Abstract: A method for strong partitioning of multi-processor applications that maintains fault containment on the VMEbus is presented. The method implements a message passing mechanism to increase fault tolerance and localize and contain detected faults on any one card connected to a VMEbus. Software implementation supports message-based inter-module communications using the available features provided by the VME cards. A portion of each VME card's global memory is allocated upon initialization for supporting the message passing mechanism. Access to the message buffer is restricted to read-only for modules that do not own that buffer. An error detection or correction code (e.g. cyclic redundancy check) is appended to the end of a data stream message to contain faults and prevent their propagation to other VME cards.