Server decision method and device thereof

• US 20020116437A1
• Filed: 04/05/2002
• Published: 08/22/2002
• Est. Priority Date: 01/31/2000
• Status: Active Grant

• Current Assignee: Fujitsu Limited
• Original Assignee: Fujitsu Limited
• Inventors: Saito, Fumio, Kambara, Junichi, Kanaoka, Ei
• Granted Patent: US 7,127,484 B2
• Time in Patent Office: Days
• US Class Current: 718/102
• CPC Class Codes: G06F 2209/505 Clust
• CPC Class Codes: G06F 9/5061 Partitioning or combining of resources

Abstract

Each node of the cluster system is classified into the main server candidate node group capable of operating the specific task, the sub-server candidate node group capable of operating the specific task and the back up task thereof and the out of candidate node grope which is excluded from the operation of the specific task and the back up operation thereof. Each of the servers belonging to the main candidate node group presents its public relations information necessary for the election judgment of the main server to all the node and runs as a candidate at a boot time and, at the same time, compares its public relations information to the public relations information of other nodes and, when it determines that it is not an adequate candidate, stops the presentation of its public relations information and withdraws its candidacy and allows the only main server approved by other nodes to exist. Each of the servers belonging to the sub-server candidate node group similarly allows the only sub-server approved by other nodes to exist.

Claims

1. A node decision method which decides the existence of the only server to operate a certain specific task within a plurality of nodes connected via a network,

the node decision method classifying said plurality of nodes into a main server candidate node group capable of operating said specific task, a sub-server candidate node group capable of the back-up operation of said specific task and the out of candidate node group which is excluded from the operation of said specific task and the back-up operation thereof,

wherein each of the servers belonging to said main server candidate node group presents its public relations information necessary for the election judgment of the main server to all other nodes and runs for a candidate at a boot time and, at the same time, compares its public relations information to the public relations information of other nodes and, when it determines that it is not an adequate candidate, stops the presentation of its public relations information and withdraws its candidacy,

wherein each of the servers belonging to said sub-server candidate node group presents its public relations information necessary for the election judgment of the sub-server to all other nodes and runs for a candidate at a boot time and, at the same time, compares its public relations information to the public relations information of other nodes and, when it determines that it is not an adequate candidate, stops the presentation of its public relations information and withdraws its candidacy.

2. The node decision method according to claim 1, wherein said main server or the sub-server periodically reports the operation of said specific task to other nodes and asserts the right of the main server or the sub-server and, when it receives the same report to assert the right of the main server or the sub-server from other nodes, it recognizes the duplicate-starting of a plurality of main servers or a plurality of sub-servers and presents its public relations information to all the nodes and runs for a candidate and, at the same time, compares its public relations information to the public relations information of other nodes and, when it determines that it is not an adequate candidate, stops the presentation of its public relations information and withdraws its candidacy.

3. The node decision method according to claim 2, wherein

said main server or sub-server periodically reports the operation of said specific task to other nodes and asserts the right of the main server or the sub-server and, when it receives the same report to assert the right of the main server or the sub-server from other nodes, it recognizes the existence of a plurality of main servers or sub-servers and presents its public relations information to all the other nodes and runs for a candidate and, when it determines that the other nodes are adequate for the main server or the sub-server by comparing its public relations information to the public relations information from other nodes, stops the specific task at an adequate time and hands over the right of the main server or the sub-server to the other node.

4. The node decision method according to claim 1, wherein said main server candidate node group is a node group which has resources necessary to operate said specific task or which is desired to positively operate said specific task,

wherein said sub-server candidate node group is a node group which does not have enough resources to operate said specific task or which is desired to passively operate said specific task,

and wherein the out of candidate node group is a node group which does not have enough resources to operate said specific task or which is not desired to operate said specific task.

5. The node decision method according to claim 1, wherein said public relations information is a duty cycle of resources and an allowance rate of distributable resources is obtained from the requirement volume of said specific task and the duty cycle of said public relations information as an election probability and, when the election probability of a candidate is smaller than the election probability of other nodes, the candidate stops the presentation of its public relations information and withdraws its candidacy.

6. The node decision method according to claim 5, wherein said allowable rate is obtained for each type of the resources and the smallest allowable rate is taken as the election probability.

7. The node decision method according to claim 1, wherein the specific task of said main server is a management task to perform setting and monitoring of each node and the specific task of said sub-server is a task to back up the management task of the main server.

