Method and apparatus for dynamic interval-based load balancing

• US 6,078,943 A
• Filed: 02/07/1997
• Issued: 06/20/2000
• Est. Priority Date: 02/07/1997
• Status: Expired due to Term

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Yu, Philip Shi-Lung
• Primary Examiner(s): Banankhah, Majid A.
• Application Number:
  US08/798,385
• Time in Patent Office: 1,229 Days
• Field of Search: 709/102, 709/103, 709/105, 709/100
• US Class Current: 718/105
• CPC Class Codes: G06F 9/505 considering the load

Abstract

A method for use in geographically distributed or clustered system wherein an arbiter assigns clients to servers. The arbiter also dynamically assigns a valid time interval to each mapping request based on network load and/or capacity parameters such as the client request rate and/or the server capacity. Alternative means for dynamically setting the valid interval in conjunction with a scheduling process, which can be either deterministic or probabilistic, are also devised.

Claims

1. In a distributed or clustered network of servers wherein clients are divided into groups which periodically send mapping requests to an arbitrator for mapping and balancing service requests among multiple replicated servers which can service the request, a computerized method for mapping servers to service requests, said method comprising the steps of:

mapping a first mapping request from a first group to a first server according to a schedule;

dynamically computing a valid interval for said mapping request to the first server as a function of one of a first group request load and a first server capacity; and

communicating the server selection and the valid interval to the first group for caching such that subsequent requests from the first group are routed to the first server during the valid interval.

2. The method of claim 1, wherein each of the servers are web servers at geographically distributed sites, the arbiter is an extended domain name server and each address mapping includes a time-to-live (TTL), and wherein the group consists of a set of clients associated with one of a common gateway and SOCKS server, said method comprising the steps of:

said mapping step comprises mapping a name request to an IP address; and

said step of dynamically computing a valid interval comprises dynamically computing the TTL as a function of one of a gateway request load and a capacity of a destination server associated with the IP address.

3. The method of claim 1, wherein said mapping step comprises the step of mapping according to one of a round robin and multi-tier round robin schedule;

and wherein the valid interval is inversely proportional to the request load and proportional to the server capacity.

4. The method of claim 1, wherein said mapping step comprises the step of mapping according to one of a probabilistic round robin and probabilistic multi-tier round robin schedule as a function of the server capacity;

and wherein the valid interval is inversely proportional to the request load.

5. The method of claim 1, wherein said step of dynamically computing a valid interval for said mapping request to the first server comprises the steps of:

estimating the request load, including one of a hidden or explicit load associated therewith, and directed to the first server; and

setting the valid interval as function of the request load associated with the group, in response to said estimating step.

6. The method of claim 5, wherein said step of setting the valid interval as function of the request load comprises the step of setting the valid interval as inversely proportional to the request load associated with the group.

7. The method of claim 1, wherein said step of dynamically computing a valid interval for said mapping request to the first server comprises the steps of:

estimating the first server capacity; and

setting the valid interval as a function of the capacity of the selected server.

8. The method of claim 7, wherein the valid interval is directly proportional to the server capacity.

9. The method of claim 5, wherein said step of dynamically computing a valid interval for said mapping request to the first server comprises the steps of:

estimating the first server capacity; and

setting the valid interval as function of both a first server capacity estimate and a request load estimate associated with the group.

10. The method of claim 9, wherein the valid interval is inversely proportional to the estimated load associated with the source group and proportional to the server capacity.

11. In a distributed or clustered network of servers wherein clients are divided into groups which periodically send mapping requests to an arbitrator for mapping and balancing service requests among multiple replicated servers which can service the request, said arbitrator comprising:

computer executable scheduling logic means for mapping a first mapping request form a first group to a first server according to a schedule;

computer executable dynamic valid interval means, coupled to said scheduling logic means, for dynamically computing a valid interval for said mapping request to the first server as a function of one of a first group request load and a first server capacity; and

means for communicating both a first server identifier in response to said scheduling logic means and the valid interval, to the first group such that subsequent requests from the first group are routed to the first server during the valid interval.

12. The apparatus of claim 11, wherein each of the servers are web servers at geographically distributed sites, the arbitrator is an extended domain name server and each address mapping includes a time-to-live (TTL), and wherein the group consists of a set of clients associated with one of a common gateway and SOCKS server, said apparatus comprising:

said computer executable scheduling logic means comprises means for mapping a name request to an IP address; and

said computer executable dynamic valid interval means comprises means for dynamically computing the TTL as a function of one of a gateway request load and a capacity of a destination server associated with the IP address.

13. The apparatus of claim 11, wherein said computer executable scheduling logic means comprises means for mapping according to one of a round robin and multi-tier round robin schedule;

and wherein the valid interval is inversely proportional to the request load and proportional to the server capacity.

14. The apparatus of claim 11, wherein said computer executable scheduling logic means comprises means for mapping according to one of a probabilistic round robin and probabilistic multi-tier round robin schedule as a function of the server capacity;

and the valid interval is inversely proportional to the request load.

15. The apparatus of claim 11, wherein said computer executable dynamic valid interval means comprises:

means for estimating the request load, including one of a hidden or explicit load associated therewith, and directed to the first server; and

means for setting the valid interval as function of the request load associated with the group, in response to said estimating step.

16. The apparatus of claim 15, wherein said means for setting the valid interval as function of the request load comprises means for setting the valid interval as inversely proportional to the request load associated with the group.

17. The apparatus of claim 11, wherein said computer executable dynamic valid interval means comprises:

means for estimating the first server capacity; and

means for setting the valid interval as a function of the capacity of the selected server.

18. The apparatus of claim 17, wherein the valid interval is directly proportional to the server capacity.

19. The apparatus of claim 15, wherein said computer executable dynamic valid interval means comprises:

means for estimating the first server capacity; and

means for setting the valid interval as function of both a first server capacity estimate and a request load estimate associated with the group.

20. The apparatus of claim 19, wherein the valid interval is inversely proportional to the estimated load associated with the source group and proportional to the server capacity.