METHOD AND APPARATUS FOR DATA TRANSFER AND STORAGE IN A HIGHLY PARALLEL COMPUTER NETWORK ENVIRONMENT

摘要

A method and apparatus for sharing a single storage device among an arbitrarily large number of parallel processes with typically no wasted memory space and continued operations even when an error or interrupt occurs is disclosed. In a networked computer system having a communication network connecting a primary storage device between one or more computer devices and a secondary storage device, when some type of interrupt is generated or an error condition occurs or is otherwise indicated, a transfer of data from primary to secondary storage is indicated. A state indicator having at least a non-backup and a backup state may be set to the backup state to indicate that a data transfer from the primary storage device to the secondary storage device should occur. Alternatively, the transfer from the primary to the secondary storage device can begin without the use of a state indicator. When the state indicator is in the backup state, a predetermined maximum number of processes will work on transferring data from the primary storage device to the secondary storage device. Other processes will transfer data from the computer to the primary storage device. The process which completes the transfer of data from the primary storage device to the secondary storage device then resets the state indicator to its normal state. Preferably, other processes which are already executing at the time when the state indicator is set to the backup state continue to execute to completion.