Use of a storage medium as a communications network for liveness determination in a high-availability cluster

• US 20040123050A1
• Filed: 12/18/2002
• Published: 06/24/2004
• Est. Priority Date: 12/18/2002
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Hare, John R., Knop, Felipe
• Granted Patent: US 7,003,645 B2
• Time in Patent Office: Days
• US Class Current: 711/147
• CPC Class Codes: H04L 12/66 Arrangements for connecting between networks having differing types of switching systems, e.g. gateways

Abstract

Liveness determination in a multinode data processing system is enhanced through the use of a shared nonvolatile memory, typically a disk, which is utilized in conjunction with defined transmission protocols to provide an independent communication path usable for both the transmission of node liveness information and for other, more general, data transfer purposes. This path is considered as having the same degree of importance as other network paths and liveness indications from the shared memory path are likewise considered at the same time as liveness indications from another path or paths.

Claims

1. A method for liveness determination in a cluster of data processing nodes, said method comprising the step of:

providing at least two paths for communication of liveness information in which at least one of said paths includes a shared nonvolatile data storage device and in which said path with said nonvolatile data storage device is employed substantially continuously and concurrently with other liveness determination paths, whereby node liveness is determinable without having to resort to using secondary communication paths that are triggered by indications of liveness failure over other communication paths.

2. The method of claim 1 in which said nonvolatile data storage device is selected from the group consisting of magnetic disks and optical disks.

3. A method for communication in a multinode data processing system which includes at least one nonvolatile data storage device, said communication method comprising:

determining that an area of said storage device reserved for write-access by a first node is available for use;

transferring information from said first one of said nodes to said area of said storage device reserved for write-access by said first node, said information including an updated sequence number;

retrieving said information by a second node having read-only access to said area of said storage device reserved for write-access by said first node; and

transferring said updated sequence number from said second node to an area of said storage device reserved for write-access by said second node, whereby said first node is able to determine when its associated write-access area is available for information transmission.

4. The method of claim 3 in which said nonvolatile data storage device is selected from the group consisting of magnetic disks and optical disks.

5. The method of claim 3 further including:

determining a write sequence number based on sequence numbers associated with packets produced by the first node;

determining a read sequence number based on a highest or lowest sequence number associated with packets read by the second node, whereby the first node can determine that its write area is available; and

monotonically varying said write sequence number and transferring it with new information to said area of said storage device reserved for write-access by said first node.

6. The method of claim 3 further including the steps of:

whenever said second node is rebooted, determining said second node'"'"'s write sequence number based on the packets present in said second node'"'"'s write area;

determining the read sequence number based on the packets present in said first node'"'"'s write area; and

transferring said read sequence number to said area of said storage device reserved for write-access by said second node, whereby said first node is thus able to determine that its write area is available for use.

7. The method of claim 3 in which said information is liveness information concerning said first node.

8. The method of claim 3 in which said area reserved for write-access by said first node includes a sector into which both control information and data are written.

9. The method of claim 8 in which said area reserved for write-access by said first node also includes at least one additional sector for data to be transferred.

10. The method of claim 8 in which information for transfer is limited in size to fit in said sector, whereby disk accessing operations are reduced in number.

11. The method of claim 3 in which said determining step is limited to determining whether or not space is available in said write-access area for said first node for information transfer, whereby access to said device for such determination is avoided.

12. The method of claim 3 in which access to said nonvolatile data storage device is handled through a liveness network abstraction layer.

13. A method for initializing information on a nonvolatile storage device having access thereto shared between at least two nodes in a multinode data processing system, said method comprising the steps of:

writing, by a first node, information to a sector on said device in a location node known to said first node and to a second node, said information containing;

a node identifier for said first node, a flag indicating whether or not the node identifier of said second node is known, and a number indicating that said information in said sector contains data of a certain type;

waiting for a random period of time;

reading, by said first node, said information from said sector;

repeating said previous steps until the unique node identifier from said sector is different from the identifier for said first node; and

selecting, for use by said first node, an area on said device for further communication, said selection being based upon difference between said identifiers.

14. The method of claim 14 further including the steps of:

writing, by a first node, information to said selected area on said device, further information containing;

said node identifier for said first node, a flag indicating whether or not the node identifier of said second node is known, and a number indicating that said information in said sector contains data of a certain type;

waiting for a random period of time;

reading, by said first node, said information from said sector;

repeating said previous steps until the node identifier from said sector is different from the identifier for said first node and until said flag indicates that the node identifier of said first node is known by said second node , whereby said first node now knows what area it is to use and also that said second node knows this as well.

15. A computer readable medium having computer executable instructions for causing a computer to provide at least two paths for communication of liveness information in which at least one of said paths includes a shared nonvolatile data storage device and in which said path with said nonvolatile data storage device is employed substantially continuously and concurrently with other liveness determination paths, whereby node liveness is determinable without having to resort to using secondary communication paths that are triggered by indications of liveness failure over other communication paths in a multinode data processing system with shared access to said nonvolatile data storage device.