Abstract: A method for scheduling functions for execution immediately or later in time by the scheduling processor or another processor of a shared memory multiprocessor computer system such as a NUMA machine. Associated with each of the multiple processors is a data structure for scheduling the execution of specified functions by the processor. These multiple data structures with associated locks allow for simultaneous access by processors to their data structures and thereby avoid the bottleneck inherent in the prior approach of providing only a single, global data structure for all processors. In addition, the method allows any processor to call a scheduling function to add a specified function to the data structure of any of the processors. Processor loads are balanced by moving scheduled specified functions from one processor to another. Scheduled functions can also be moved from one processor to another to allow a processor to be taken offline for service.