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
Load average can exceed your number of cores. There's little you can do to prevent it entirely. Little spikes are alright, you'd hate to place a blanket rule that says "You have 1 CPU core therefore your load can never exceed 1 or your container is auto-suspended."
Just like I only have 12 cores, but tell that to the load average of 40+ (that's not a real number, my load averages are all under 2, if not under 1).
Here's an explanation of load: http://blog.scoutapp.com/articles/2009/07/31/understanding-load-averages
Edit: Don wins.
@Mun Roger. I want to be schooled on this if ya'll don't mind..So how does a virtual machine do that..
I just thought that wasn't possible to affect the node that way. Your VM has a virtual kernel doesn't it etc?
Just help me and I'll let it go Haha thanks @Mun!
Roger didn't know that to be honest. Especially for a VM.
I need to read up on virtualization a bit I wasn't aware that actual cores were allocated per VM and all
Thought even the "cores" a virtual machine got were split from an actual one so it was one to many and not 1:1 etc
Um.. depends on virtualization technology?
Also, just because your VM is virtualized doesn't mean it won't affect the node.
I can die happy.
Aright well in that case...the guy's a douche!
@Mun I dub thee Mun the mentor. I am nate, learner of great wizards Jarland and friends.
@natestamm basically all the load still needs to be transferred from the vm to the host node where it is then processed.
In this case however the local kernel on a said VM will look at what it has available (i.e. a 4 core instance) and see how much you alone are using. I.e. a core pegged at 100% would read ~1.00.
Now the main node will read for all users and could be much higher. Like ~10.00 if there was ten cores worth of load by 10 different vms all pegging 1 core at 100%.
Simply put, having it on a vm doesn't create more umpph for the server. It just isolates instances with a slight cost for overhead.
OpenVZ isn't real virtualization,so its recourse separation is not as strong as Xen, KVM etc.
Well that's kinda why I said this before hand too:
Although meh, it was a half-thought-out post.
So the answer to all this is to avoid selling vps's at LEB price range and offer undersold premium vps's to clients?
/jk hhh
I think people should be able to load a 3GB sql dump on any VPS service that is paid. But that's me.
" $6 VPS with us and literally felt obligated to be able to use 100% of the entire system's CPU "
i really hate providers with the above statement because i had a bad experience like this sometime ago , a user pays for vps resources, he should be able to use it to 100% what the resources allocation is if you have a problem with this go back to selling shard hosting. if you have problems with cpu issues on your node, it just means your overselling and putting to many users on a single disk and expect users to limit a resources they have paid for. blueVM, I think you need to rethink what a private VPS resources is.
I got nothing to do with the guy or blueVM, the above is just my personal opinion, I just hope providers that limit cpu do fail and get out of the market
Name 1 LEB provider who does not (or will not in case of abuse) limit CPU.
It looks like you're mistaking LEB VPS for dedi.
You just demonstrated how little knowledge you have on this matter. If you want to use 100% go back to buying a dedicated server, not a VPS.
If you seriously think buying a VPS on a E3-1230 server for $7 or w/e entitles you to use the entire processor you're living in a dream world.
@kandosan sounds like you want to have all cores assigned to you on a single VM and because it's assigned to you (SHARED) you and only you should be able to use it all?
will cpu throttling help, to say give the guy just 100 or 200 mhz and let him use that all he wants?
In that case what would he use it for? 100mhz would be slow.........slow............slow............ wait... let me try it on myself
yes it's slow, but he deserves it right?
@jcaleb Nah, it's counter productive to screw with such customers. The best course of action is to hit "cancel & refund" and wish them luck with their new host.
@NickO, use what you pay for that's what i meant, if you pay for 1 or 2 cores, I am expecting to use them at 100% i don't want to hear my provider complain that I'm using 2 cores at 100% its what i paid for.
@dnom , yes should go with dedi for future projects
You are purchasing the product that has "Fair share" cpu cores. You are not buying dedicated cores.
Oktay
well its fun to think that can be applied for this matter. but yes, in business must be practical. no emotions attached
yeah, most providers allows this. those who don't are bad companies.
I think this is the same confusion that my customer had.
A load of 100 does not mean he uses 100% of the resources, like a load of 65 does not mean he uses 65%.
It means he would need the equivalent of 100 or 65 cores to clear the backlog of instruction on his CPU.
A customer with that load must be shutdown or throttled to protect other customers if the server has few cores, or example, an E3 has 8 logical cores, if the customer has 4 vcpu and locks them up with loads like those (practically he will use them exclusively, OVZ does not allow a correct separation and time sharing) then for all the other customers only 4 cores remain, logical cores, for that matter.
That is not how shared resources should work, someone uses half the server and the rest should be happy with the crums.
The abusive customer must be removed or throttled, otherwise there wont be a "fair share" of the resources.
Even on servers with many cores, locking up some will create unneeded overhead and will be a few cores less for the other people.
You can have a load of 100 on your dedicated server, there nobody else will suffer if that happens.
It still blows my mind that anyone could even manage to generate a load of 65.
damn
Hahaha, just skimmed the logs, what an asshole.
Doubly so because he's being so petty about this
Had I been in that IRC, I would have been dogpiling on him too and suggesting he go somewhere else (he won't, because he's cheap)
Andre Lopes sounds mentally ill, unfortunately that's who you have to deal with when you deal with the general public.
You're right to post about this fwit.
You don't understand VPSes. You're never offered 1 or 2 cores, you're offered a share of 1 or 2 cores. Because CPU usage tends to be used in bursts, allowing a VPS to use most of the CPU resources temporarily is a sensible arrangement. The problem occurs when you get dickheads like Andre.
Why is it that in the LEB market you are given a "share" without limits and can abuse it, but when you're paying 15$+ you often have hard limits such as 250MHz, 300MHz, 500MHz, etc.
Probably because the low price comes from overselling, so people expect that they can burst when they need to but frequently will not be using a lot of CPU