Fault-tolerant method of communicating between processes in a multi processor system by keeping track of the current node location of messages

• US 5,404,501 A
• Filed: 12/21/1992
• Issued: 04/04/1995
• Est. Priority Date: 08/31/1989
• Status: Expired due to Fees

• Current Assignee: Motorola, Inc. (Motorola Solutions, Inc.)
• Original Assignee: Motorola, Inc. (Motorola Solutions, Inc.)
• Inventors: Weisshaar, Bernhard P., Carr, Patricia A.
• Primary Examiner(s): Kulik, Paul V.
• Assistant Examiner(s): Loomis, John C.
• Application Number:
  US07/993,491
• Time in Patent Office: 834 Days
• Field of Search: 371/32, 371/33, 371/34, 395/200, 395/575, 395/650
• US Class Current: 719/314
• CPC Class Codes: G06F 11/1443 Transmit or communication errors

Abstract

A method for communicating between processes in a data processing system comprising a plurality of processor hosts each coupled to a network and a plurality of processes resident on different ones of said processor hosts, wherein messages are transmitted from one process to another in a logical ring. A message-holding queue is maintained at any host originating a message transmission, and it contains a copy of the message, an identifier of the initiating processor host, and an identifier of the target processor host. A forward notification message is returned to the originating host from a forwarding host when the forwarding host relays the message to another host, and an identical forward notification message is returned to the originating host by the target host when it receives the message. If the host on which the message currently resides becomes separated from the initiating host or terminates operation, or if any process becomes inactive after receiving the message and before relaying it to the next host or is unable to relay the message, flags may be set in the message, and, if necessary, the originating host retransmits the message into the logical ring.

Claims

1. A method for monitoring communication between processes in a data processing system comprising a plurality of processor hosts, including first, second, and third processor hosts each coupled to a network and a plurality of processes resident on different ones of said processor hosts, wherein said first processor host comprises a message-holding queue for storing information regarding messages transmitted among said processes said information including a copy of messages which are transmitted and identifiers for identifying the source and current location of said messages, said method comprising the steps of:

(a) transmitting a first message from a first process resident on said first processor host to a second process resident on said second processor host;

(b) storing in an entry of said message-holding queue a copy of said first message, a first identifier identifying said first process as the source of said first message, and a second identifier identifying said second processor host as the processor host where said first message currently resides;

(c) transmitting said first message from said second process to a third process resident on said third processor host;

(d) generating a forward notification message by said second processor host, said forward notification message comprising said first identifier and a third identifier identifying said third processor host as the processor host where said first message currently resides; and

(e) transmitting said forward notification message from said second processor host to said first processor host.

2. The method recited in claim 1 and further comprising the steps of:

(f) receiving said forward notification message by said first processor host; and

(g) responsive to the receipt by said first processor host of said forward notification message, storing said third identifier in said message-holding queue.

3. The method recited in claim 2 and further comprising the steps of:

(h) when said first message is received by said third processor host, generating an additional forward notification message by said third processor host, said additional forward notification message comprising said first identifier and said third identifier identifying said third processor host; and

(i) transmitting said additional forward notification message from said third processor host to said first processor host.

4. The method recited in claim 3 and further comprising the steps of:

(j) receiving said additional forward notification message by said first processor host; and

(k) responsive to the receipt by said first processor host of said additional forward notification message, storing said third identifier in said message-holding queue.

5. The method recited in claim 4, and further comprising the step of:

(l) repeating steps (c) through (k) in an effort to transmit said first message successively from a process on one processor host to a process on a different processor host until said first message returns to said first processor host,

wherein if said first message returns to said first process in said first processor host, deleting said entry in said message-holding queue,

and wherein if any processor host on which said first message is resident becomes disconnected from said first processor host or becomes inoperative,

i) transmitting a failure message over said network regarding the status of said disconnected or inoperative processor host;

ii) receiving said failure message by said first processor host;

iii) dequeuing said copy of said first message from said message-holding queue by said first processor host;

iv) transmitting said first message by said first processor host to a process on one of said processor hosts; and

v) repeating steps (b) through (k) in an effort to transmit said first message successively from a process on one processor host to a process on a different processor host until said first message returns to said first processor host.

6. The method recited in claim 4, and further comprising the step of:

(l) repeating steps (c) through (k) in an effort to transmit said first message successively from a process on one processor host to a process on a different processor host until said first message returns to said first processor host,

wherein if said first message returns to said first process in said first processor host, deleting said entry in said message-holding queue,

and wherein if any process becomes inactive after receiving said first message and before transmitting said first message to a process located on a different processor host,

i) transmitting a failure message over said network regarding the status of said inactive process;

ii) receiving said failure message by said first processor host;

iii) dequeuing said copy of said first message from said message-holding queue by said first processor host;

iv) transmitting said first message by said first processor host to a process on one of said processor hosts; and

v) repeating steps (b) through (k) in an effort to transmit said first message successively from a process on one processor host to a process on a different processor host until said first message returns to said first processor host.

7. The method recited in claim 4, and further comprising the step of:

(l) repeating steps (c) through (k) in an effort to transmit said first message successively from a process on one processor host to a process on a different processor host until said first message returns to said first processor host,

wherein if said first message returns to said first process in said first processor host, deleting said entry in said message-holding queue,

and wherein if any process is unable to transmit said first message to a process located on a different processor host,

i) transmitting a failure message over said network regarding the status of said incapacitated process;

ii) receiving said failure message by said first processor host;

iii) dequeuing said copy of said first message from said message-holding queue by said first processor host;

iv) transmitting said first message by said first processor host to one of said processor hosts; and

v) repeating steps (b) through (k) in an effort to transmit said first message successively from a process on one processor host to a process on a different processor host until said first message returns to said first processor host.