give a VPS 512MB, that's all it can use (leaving out burst memory, swap, etc.)
So that leaves CPU. If you have an 8-core box and you give a VPS 1 core, they can max out a core's worth, but that's the limit of what they can do, right? So in theory I could be running 1 core at 100% but the other would be sitting there idle or in use by other VPSes?
Just wondering how tall the inter-VPS walls are.
Comments
I was wondering as well. Anyone care to share?
In KVM yes, in OpenVZ there is an ioprio option
Well, we already know lol
Hmmm, most providers give that "fair share" based in a "cpuunits" param, I guess everybody gets "1000", so is fair.
Some providers have set "cpulimit" to give a fixed percentage of CPU. That's OpenVZ
For KVM, afaik the cpu scheduling just sucks.
For xen apparently you have fine control too
Isn't it also possible to limit your container to a certain percentage of a core?
Edit: too late ;(
Linux cgroups are used by savvy providers to partition host resources to client KVMs (example: OpenITC): https://www.kernel.org/doc/Documentation/cgroups/cgroups.txt
It´s possible to limit cpu in kvm too, but not implemented in solusvm (only in beta). http://libvirt.org/formatdomain.html#elementsCPUAllocation
You can use cgroups with RHEL6-based VZ kernels.
They do that?
Very interesting information BTW, bookmarked
I thought KVM is already using cgroups to allocate and cap things.