Apparatus and method for computer job workload distribution

• US 6,202,080 B1
• Filed: 12/11/1997
• Issued: 03/13/2001
• Est. Priority Date: 12/11/1997
• Status: Expired due to Fees

• Current Assignee: Nortel Networks Limited (Nortel Networks Corporation)
• Original Assignee: Nortel Networks Limited (Nortel Networks Corporation)
• Inventors: Lu, James, Juan, Shao-Min, Asava, Anand
• Primary Examiner(s): BANANKHAH, MAJID A
• Application Number:
  US08/988,682
• Time in Patent Office: 1,188 Days
• Field of Search: 709/100, 709/102, 709/103, 709/104, 709/105, 709/217, 709/219, 709/223
• US Class Current: 718/105
• CPC Class Codes: G06F 9/5083 Techniques for rebalancing the load in a distributed system

Abstract

A job distribution system implemented by a computer network having a plurality of nodes, each having a cross-mounted local directory. The job distribution system has a job balance object, a node-processing object, and an updatable node distribution table. The job distribution system is on a first node of the plurality of nodes. A node daemon object is on each of the plurality of nodes. The job balance object has access to the local directory of each remaining node. The updatable node distribution table has a plurality of distribution data structures which correspond to each of the remaining nodes, and are updated by the node daemon object with a job status value. With these objects, the job balance object distributes the plurality of jobs by averaging the job status value of each of the remaining nodes to distribute a plurality of the pending job requests.

Claims

1. A job distribution system implemented by a computer network with a plurality of nodes having a local node and at least one remote node, each of the nodes of the plurality of nodes capable of executing one or more job commands, and each of the nodes of the plurality of nodes having a local directory cross-mounted with the local directories of the plurality of network nodes, the job distribution system comprising:

a job balance object on the local node of the plurality of nodes, said job balance object having access to the local directory of the at least one remote node and being capable of reading a node distribution table and distributing a plurality of job requests to the at least one remote node;

a node daemon object on the at least one remote node of the plurality of nodes, the node daemon object capable of updating a node distribution table with a job status value; and

an updatable node distribution table having a plurality of distribution data structures, each data structure of said plurality of distribution data structures corresponding to each of the at least one remote node having said node daemon object, and being capable of being updated by said node daemon object with a job status value, wherein said job balance object reads said plurality of distribution data structures to determine which of a plurality of job requests are pending on the at least one remote node and distributes said plurality of job requests to the at least one remote node.

2. The job distribution system of claim 1 wherein said job balance object distributes said plurality of jobs one or more jobs according to an average of said job status value per each of the at least one remote node.

3. The job distribution system of claim 1, wherein said node daemon object is capable of spawning at least one child process to execute a job request of said plurality of job requests.

4. The job distribution system of claim 1 wherein said updatable node distribution table is stored on one of the local directories of said plurality of nodes.

5. The job distribution system of claim 1 wherein said node distribution table is stored on the local directory of the local node.

6. The job distribution system of claim 1 wherein said job status value is an amount of pending jobs requests on the at least one remote node.

7. The job distribution system of claim 1 wherein each of said distribution data structures has a WorkLoad field, a JobsOnNode field, a JobRequestQue field, and a JobResponseQue field updatable with said job status value.

8. The job distribution system of claim 1 wherein said job balance object is capable of copying said plurality of job requests to a local directory of the at least one remote node.

9. The job distribution system of claim 1 wherein said job balance object sustains an average of pending job requests across the plurality of nodes.

10. The job distribution system of claim 1 wherein said object assigns a new job to a node having a lowest amount of pending jobs with respect to the at least one node.

11. A workload distribution method for a computer network with a plurality of nodes in a local-remote node relationship having a network file system that cross-mounts a local directory of each node with the local directories of each other node, the workload distribution method comprising the steps of:

providing an updatable node distribution table with a plurality of distribution data structures, each data structure of the plurality of distribution data structures correlating to a node of the plurality of nodes;

providing a job balance object on a local node of the plurality of nodes;

submitting a plurality of job requests by each node of the plurality of nodes;

using at least one node daemon object to update each of the distribution data structures that correspond to each of the nodes of the plurality of nodes with a first job status value;

using the job balance object to read each of the distribution data structures to determine which of the plurality of job requests are pending on each of the plurality of nodes;

using the job balance object to balance the plurality of job requests among the plurality of nodes according to a balancing criteria; and

updating the distribution data structures that correspond to each of the nodes of the plurality of nodes with a second job status value when a job request of the plurality of job requests is executed.

12. The workload distribution method of claim 11 wherein the balancing criteria is based on an amount of pending jobs on a node of the plurality of nodes being less than an average of job requests for all the nodes of the plurality of nodes.

13. The workload distribution method of claim 11 wherein the balancing criteria further comprises a capability to redistribute the plurality of job requests to sustain a balance of job requests of each node of said plurality of nodes.

14. The workload distribution method of claim 11 wherein the step of balancing the job requests further comprises the step of copying one or more of the plurality of job requests to the other nodes.

15. The workload distribution method of claim 11 wherein the second job status value is a pointer to a job response file.

16. A method of distributing computer jobs to a plurality of remote nodes on a network from a local node comprising the steps of:

maintaining on a remote node an updatable node distribution table corresponding to each of the plurality of remote nodes;

updating a job status value in each of said tables as jobs are assigned and when the jobs are completed;

reading said tables via the job balance object on the local node; and

balancing, via the job balance object on the local node, an assignment of a plurality of new jobs to the plurality of remote nodes as a function of jobs remaining to be executed at that node as compared to the average number of jobs for all of the plurality of remote nodes.