8. A server decision device which decides the existence of the only server to operate a certain specific task within a plurality of nodes connected via a network,

the server decision device classifying said each server into the main server candidate node group capable of operating said specific task, the sub-server candidate node group capable of operating the back-up task of said specific task and the out of candidate node group which is excluded from the operation of said specific task and the back-up operation thereof,

wherein a main server processing unit is provided in which each of the nodes belonging to said main server candidate node group presents its public relations information necessary for the election judgment of the main server and runs as a candidate at a boot time and, at the same time, compares its public relation information to the public relations information of other nodes and, when it determines that it is not an adequate candidate, stops the presentation of its public relations information and withdraws its candidacy,

and wherein a sub-server processing unit is provided in which each of the nodes belonging to said sub-server candidate node group presents its public relations information necessary for the election judgment of the sub-server to all the nodes and runs as a candidate at a boot time and, at the same time, compares its public relations information to the public relations information of other nodes and, when it determines that it is not an adequate candidate, stops the presentation of its public relations information and withdraws its candidacy.

9. The server decision device according to claim 8, the main server candidacy processing unit of the main server and the sub-server candidacy processing unit of the sub-server which were selected at the boot time periodically report the operation of said specific task to all the other nodes and assert the right of the main server or the sub-server at the operation starting time after boot and, when they receive the same report to assert the right of the main server or the sub-server from other nodes, they recognize the duplicate starting of a plurality of main servers or sub-servers and present their own public relations information to all the other nodes and, at the same time, compare their public relations information to the public relations information of other nodes and, when they determine that they are not adequate candidates, stop the presentation of their public relations information and withdraw their candidacies.

10. The server decision device according to claim 9, wherein, during the operation, said main server candidacy processing unit or the sub-server candidacy processing unit periodically report the operation of said specific task to other nodes and asserts the right of the main server or the sub-server and, when it receives the same report to assert the right of the main server or the sub-server from other nodes, it recognizes the existence of a plurality of main servers or a plurality of sub-servers and presents its public relations information to all the nodes and runs as a candidate and, at the same time, when it determines that the other node is an adequate candidate by comparing its public relations information to the public relations information of the other node, stops the specific task at an adequate time and hands over the right of the main server or the sub-server to the other node.

11. The server decision device according to claim 8, wherein said main server candidate node group is a node group which has necessary resources to operate said specific task or which is desired to positively operate said specific task,

wherein said sub-server candidate node group is a node group which has necessary resources to operation said specific task or which is desired to passively operate said specific task,

and wherein the out of candidate node group is a node group which has not the necessary resources to operate said specific task or which is not desired to operate said specific task.

12. The server decision device according to claim 8, wherein said public relations information is the duty cycle and said main server candidacy processing unit or sub-server candidacy processing unit obtains the duty cycle of distributable resources from the requirement volume of said specific task and the duty cycle of said public relations information as an election probability and, when its election probability of a candidate is smaller than the election probability of the other node, it stops the presentation of its public relations information and withdraws its candidacy.

13. The server decision device according to claim 12, wherein said main server candidacy processing unit or sub-server candidacy processing unit obtains said allowable rate for each type of the resources and the smallest allowable rate among them is taken as an election probability.

14. The server decision device according to claim 8, wherein the specific task of said main server is a management task to perform setting and monitoring of each node of said cluster system, and the specific task of said sub-server is a task to back up the management task of the main server.

15. A node decision method which decides the existence of the only server to operate a certain specific task within a plurality of nodes connected via a network,

said each server being constituted by the server candidate node group capable of operation said specific task and the out of candidate node group which is excluded from the operation of said specific task,

wherein each of the nodes belonging to said node group presents its public relations information necessary for the election judgment of a specific task server to all the other nodes and runs as a candidacy and, at the same time, compares its public relations information to the public relations information of other nodes and, when it determines that it is not an adequate candidate, stops the presentation of its public relations information and withdraws its candidacy.

16. A node decision device which decides the existence of the only server to operate a certain specific task within a plurality of nodes connected via a network,

the node decision device divides said plurality of nodes into the server candidate node group capable of operating said specific task and the out of candidate node group which is excluded from the operation of said specific task,

wherein each of the nodes belonging to said server candidate node group presents its public relations information necessary for the election judgment of a specific task server to all the other nodes and runs as a candidate and, at the same time, compares its public relations information to the public relations information of other nodes and, when it determines that it is not an adequate candidate, stops the presentation of its public relations information and withdraws its candidacy.