Load-balancing framework for a cluster

• US 7,757,236 B1
• Filed: 06/28/2004
• Issued: 07/13/2010
• Est. Priority Date: 06/28/2004
• Status: Active Grant

• Current Assignee: Oracle America, Inc. (Oracle Corporation)
• Original Assignee: Oracle America, Inc. (Oracle Corporation)
• Inventors: Singh, Servesh Pratap
• Primary Examiner(s): An; Meng-Ai
• Assistant Examiner(s): Truong; Camquy
• Application Number:
  US10/878,894
• Time in Patent Office: 2,206 Days
• Field of Search: 718/105, 719/328, 709/224
• US Class Current: 718/105
• CPC Class Codes: G06F 2209/505 Clust
• CPC Class Codes: G06F 9/5083 Techniques for rebalancing the load in a distributed system

Abstract

Various embodiments of a system and method for balancing nodal loading in a cluster are disclosed. In a cluster, each node may execute an instance of a distributed application component. A pluggable load balancer may receive requests targeted at the application component. For each request, the load balancer may invoke a pluggable load-balancing function to determine a next node to which the request should be forwarded to balance the workload of the cluster nodes. Each node of the cluster may include a load-balancing framework that provides a standard load-balancing API. In order to determine the next node to which a request should be milted, the load balancer may send requests for node information conforming to the API, to the cluster members. This information may include nodal CPU speed, memory capacity, number of queued requests, and number of requests serviced during a specified time period.

Claims

1. A system, comprising:

a plurality of computers configured to implement;

a cluster comprising a plurality of nodes configured to execute a distributed application component, wherein each node comprises a load-balancing framework providing a standard load-balancing application programming interface (API);

a pluggable load balancer, wherein the pluggable load balancer is configured to communicate with the plurality of nodes using the standard load-balancing framework API to forward a request to one of the plurality of nodes according to a load-balancing algorithm; and

a naming service configured to receive a lookup request for the application component and in response return a stub corresponding to one of the plurality of nodes, wherein the pluggable load balancer is comprised in the naming service, and wherein the stub corresponds to a next node as determined by the pluggable load balancer;

wherein the load-balancing framework on each node provides for the pluggable load balancer to be selectable for plugging in to the system, wherein the pluggable load balancer is selectable from among a plurality of different pluggable load balancers not currently plugged in to the system, and wherein the load-balancing framework on each node is configured to support communication with each of the plurality of different pluggable load balancers using the same standard load-balancing framework API.

2. The system, as recited in claim 1, wherein the load-balancing framework is configured to send load-balancing support data to the pluggable load balancer in compliance with the API.

3. The system, as recited in claim 2, wherein the load-balancing support data comprises one or more of the following:

a node CPU speed, a number of jobs in execution, a number of jobs in queue awaiting execution, and a number of jobs executed during a time period.

4. The system, as recited in claim 2, wherein the load-balancing framework is configured to send its load-balancing support data to the pluggable load balancer on a periodic or aperiodic basis.

5. The system, as recited in claim 2, wherein the load-balancing framework is configured to send its load-balancing support data to the pluggable load balancer in response to a request from the pluggable load balancer.

6. The system, as recited in claim 2, wherein the pluggable load balancer adjusts its load-balancing algorithm according to the load-balancing support data received from the load-balancing framework of one or more of the nodes.

7. The system, as recited in claim 1, wherein each cluster node comprises an additional pluggable load balancer configured to:

monitor the workload of its node;

when the workload exceeds a predetermined limit, forward received jobs to another of the plurality of nodes.

8. The system, as recited in claim 1, wherein the pluggable load balancer is configured via one or more deployment descriptors on each node.

9. The system, as recited in claim 1, further comprising a client comprising an augmented stub for interfacing to the application component, wherein the augmented stub comprises:

a stub corresponding to one of the plurality of nodes;

wherein the augmented stub is configured to forward a client request to the one of the plurality of nodes using the stub;

wherein another pluggable load balancer is comprised in the augmented stub; and

wherein the stub corresponds to a next node as determined by the another pluggable load balancer.

10. The system, as recited in claim 1, further comprising another pluggable load balancer that is a separate, dedicated load-balancing process.

11. The system, as recited in claim 10, wherein the another pluggable load balancer is configured to:

monitor a characteristic of a node;

detect when a value of the characteristic reaches a specified limit;

in response to detecting when the value of the characteristic reaches the specified limit, forward received requests to a next node.

