KR20110046897A - Method to elect master nodes from nodes of a subnet - Google Patents

Abstract

PURPOSE: A method for selecting a master node from the nodes of a subnet is provided to promptly select a master node through the volunteer vote within a unit time between network nodes. CONSTITUTION: A candidate node is selected. The candidate node is determined as a node having an IP(Internet Protocol) address which is the smallest node within a subnet. When having the least Internet protocol address, the node is transited to a candidate node status. The node becomes the candidate node, and the other nodes are transited to a vote node state. The other nodes start voting.

Description

How to elect a master node on a subnet {METHOD TO ELECT MASTER NODES FROM NODES OF A SUBNET}

The present invention relates to a method for selecting a master node in a subnet, and more particularly, a method for selecting a master node for collecting and managing information (subnode states, link state information) of a subnet to which one belongs to a subnet. It is about.

The network infrastructure must be scalable to a large number of nodes. In addition, the network should have a countermeasure against various types of node failures or unusable links. Ethernet is a realistic solution for open networks. There are two ways in which Ethernet-based networks can adapt to failure. One of them is to have a multi-port network interface card. The firmware of the network interface card detects the failure of the active port and activates the standby port. Another approach is to have software that detects the failure and performs the recovery algorithm. Most software approaches apply a periodic heartbeat mechanism.

This conventional heartbeat-based configuration inherently had a problem of scalability. Therefore, the present inventors proposed a multi-subnet architecture to solve this problem of scalability. Multi-subnet architectures divide the network into multiple subnets and limit the number of nodes a subnet can accept to provide scalability and fast failover. Thus, the network can continue to expand by adding subnets. All nodes on a subnet send and receive Ethernet-based heartbeats to each other. Nodes can detect and overcome failures in the path by checking heartbeats sent from other nodes. Since heartbeats are Ethernet frames and cannot be transmitted between subnets, a master node must be assigned and managed in each subnet to recover from failures between subnets. Master nodes communicate with each other using IP packets to exchange subnet states. When a failure occurs in one subnet, the master node plays an important role in detecting the failure through a heartbeat and informing the other master node to recover from the failure on the interconnected subnet communication path. In general, master nodes are assigned directly by an administrator in each subnet. Also, in a typical network system, node failures do not occur frequently, so failures are not considered. It is also common for network administrators to detect and recover from failures in the event of a failure. However, in special cases such as the network system of the ship, failure may occur due to the attack of the opponent during the battle, and the failure must be recovered quickly and automatically, so the master node is automatically selected quickly among the nodes instead of arbitrarily designated by the administrator. In addition, there is a need to designate a master node to replace a master node if it fails.

An object of the present invention is to solve the above problems and to provide a method for selecting a master node as quickly as possible on a subnet within a predetermined time and for designating a second master node in preparation for a failure of the master node. It is done.

The above object of the present invention is a method for selecting a master node among a plurality of normal nodes in one subnet in a network composed of a plurality of subnets having a limited number of nodes, wherein the plurality of normal nodes are selected as candidate nodes and voting nodes. The first step of classifying the candidate node and the candidate node in the second step of receiving a voting message from the voting node within a predetermined unit time, and when the number of voting messages received by the candidate node from the voting node becomes equal to or more than a predetermined ratio of the normal number of nodes. And a third step of selecting the candidate node as the primary master node.

Another object of the present invention is to select a node having the smallest internet protocol address as a candidate node among a plurality of normal nodes in one subnet in a network composed of a plurality of subnets having a plurality of nodes, and vote the remaining nodes. A method of selecting a candidate node as a master node in a corresponding subnet after selecting the node, comprising: a first step of setting a period of a certain MASTER_TIMER, a second step of receiving a voting message from a voting node, and a period of MASTER_TIMER A third step of counting whether the number of voting messages received from the voting node is equal to or greater than a predetermined ratio of the plurality of nodes, and when the ratio exceeds the predetermined ratio, a fourth step of deleting the MASTER_TIMER and setting the corresponding candidate node as a master node. Selecting a master node in the subnet, characterized in that including Achievement is also possible.

