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
That's what it was before. It's now changed to 1000% and 10000 units.
wow...
Also hard set cpulimit to 100 in case it was 25 before as someone suggested.
So all of your clients were getting screwed unless this one guy complained?
What about all other clients you sold services to? Do they still get the limited cpu power until they also make a 30-post thread on LET?
Just for giggles, I decided to sysbench my own service similar to the $5 special from the LEB front page on January 6th:
`./sysbench --num-threads=1 --test=cpu --cpu-max-prime=50000 run
sysbench 0.4.12: multi-threaded system evaluation benchmark
Running the test with following options:
Number of threads: 1
Doing CPU performance benchmark
Threads started!
Done.
Maximum prime number checked in CPU test: 50000
Test execution summary:
total time: 86.9962s
total number of events: 10000
total time taken by event execution: 86.9922
per-request statistics:
min: 8.50ms
avg: 8.70ms
max: 13.86ms
approx. 95 percentile: 8.98ms
Threads fairness:
events (avg/stddev): 10000.0000/0.00
execution time (avg/stddev): 86.9922/0.00`
OK, so...
`./sysbench --num-threads=4 --test=cpu --cpu-max-prime=50000 run
sysbench 0.4.12: multi-threaded system evaluation benchmark
Running the test with following options:
Number of threads: 4
Doing CPU performance benchmark
Threads started!
Done.
Maximum prime number checked in CPU test: 50000
Test execution summary:
total time: 86.6999s
total number of events: 10000
total time taken by event execution: 346.7389
per-request statistics:
min: 11.53ms
avg: 34.67ms
max: 59.69ms
approx. 95 percentile: 40.40ms
Threads fairness:
events (avg/stddev): 2500.0000/0.71
execution time (avg/stddev): 86.6847/0.01
`
Interesting that a special for 25% of the usual price gives you.. 25% of the service. Cancelling out of general principle.
I can't get this damn formatting to work. Anyhow- I paid more than $5/mo for this deal, I got the "Self-Managed 300GB VPS" deal on this one.
Or anyone can drop me a PM with issues. No core speed was advertised and they are vCPU's, not full cores as advertised on the KVM plan from the same offer. Same 4gb plan offered for years.
One would assume that if one vCPU can obtain a certain benchmark, 4 vCPUs can at least obtain double that benchmark score, if not 4x.
The current design seems broken, and given that you admit such by altering settings for one client, why not alter them for all your clients?
CPU Governator / CPU Dynamic Scale, can be disabled in the nodes.
You shouldn't set a limit for 1000% / 10000.
Note that the solusvm's limits aren't that good, it doesn't limit properly. It sucks.
Thanks @MikePT - I will check into that.
So you got the same vps as me, but managed? Interesting.
Self-Managed. It wasn't one of the $5 specials, but the specs were pretty much precisely the same. I don't disagree with their practice of keeping any random retard from pegging the CPU, but if I am told I have 4 cores, I don't generally expect to be limited to using (the equivalent to) one of them.
I've read too fast, sorry. Well, if I order one vps with 4 cores, I want to be able to burst that 4 cores (400%). If not then I would order a vps with just one core :P
So, we're in agreement.
@Nick I suggest to have it more clear in your advertisement that a vCPU is not a full node core but a capped one. And rise a little the speed cap to something more than ~800MHz with fair share, of course. This way, you will not have any other troubles with your clients. Anyway, 5$ per month for 300GB hard disk, 5TB bandwidth and 4GB of memory, even with the speed of a single full core, couldn't be stated as a bad deal...
In any case, most of providers that do advertise vCPU, do not offer the full speed of a plain core but a capped one. If speed is critical, then, any potential byuer should ask the provider what are the hard limits of the core's speed.
Also, interestingly enough, my VPS seems to have me locked down to 1.00 CPU time, which I expect to not be as well managed for the hard bursting numerics.
top - 03:33:35 up 63 days, 4:52, 1 user, load average: 1.01, 1.05, 1.00
So typical of a CC company, its what you don't tell people that is actually important, this thread is so completely ridiculous.
Those of you who know why, know; those of you who don't, have no reason to know better or you know... work for HVH.
I am pretty shocked this is the standard we have slipped too these days.
First it was the 'dedicated' servers that turned out to be KVM VM's, now this.
Ya know, I've been toying with making lower-and-lower end VPS just to amuse myself, but so far I can't find anyone who will give me an ne2k and VNC so I can setup Lantastic.. I don't think I'd ever need more than a quarter-CPU..
When I thought about opening this thread or not I thought exactly about this, that it would become a mess and a CC bashing thread etc. It was not my intention anyhow. But thanks to me now, it as been resolved Well, thanks @teamacc in first place
There will always be CC bashing, and the reason for doing so is pretty clear in this case. Someone has to call them on their shit.
Instead of playing around with the CPU limits, they'd better spend their efforts on getting IPv6 up and running.
A bit off-topic but just curious (as an end-user), when "1 CPU core" is advertised for a KVM VPS, is there a standard percentage at which it's capped? Or could this percentage vary from provider to provider (depending on how they decide to set KVM up)? In other words, what does "1 CPU core" mean in the case of KVM? (This is intended as a general question -- not about HVH in particular.)
Not really. It's generally thought that you won't be able to monopolize what is available to you for a VPS, but not that you'll be limited to ((1/vCPUs)*vCPUs)%. OpenVZ already allows for massive overloading [roughly 4x in my experience] of resources (as compared to KVM), so doing so is an incredible dick move.
Aboveclouds.co.uk does the same thing. The cpu limit is set at 100 regardless of how many cores you order (normally two cores should be set at 200, 4 at 400 etc). So you only get one core max regardless how many you order.
Seems scandalous. Am beginning to lose faith in how providers administer OpenVZ ...