Method and apparatus for reliably choosing a master network manager during initialization of a network computing system

• US 6,941,350 B1
• Filed: 10/19/2000
• Issued: 09/06/2005
• Est. Priority Date: 10/19/2000
• Status: Active Grant

• Current Assignee: Stripe, Inc.
• Original Assignee: International Business Machines Corporation
• Inventors: Frazier, Giles Roger, Pfister, Gregory Francis, Thurber, Steven Mark, Van-Mierop, Dono
• Primary Examiner(s): EDELMAN, BRADLEY E
• Application Number:
  US09/692,346
• Time in Patent Office: 1,783 Days
• Field of Search: 709/208, 709/209, 709/223, 710/110, 714/4, 713/300
• US Class Current: 709/209
• CPC Class Codes: G06F 15/17375 One dimensional, e.g. linear array, ring
• CPC Class Codes: H04L 63/061 for key exchange, e.g. in peer-to-peer networks cryptographic mechanisms or cryptographic arrangements for key agreement H04L9/0838

Abstract

A method in a node within network computing system for selecting a master network manager, wherein the first node is associated with a first priority. Requests are sent to the network computing system to discover other nodes within the network computing system. A second priority from the request is identified in response to receiving a response to one of the requests from another node within the network computing system. The first node shifts to a standby mode if it discovers a master subnet manager or the second priority is higher than the first priority. The first node shifts to a master mode if a response containing a priority higher than the first priority is absent in responses received by the first node and the first node has completed checking all other nodes in the network computing system. In the case where the priority received is equal, the comparison is further made on the globally unique identifier which is received from the same node, in which case the node with the lowest globally unique identifier wins the arbitration.

Claims

1. A method in a node within network computing system for selecting a master subnet manager, the method comprising:

sending requests from a first node to the network computing system to discover other nodes within the network computing system, wherein the first node contains a first subnet manager having a first priority;

receiving a response to one of the requests from a second node within the network computing system, wherein the second node within the network computing system contains a second subnet manager having a second priority;

comparing the second priority to the first priority;

shifting the first subnet manager to a standby mode if the second priority is higher than the first priority; and

shifting the first subnet manager to a master mode if no response is received from any node containing a subnet manager having a priority higher than the first priority and if in the discovery of the other nodes within the network computing system is complete.

2. The method of claim 1, wherein the node is associated with a first unique value and wherein the second node is associated with a second unique value, the method further comprising:

comparing the first unique value to the second unique value if the first priority is equal to the second priority; and

shifting the first subnet manager to a standby mode if the first unique value identify is less than the second unique value.

3. The method of claim 2, wherein the first unique value and the second unique value are globally unique identifiers.

4. The method of claim 1, wherein the network computing system is a system area network.

5. The method of claim 1, wherein the requests are system management packets.

6. The method of claim 1 further comprising:

polling a master subnet manager in the network computing system in response to the first subnet manager shifting to a standby mode.

7. The method of claim 6, wherein the polling occurs periodically.

8. The method of claim 6 further comprising:

reinitiating the steps of sending, receiving, comparing, shifting to a standby mode, and shifting to a master mode if a response to polling of the master subnet manager is absent.

9. The method of claim 8, wherein the response to polling of the master subnet manager is considered absent if a response is not received from the master subnet manager within a selected period of time and commensurate to the poll operation not completing after the poll operation has been retried a predetermined number of times.

10. The method of claim 1, wherein the steps of sending receiving, comparing, shifting to a standby mode, and shifting to a master mode are initiated prior to initialization of an operating system for the first node.

11. The method of claim 1, further comprising:

shifting the first subnet manager from master code mode into standby mode in response to receiving a message to handover mastership of the network computing system.

12. The method of claim 1, further comprising:

shifting the first subnet manager to a non-active mode from the standby mode in response to receiving a message to shift to non-active mode from a master subnet manager.

13. A data processing system in a first node comprising:

a bus system;

a host channel adapter connected to the bus system, wherein the host channel adapter provides a communications link to a network computing system;

a memory including a set of instructions connected to the bus system; and

a processing unit connected to the bus system, wherein the processing unit executes the set of instructions to send requests from a first node to the network computing system to discover other nodes within the network computing system, wherein the first node contains a first subnet manager having a first priority;

receive a response to one of the requests from a second node within the network computing system, wherein the second node within the network computing system contains a second subnet manager having a second priority;

compare the second priority to the first priority;

shift the first subnet manager to a standby mode if the second priority is higher than the first priority; and

shift the first subnet manager to a master mode if no response is received from any node containing a subnet manager having a priority higher than the first priority and if discovery of the other nodes within the network computing system is complete.

