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.
Comments
Know your audience.
No no no no no. You aren't restricted to one per core. On my home rig I've got a dual core proc, and yet I run a dns server, storage server (although I really haven't transferred all my data to it yet), and a small server I use to link my network to work.
There's really hacky ways to oversell disk. RAM is easy enough to oversell, just overcommit. That's what you created the linux-swap partition for.
Just going to say.. that SSD's going to die real fast.
KSM
And probably balloning, but that thing is obvious for the client, so... no.
Who said you can't share CPU on Xen...?
It works exactly the same like with KVM. eg you can have a fast 8 core CPU (E3v2) and run as many VPS your RAM and HDD can hold. Some VPS can have 1 vCPU assigned, others 2, others 4 etc.
Weight: http://wiki.xen.org/wiki/Credit_Scheduler#Weight
Few facts because there is some absolute bollocks in this thread,
Xen uses weighted resources for CPU so you can assign your 256mb plans a weight of 100 and your 512mb plans a higher weight which means that the 512mb plan will win when it comes to a fight for cpu resource.
Xen handles CPU distribution better than any other virt type, the best way is to set everyone on the same weight which is also known as equal share.
You can oversell disk space on Xen, is almost always fucks up, solusvm makes a mess of things and it comes with a performance hit, disk IO throughput on the stock Xen 3.x + CentOS 5 that ships with SolusVM is horrible, always go for Xen 4.x and CentOS 6 if possible for a 30 - 40% increase.
You cannot oversell ram in Xen, ballooning simply takes away from peter to pay paul, the total amount allocated at any given time can never be higher than the total physical amount of ram on your host node, it is horribly unstable and the guest VPS's can disable it themselves so it is never used in hosting unless people have forgotten to disable it or have been smoking carrots during node set up's
Xen PV is faster in general than Xen HVM, but not quite as flexible and frankly with KVM who the hell would bother with Xen HVM these days.
You will hear lots of stories about OpenVZ being the fastest which simply are not true any more, OpenVZ stable is lagging behind the others now, with Xen 4.2 and 3.x kernels you get a 30 - 40% boost in overall performance and KVM is only 3% loss on native.
/rant
Either this thread is some huge elaborate joke or people don't know anything about KVM/XEN
So much misinformation
It is such a waste of disk space to allocate a lot of disk, that is not going to be used. I'm tempted just to use OpenVZ because of this. I am assuming due to how Xen creates a virtual disk for every VPS it can not be oversold? But OpenVZ has a poor reputation
You're going in circles.
If OpenVz has a poor reputation, it's because of the ease with which irresponsible or ignorant providers can oversell, to the detriment of performance.
Now you're asking how to do the same thing with Xen. If that was possible, in the same fashion as OpenVz, where do you think it would lead?
Disk space really should not be a concern these days, 4 x 2TB Disks + 32GB p/node gives a very healthy balance. which is why I LOL every time I see the stupid 50GB of space with 2GB plans... oh.. I see what you did there
@ShardHost you hit the nail on the head. it was a joke.
We probably have 1TB spare per node, more in some cases but who likes a jam packed server. I'd rather not have to migrate a server with 2TB (~1.8TB in reality) space used.
Filling hard drives to the max on a production server, unless it's a storage server is just asking for trouble.
So you don't give much space just in case you need to migrate at some point and everyone has actually used all the space you gave them?
But serial though? haha.
+1 @AnthonySmith thank you for the long post. I have a deep palm imprint on my forehead up until that point.
How do you think it looks to the community when someone opens thread after thread asking for help on overselling services?
Like he does not have the first idea what he is doing and is not even capable of googling it?
@AnthonySmith If a array is in a bad state then migrations are often the only option.
But migrations can mean downtime, in some cases, even more so if you have VPSes with huge amounts of data.
I don't think that the person in question is stupid by any means. Don't get that impression. I mean, he is running a hosting company with at least a few clients. I just think that he is misguided in his attempts to expand his business; buying the absolute cheapest dedi's out there and trying to oversubscribe them to make more profit.
So you expect me to believe you base your entire business model around the fact that at some point you might have to migrate and the difference between migrating 1TB and 2TB is enough for you to offer less disk space globally, while at the same time in another thread offering 150GB for $7 p/month
Seriously I was not born yesterday, I am not stupid.
@AnthonySmith That's called balancing resources.
I wouldn't put 60 VPSes w/ 150GB Disk to a single node, where as I could put 60 VPSes w/ 30GB Disk on a node without a worry and performance would be like new.
Bandwidth & Disk space are not oversold, where as the RAM is.
So, after reading this thread, I still don't have a solid answer as to whether CPU cores can be shared in Xen and whether disk space can be oversold in Xen. Every other post conflicts with the previous.
Note this post has nothing to do with ShoveHost, ShoveHost will never offer Xen VPS and always just OpenVZ VPS. So don't let it reflect badly on that. Thank you.
@shovenose
http://wiki.xen.org/wiki/Credit_Scheduler
And what is this???
Honestly, setup your own node and find the answers yourself.
@tchen yeah I read that, didn't really answer the question.
Not it can not, 1 VPS per core
Xen inside esx and you can oversell as much ram/cpu as you like.
Everyone gets 2 cores and 2 gig of memory. All 256 of them on the a dual core machine with 4g of ram.
Well then screw it, not using Xen
thanks for the help everybody, really appreciate it.
What's the purpose of a cpu scheduler if VMs are exclusively pinned? @spencer why are you so adamant that its core-bound?
BTW, search 'spare file' for your topic on thin provisioning.
sparse, you mean?
This reminds me of the time I went to the dentist and he walked out into the waiting room and asked if anyone knew how to remove a tooth... oh wait.
Haha yes. Thanks for catching it. -embarassed-
c'mon stop messing with him :P
I think the funniest part is @Shovenose in his signature says he owns bettervps, which supposedly sells Xen plans.
EDIT: He must have saw this and removed it.
Here is a screenshot before he removed it:
http://gyazo.com/590e02d0ceb32e613ad2254eb7ec0b00