Talk about CPU usage 100% again.

I used to hate those ops. When I saw CPU usage 100% , I shouted wildly, and I once said, if the CPU is not used, is it used for idle?

It seems to me that what I said is quite reasonable. But, I was wrong ...

But that doesn't mean that OPS is right, and that after 100% of the CPU they scream wildly, most of them feel like they're on the job, not on a sense of responsibility in order to find and solve the problem (trust me, please). Why else would they always leave things to research instead of trying to check them out?

The modern computer based on CPU is a typical m/m/1 queueing system, and the CPU is the service desk. and pending's task is to queue entities.

I initially thought that CPU usage was 100% and that the CPU was being fully utilized, but I overlooked what might happen next, and the next situation was very dangerous. As a typical queueing system, the task arrives in accordance with the typical poisson distribution , while the queue time of the task conforms to the exponential distribution , which must be learned after the queuing theory.

CPU usage 100% means that tasks under the Poisson distribution continue to arrive in accordance with the probability distribution, and if the CPU does not sell some idle time, the exponential distribution means that the new arrival queue time will reach infinity, resulting in a memory overflow. Therefore, according to the exponential distribution of queuing time constraints, CPU utilization 100% can only maintain a short period of time, if the CPU lasts 100%, it will inevitably lead to a large number of tasks hang not responding, this is what DDoS attackers want to achieve.

There is no formula here, in short, the CPU will continue to be 100% of the total problem, otherwise, according to the mission arrival rate Poisson distribution does not make sense.

2017, goodbye.