Guaranteeing delivery of multi-packet GSM messages

A processing node within a distributed data processing system comprising:

• a processing unit with a target task executing thereon, said target task being a local task of a distributed job having multiple tasks executing at different nodes across a global shared memory (GSM) environment;
• a local memory coupled to the processing unit and having real address (RA) space;
• a host fabric interface (HFI) including a HFI window assigned to the target task;
• and
• program instructions executing on the processor and providing processing logic for ensuring complete delivery of a global shared memory (GSM) message from an originating task on a first node coupled to a network fabric of a distributed network via a HFI to a target task on a second node also coupled to the network fabric via a target HFI, wherein the processing logic instructions comprises processing logic for:
• receiving a first GSM packet of a plurality of GSM packets generated from a single GSM message being sent from the originating task;
• assigning a sequence number to track receipt of the GSM message from the originating task, said sequence number being unique from other sequence numbers assigned to GSM messages from the same originating task that have not been completely received;
• updating a count value within a tuple of the first GSM packet, the count value comprising a count including the first GSM packet and one or more subsequently received GSM packets of the plurality of GSM packets that have been received for the GSM message;
• and
• determining when receipt of the GSM message is complete by comparing the count value within the tuple of the first GSM packet with a count total retrieved from the packet header, which indicates a total number of GSM packets being transmitted for the single GSM message.

A computer program product comprising:

• a computer readable storage medium;
• and
• program code on the computer readable storage medium that when executed on a processing node performs the function of ensuring complete delivery of a global shared memory (GSM) message from an originating task on a first node coupled to a network fabric of a distributed network via a host fabric interface (HFI) to a target task on a second node also coupled to the network fabric via a target HFI, wherein said program code comprising code for:
• receiving a first GSM packet of a plurality of GSM packets generated from a single GSM message being sent from the originating task;
• assigning a sequence number to track receipt of the GSM message from the originating task, said sequence number being unique from other sequence numbers assigned to GSM messages from the same originating task that have not been completely received;
• updating a count value within a tuple of the first GSM packet, the count value comprising a count including the first GSM packet and one or more subsequently received GSM packets of the plurality of GSM packets that have been received for the GSM message;
• and
• determining when receipt of the GSM message is complete by comparing the count value within the tuple of the first GSM packet with a count total retrieved from the packet header, which indicates a total number of GSM packets being transmitted for the single GSM message.