Each resource in any system is supposed to be working at 100% efficiency, unless they are bound to some lower limit due to restrictions. CPUs are supposed to run at a maximum of 100% at nodes.
CPU usage is reported out of 100% for each core, so the total amount of CPU power to be used in a dual core system would be 200%, and a quad core, 400%, depending on cores in the CPU. So, when we have multiple CPUs in a system, how should we calculate CPU busy percentage across all the nodes?
To calculate how busy the CPUs are within a node, we need to normalize the CPU over the node level; for this, we add together all the CPUs in a node, and then divided by the number of CPUs in the node, to arrive at a number with a maximum value ...