Abstract: A computer system performs memory to memory transfer, task scheduling and I/O request handling via a group of dedicated processors (e.g. a memory interface unit, an I/O unit, a data transfer unit, and a task control unit). The memory interface unit facilitates data interaction between the memory and the remainder of the system. The I/O unit is coupled to the memory interface unit and performs high level I/O job functions including I/O job scheduling, I/O job path selection, gathering of job statistics and device management. The data transfer unit is coupled to the memory interface unit and moves data between memory locations. The task control unit, coupled to the memory interface unit, allocates and deallocates events, maintains the status of tasks running on the system and schedules the execution of tasks. A hierarchical error reporting scheme is used by all of the processors.

Abstract: An interface controller coupled between the main memory system and the I/O system of a large data processing system which controller is able to receive memory access requests from a number of different peripheral devices. The memory interface controller is provided with a data array for holding a number of data words fetched from memory which data array in turn is addressed by the output of an address array. The address array is an associative memory that can associate a given main memory address, of data in the data array, with a data array address containing data stored in main memory at that main memory address so that actual main memory access need not be required.

Abstract: The present invention involves the use of multiple I/O processors (204) and (206), configured in parallel in an I/O system (104), to increase system performance as well as enhance system resiliency. Performance is increased by programming one of the multiple I/O processors with the added ability to allocate received I/O job requests among the active I/O processors including itself, thus allowing for parallel processing. The I/O system resiliency is enhanced by ensuring that any one of the I/O processors can assume the tasks of any other I/O processor which may fail. The I/O system described above employs a load balancing algorithm to evenly distribute the I/O job functions among the active I/O processors.

Abstract: A mechanism for queuing a set of happened events in order of their occurrence and allowing for multiple occurrences to result in multiple processing iterations which mechanism maintains a multi-event table which is really a table of multi-event designations to be allocated to different processes upon request where the requesting processes assign a particular function with each multi-event entry and each of its own related event designations.