In a method for processor overload control in a wireless or other network, a processor occupancy level (“PO”) of a network processing unit is monitored and compared to a target PO. If the measured PO exceeds the target PO, one or more network load sources are controlled to reduce the data load of the processing unit, until the measured PO falls below a designated PO. “Load source” refers to a source of extant and/or potential data traffic through the processing unit, and/or a network control factor relating thereto, the control of which results in a reduction (or prevents an increase) in the data traffic handled by the processor. Examples include existing data flows, new calls, and BE flow window size. The load sources are controlled according to differentiated QOS levels, wherein non delay-critical data flows (e.g., BE flows) are reduced before delay-critical data flows, e.g., EF or AF flows.