14. The data processing system of claim 13, wherein the first node is associated with a first globally unique identification and wherein the second node is associated with a second globally unique identification and wherein if the data system first node has an equal priority as compared to a highest priority received from any other node in the network computing system, the processing unit compares the first globally unique identification to the second globally unique identification to determine whether to shift into a master mode.

15. A network computing system comprising:

fabric, wherein the fabric facilitates transfer of date; and

a plurality of nodes connected to the fabric, wherein a first node within the plurality of nodes contains a first subnet manager that has a first priority and sends requests to discover other nodes within the network computing system, receives a response to one of the requests from a second node containing a second subnet manager that has a second priority, comparing the second priority to the first priority, shifts the first subnet manager to a standby mode if the second priority is higher than the first priority node, and shifts the first subnet manager to a master mode if discovery of the other nodes within the network computing system is complete and a response containing a subnet manager having a priority higher than the first priority the node is absent in responses received by the node.

16. An apparatus for selecting a master subnet manager, the apparatus comprising:

sending means for sending requests from a first node to a network computing system to discover other nodes within the network computing system, wherein the first node contains a first subnet manager having a first priority;

receipt means for receiving a response to one of the requests from a second node within the network computing system, wherein the second node within the network computing system contains a second subnet manager having a second priority;

comparison means for comparing the second priority to the first priority;

first shifting means for shifting the first subnet manager to a standby mode if the second priority is higher than the first priority; and

second shifting means for shifting the first subnet manager to a master mode if no response is received from any node containing a subnet manager having a priority higher than the first priority is and if discovery of the other nodes within the network computing system is complete.

17. The apparatus of claim 16, wherein the first node is associated with a first unique value and wherein the second node is associated with a second unique value, the method further comprising:

comparing means for comparing the first unique value to the second unique value if the first priority is equal to the second priority; and

third shifting means for shifting the first subnet manager to a standby mode if the first unique value identify is less than the second unique value.

18. The apparatus of claim 17, wherein the first unique value and the second unique value are globally unique identifiers.

19. The apparatus of claim 16, wherein the network computing system is a system area network.

20. The apparatus of claim 16, wherein the requests are system management packets.

21. The apparatus of claim 16 further comprising:

polling means for polling a master subnet manager in the network computing system in response to the first node shifting to a standby mode.

22. The apparatus of claim 21, wherein the polling occurs periodically.

23. The apparatus of claim 21 further comprising:

reinitiating means for reinitiating the sending means, receipt means, comparison means, first shifting means, and second shifting means if a response to polling of the master is absent.

24. The apparatus of claim 23, wherein the response to polling of the master subnet manager is considered absent if a response is not received from the master subnet manager within a selected period of time and commensurate to the poll operation not completing after the poll operation has been retried a predetermined number of times.

25. The apparatus of claim 16, wherein the sending means, receipt means, comparison means, first shifting means, and second shifting means are initiated prior to initialization of an operating system for the first node.

26. The apparatus of claim 16, further comprising:

third shifting means for shifting the first subnet manager from master mode into standby mode in response to receiving a message to handover mastership of the network computing system.

27. The apparatus of claim 16, further comprising:

third shifting means for shifting the first subnet manager to non-active mode from standby mode in response to receiving a message to shift to non-active mode from a master subnet manager.

28. A computer program product in a tangible computer readable medium for use in a first node within network computing system for selecting a master subnet manager, the computer program product comprising:

first instructions for sending requests from the first node to the network computing system to discover other nodes within the network computing system, wherein the first node contains a first subnet manager having a first priority;

second instructions for receiving a response to one of the requests from a second node within the network computing system, wherein the second node contains a second subnet manager having a second priority;

third instructions for comparing the second priority to the first priority;

fourth instructions for shifting the first subnet manager to a standby mode if the second priority is higher than the first priority; and

fifth instructions for shifting the first subnet manager to a master mode if no response is received from any node containing a subnet manager having a priority higher than the first priority and if discovery of the other nodes within the network computing system is complete.