Compared to the conventional method in which the master node is artificially designated by the administrator, the method proposed in the present invention can select a master node quickly by autonomous vote within a unit time among network nodes, even when a failure occurs in the master node. The advantage is that you can immediately elect a new master node to carry out its mission.

In particular, the method of selecting a master node according to the present invention can be most advantageously applied to a network system in which a failure is fatal. In particular, the primary master node can designate the first master node or the third master node in advance. Even if the car master node fails, it can respond quickly.

Hereinafter, the advantages, features and preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a system configuration diagram applied to the present invention. The system to which the present invention is applied is composed of a plurality of subnets. If you want to expand the network, you can easily achieve this by adding subnets. One subnet consists of an appropriate number of nodes and LAN switches, which are connected by LAN switches to communicate with each other. The proper number of nodes included in one subnet is determined by the time and network complexity of setting the master node, which is not the object of the present invention, and thus detailed description thereof will be omitted. Each subnet selects at least one master node from among the normal nodes to which it belongs. Although one master node may be provided in one subnet, it is preferable to have a plurality of master nodes as the first master node, the second master node, and the third master node in one subnet to effectively cope with failure. have.

The master node is responsible for maintaining the information path between one subnet node and exchanging information with the master node of another subnet. Since the master node has a Fault Tolerant Ethernet (FTE) function, the heartbeat frame-based Ethernet is transmitted from another node in the subnet. By using the heartbeat received from another node, the available path information between the nodes can be known.

At startup or when all master nodes fail, each node in the subnet votes to select a master node. The case where the master node has failed will be briefly described. The master node according to the present invention periodically broadcasts a WIN message. If the nodes other than the master node do not receive this message within a predetermined timeout time, the node is determined to be a failure of the master node and the master node selection procedure is performed. Therefore, each node checks these conditions on its own and performs a master node selection procedure when the conditions are met. At this time, if it is not the node having the minimum IP address, it becomes a VOTE node, and sends a VOTE message to the candidate node. If it is a node with a minimum IP address, it becomes a candidate node and waits for a VOTE message.

2 is a state transition diagram for selecting a master node among nodes in one subnet according to the present invention. Each node starts in the normal node state. First, candidate nodes to be elected as master nodes are determined. The candidate node is defined as the node with the smallest Internet Protocol (IP) address in the subnet. In the present invention, since each node periodically broadcasts its own IP address after booting, it will be described on the assumption that it knows the IP addresses of other nodes. If it is the node with the smallest internet protocol address, it transitions to the candidate node state and becomes a candidate node. The other nodes transition to the voting node state and start voting.

3 is a C-language describing a state transition of a candidate node state, and FIG. 4 is a C-language describing a state transition of a voting node state. First, a state transition process from a candidate node state to a normal node state or a master node state will be described with reference to FIG. 3.

Set an arbitrary time called MASTER_TIMER (STEP 1). If the message received during the MASTER_TIMER period is VOTE, if the VOTE message is received from 50% of all normal nodes, the candidate node becomes the master node, so the timer is deleted and the WIN message is broadcasted (STEP 2). . If the message received during the period of MASTER_TIMER is WIN, it indicates that another candidate node has already been selected as the master node. Therefore, the timer is deleted and the node sending the message is determined as the master node (STEP 3).

If the candidate node receives the VOTE message after becoming the master node, it is ignored. After the candidate node becomes the master node and receives the WIN message and the candidate node's IP address is greater than the sender's IP address, the candidate node gives up its status as the master node and receives the WIN message but receives the corresponding candidate node. If the IP address is not greater than the sender's IP address, the corresponding node is ignored because the candidate node is the master node (STEP 4).

