CPU saturation
also: load saturation, CPU overload
CPU saturation occurs when the number of runnable processes exceeds the number of available CPU cores, causing processes to wait in the scheduler queue. It indicates the system cannot handle all computational demands immediately.
CPU saturation is a performance state where the CPU(s) are fully utilized and additional processes must wait their turn to execute. It's measured by the load average relative to the number of CPU cores—a load average of 4.0 on a 4-core system means saturation, while 8.0 on that same system means severe saturation with queued work.
When saturated, context switching overhead increases as the kernel rapidly switches between runnable processes. This causes longer latency for individual tasks and reduced overall throughput. You can observe saturation using tools like top, htop, or sar, which show load average and CPU utilization percentages.
Example: A web server experiencing a traffic spike might show 95-100% CPU utilization with a load average of 12 on an 8-core system. Requests queue up waiting for CPU time, increasing response latencies. This differs from CPU utilization (percentage of time CPU is busy)—you can have high utilization without saturation if work is well-distributed, or saturation with moderate utilization if context-switching overhead is high.