I don't see how a cloud provider can complain about high CPU-usage if they bill by the machine and by the minute. Otherwise 20% of your bill is for idle CPUs.
I don't see how a cloud provider can complain about high CPU-usage if they bill by the machine and by the minute. Otherwise 20% of your bill is for idle CPUs.
It really boils down to the business impact you have on them. Sure, you pay for the resource, but they are not guaranteed. And it is up to their discretion to de-prioritize your VMs on the HV, so that you're the last to get scheduled when the host's CPU is starved. This is the whole 'CPU credits' thing on AWS that they document in [0] where you can burst above the baseline. Interestingly, note that the baseline is in the 30% ballpark in their illustration on the bottom of the page :)
[0]: https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/burstabl...
It's also fraud because AWS get to charge me when I'm not scheduled, but they also charge the users who are scheduled on the contended CPU. They should bill by the CPU cycles I consume.
Anyway, we've moved from AWS. Lucky escape.
> Traditional Amazon EC2 instance types provide fixed CPU resources, while burstable performance instances provide a baseline level of CPU utilization with the ability to burst CPU utilization above the baseline level.
If the provider you are working for is doing this crap, then it sucks.