12. A method, comprising:

using one or more computers to perform;

selecting a current pluggable load balancer, from among a plurality of different pluggable load balancers, for a server cluster, wherein each node of the cluster comprises a pluggable load-balancing framework, and wherein the pluggable load balancer is specified according to the pluggable load-balancing framework;

the pluggable load balancer forwarding a request to one of the nodes according to a load-balancing algorithm and in compliance with a load-balancing API comprised in the pluggable load-balancing framework;

selecting a different one of the plurality of different pluggable load balancers to replace the current pluggable load balancer, wherein each of the plurality of different pluggable load balancers uses the same standard load-balancing framework API to communicate with the nodes of the cluster; and

sending, by the load-balancing framework, load-balancing support data on a periodic or aperiodic basis to the pluggable load balancer in compliance with the API, wherein the load-balancing support data comprises one or more of the following;

a node CPU speed, a number of jobs in execution, a number of jobs in queue awaiting execution, or a number of jobs executed during a time period.

13. The method, as recited in claim 12, further comprising sending the load-balancing support data to the pluggable load balancer in response to a request from the pluggable load balancer.

14. The method, as recited in claim 12, further comprising adjusting a load-balancing algorithm according to the load-balancing support data.

15. The method, as recited in claim 12, further comprising:

monitoring the workload of a node;

when the workload exceeds a predetermined limit, forwarding received jobs to another of the cluster nodes.

16. The method, as recited in claim 12, wherein the pluggable load balancer is configured via one or more deployment descriptors on each node.

17. The method, as recited in claim 12, further comprising a naming service returning a stub corresponding to a next node as determined by the pluggable load balancer.

18. The method, as recited in claim 12, further comprising an augmented stub forwarding the request to a next node as determined by the pluggable load balancer.

19. The method, as recited in claim 12, wherein the pluggable load balancer is a separate, dedicated load-balancing process.

20. The method, as recited in claim 19, further comprising:

monitoring a characteristic of one of the plurality of nodes;

detecting when a value of the characteristic reaches a specified limit;

in response to detecting when the value of the characteristic reaches the specified limit, forwarding received requests to a next node.

21. A computer-readable storage medium storing program instructions, wherein the program instructions are computer-executable to implement:

a load-balancing framework providing a standard load-balancing application programming interface (API);

a pluggable load balancer, wherein the pluggable load balancer is configured to communicate with a plurality of nodes of a server cluster using the standard load-balancing framework API to forward a request to one of the plurality of nodes according to a load-balancing algorithm; and

a naming service configured to receive a lookup request for the application component and in response return a stub corresponding to one of the plurality of nodes, wherein the pluggable load balancer is comprised in the naming service, and wherein the stub corresponds to a next node as determined by the pluggable load balancer;

wherein the load-balancing framework provides for the pluggable load balancer to be selectable for plugging in to the system, wherein the pluggable load balancer is selectable from among a plurality of different pluggable load balancers not currently plugged in to the system, and wherein the load-balancing framework is configured to support communication with each of the plurality of different pluggable load balancers using the same standard load-balancing framework API.

22. The computer-readable storage medium as recited in claim 21, wherein the program instructions are further computer-executable to send load-balancing support data to the pluggable load balancer in compliance with the API.

23. The computer-readable storage medium as recited in claim 22, wherein the load-balancing support data comprises one or more of the following:

a node CPU speed, a number of jobs in execution, a number of jobs in queue awaiting execution, and a number of jobs executed during a time period.

24. The computer-readable storage medium as recited in claim 22, wherein the program instructions are further computer-executable to send the load-balancing support data to the pluggable load balancer on a periodic or aperiodic basis.

25. The computer-readable storage medium as recited in claim 22, wherein the program instructions are further computer-executable to send the load-balancing support data to the pluggable load balancer in response to a request from the pluggable load balancer.

26. The computer-readable storage medium as recited in claim 22, wherein the program instructions are further computer-executable to adjust a load-balancing algorithm according to the load-balancing support.

27. The computer-readable storage medium as recited in claim 21, wherein the program instructions are further computer-executable to:

monitor the workload of a node;

when the workload exceeds a predetermined limit, forward received jobs to another of the cluster nodes.

28. The computer-readable storage medium as recited in claim 21, wherein the pluggable load balancer is configured via one or more deployment descriptors on each node.

29. The computer-readable storage medium as recited in claim 21, wherein the program instructions are further computer-executable to implement an augmented stub configured to forward the request to a next node as determined by the pluggable load balancer.

30. The computer-readable storage medium as recited in claim 21, wherein the program instructions are further computer-executable to implement another pluggable load balancer as a separate, dedicated load-balancing process.

31. The computer-readable storage medium as recited in claim 30, wherein the program instructions are further computer-executable to:

monitor a characteristic of one of the plurality of nodes;

detect when a value of the characteristic reaches a specified limit;

in response to detecting when the value of the characteristic reaches the specified limit, forward received requests to a next node.