If MASTER_TIMER is finished, repeat Step 1 to STEP 4 2 or 3 times, and if the candidate node is not selected as a master node after repeating, repeat Step 1 to STEP 4 for the next candidate node ( STEP 5).

Next, a state transition process from the voting node state to the normal node state will be described with reference to FIG. 4.

The node having the smallest IP address is determined as a candidate node, a VOTE message is sent to the candidate node, and an arbitrary VOTE_TIMER period is set (ST 1).

During the VOTE_TIMER period, if a VOTE message is received, it is ignored (ST 2). If a WIN message is received during the VOTE_TIMER period, the node that sent the WIN message is determined as the master node (ST 3). When you have a master node, if you receive a VOTE message, ignore it. If you receive a WIN message, determine the node with the smallest IP address as the master node (ST 4).

If the VOTE_TIMER period ends, steps ST 1 to ST 4 are performed 2-3 times more. If the selection of the current candidate node as the master node fails, another candidate node is designated and steps ST 1 to ST 4 are repeated (ST 5).

Although specific embodiments of the present invention have been described and illustrated above, it will be apparent that various modifications may be made by those skilled in the art without departing from the technical spirit of the present invention.

1 is a system configuration diagram applied to the present invention.

2 is a state transition diagram for selecting a master node among nodes in one subnet in accordance with the present invention.

3 is an embodiment of a C-language code illustrating a state transition of a candidate node state.

4 is an embodiment of a C-language code illustrating a state transition of a voting node state.

Claims (12)

In a network composed of a plurality of subnets having a limited number of nodes, a method of selecting a master node among a plurality of normal nodes in one subnet, Classifying a plurality of normal nodes into candidate nodes and voting nodes; The candidate node receiving a voting message from the voting node within a predetermined unit time; And And selecting a candidate node as a first master node when the number of vote messages received by the candidate node from the voting node is equal to or greater than a predetermined ratio of the number of normal nodes. How to select. The method of claim 1, And in the first step, classifying the candidate node and the voting node is performed with an internet protocol address. 3. The method of claim 2, wherein the candidate node is selected as a node having the smallest Internet protocol address, and before the first step, each normal node periodically broadcasts information including its Internet protocol address after booting. The method for selecting a master node in the subnet, characterized in that further comprising step 1-1. The method of claim 1, And selecting a master node in the subnet, wherein the predetermined ratio of the third step is 50%. The method of claim 1, And further comprising a fourth step in which the first master node selected after the third step arbitrarily selects a second master node to be sequentially operated as a master node in preparation for its failure. How to select a master node. The method of claim 4, wherein And in the fourth step, the primary master node arbitrarily further selects the third master node in case the secondary master node fails. The method of claim 1, And after the third step, the first master node further comprises a fifth step of periodically transmitting a WIN message. In a network composed of multiple subnets with multiple nodes, the node with the smallest internet protocol address is selected as a candidate node among the plurality of normal nodes in one subnet, the remaining nodes are selected as voting nodes, A method of selecting the candidate node as a master node, Setting a period of time called a constant MASTER_TIMER; A second step of receiving a voting message from the voting node; Counting whether the number of voting messages received from the voting node is equal to or greater than a predetermined ratio of a plurality of nodes during the MASTER_TIMER period; And And a fourth step of deleting the MASTER_TIMER and setting a corresponding candidate node as a master node when the ratio exceeds a predetermined ratio in the third step. The method of claim 8, And selecting a master node in the subnet, wherein the predetermined ratio of the third step is 50%. The method of claim 8, And further comprising a fifth step in which the first master node selected after the fourth step arbitrarily selects a second master node to be sequentially operated as a master node in preparation for its failure. How to select a master node. The method of claim 10, And in the fifth step, the first master node selects a third master node arbitrarily in case of failure of the second master node. The method of claim 8, And after the fourth step, the first master node further includes a sixth step of periodically transmitting a WIN message.

Images