KR101463468B1 - Cluster structure based on omg standard middleware and compatible omg standaed dds - Google Patents

Abstract

The present invention relates to an OMG standard middleware-based cluster structure compatible with an OMG standard data distribution service (DDS) in a cluster in which a computer system is composed of a plurality of nodes. The present invention comprises the nodes which comprise copy caches comprising a data storage class, a copy class, and a node managing device and support the OMG standard DDS, and a legacy node which supports the OMG standard DDS without having the copy caches. When the legacy node is newly registered in the cluster structure, the copy class transmits data inputted from the data storage class to other nodes and the legacy node through a middleware operated with the OMG standard DDS and inputs the data transmitted from the other nodes and the legacy node in the data storage class.

Description

[0001] CLUSTER STRUCTURE BASED ON OMG STANDARD MIDDLEWARE AND COMPATIBLE OMG STANDARDED DDS [0002] The present invention relates to a cluster structure based on OMG standard middleware compatible with OMG standard DDS (Data Distribution Service)

The present invention provides a cluster structure based on OMG standard middleware compatible with OMG standard DDS (Data Distribution Service) in a cluster in which a plurality of nodes constitute one computer system.

Generally, in a cluster where multiple nodes are connected by a network, a time delay of "data request time ~ data response time" occurs when one node uses updated data from another node.

A replicated cache is designed to overcome the time delay. The replication cache defines the data items to be shared among the cluster nodes, and when a data item changes in one node, the changes are replicated in real time to the memory of another node in the cluster. As a result, each node is able to take data from the node's own memory without accessing the network since the latest data is already reflected in the memory before the demand of the data occurs.

However, since only the current concept of the duplicate cache is disclosed, each manufacturer manufactures and manages the duplicate cache middleware individually in accordance with the types and characteristics of the nodes. As a result, the replica cache manufactured by a specific manufacturer can communicate only with nodes having the same replicated cache middleware, does not use the replicated cache, and communicates with a node operating in a specific standard (for example, OMG DDS standard) .

Therefore, conventionally, when introducing a node (a legacy node) that does not use a duplicate cache in a cluster using a duplicate cache, inter-node compatibility is not ensured, which is disadvantageous to expansion of the cluster.

Accordingly, it is an object of the present invention to provide a cluster structure based on OMG standard middleware that can provide interoperability with OMG DDS standard by defining replicated cache middleware and providing instantness of data acquisition and development productivity provided by replicated cache middleware have.

Another object of the present invention is to provide a cluster structure based on OMG standard middleware that can provide interoperability with legacy nodes or future introduction nodes using OMG DDS by defining duplicate cache middleware capable of interoperating with OMG DDS standard .

In order to achieve the above object, a cluster structure based on OMG standard middleware according to an embodiment of the present invention includes a duplication cache including a data storage layer, a replication layer, and a node manager, and includes an OMG standard DDS (Data Distribution Service) A plurality of supporting nodes; And a legacy node that does not have a duplicate cache but supports an OMG standard DDS, wherein the replica cache of each of the nodes is operable when the legacy node is newly joined to the cluster structure, The data input from the data storage layer is propagated to the other node and the legacy node through the middleware, and the data propagated from the other node and the legacy node are input to the data storage layer.

The middleware includes a Publisher for issuing a corresponding content to a topic when data is changed in a data storage layer and a Subscriber for receiving data issued by the other node to the topic.

The node manager of each node registers a legacy node as a new node when the legacy node is newly joined and the received node ID and startup time are received from the node and the received value is compared.

If the number of times that the heartbeat message is not received within the set time from the specific node, the node manager of each node judges that the node has withdrawn and removes the node from the node list.

The node manager of each node checks the start time of each node through the node list, and only the node manager of the node having the fastest start time transmits the node ID and the start time to the legacy node.

Each node issues a change content provided by a data storage layer through a publisher at a time of data change, a legacy node receives data changed from the topic through a subscriber, and performs communication between each node and a legacy node ,

The subscriber of each node is subscribed to the same domain and topic as the publisher of the legacy node.

The legacy node publishes a message to a subscribed topic, the subscriber of each node receives the message, writes the message to the data storage layer, and performs communication between the legacy node and each node. And subscribed to the same domain and topic as the subscriber of the node.

The present invention can extend the cluster seamlessly to legacy, which introduces a duplicate cache in the cluster to perform communication with the OMG DDS, and can update a node that does not need or can not mount the duplication cache If the new node is communicating with the OMG DDS, it is also possible to smoothly expand the new node even if it is introduced into the operating cluster. Therefore, the design / administrator of the cluster can obtain the freedom of the system configuration whether or not the copy cache is mounted.

FIG. 1 is a diagram illustrating a cluster structure based on OMG standard middleware according to an embodiment of the present invention; FIG.
FIG. 2 is an exemplary diagram illustrating a data sharing process according to the present invention. FIG.
3 is a diagram illustrating a process of adding a new node according to the present invention;
4 is an exemplary diagram illustrating a method of communicating between a replicated cache cluster and a non-replicated cache node of the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings, wherein like reference numerals are used to designate identical or similar elements, and redundant description thereof will be omitted.

The present invention defines replication cache middleware interoperable with the OMG DDS standard to provide interoperability of legacy nodes or future introduction nodes using OMG DDS with instantness of data acquisition and development productivity provided by the replicated cache middleware I suggest a solution.

FIG. 1 illustrates a cluster structure based on OMG standard middleware according to an embodiment of the present invention.

FIG. 1 illustrates the configuration of the data storage layer 10 and the replication layer 11 and interworking with the node 200 that does not use the replication cache. In FIG. 1, each node supporting the OMG standard DDS includes a duplicate cache 100, and the legacy node 200 supports OMG standard DDS, but does not include a duplicate cache. The legacy node 200 may be a DDO application.

In each node, the replication cache 100 is composed of a data storage layer 10, a replication layer 11, and a node manager 12.

The replication cache 100 provides an application with a programmable interface to the data storage layer 10. There are two types of application programming interfaces (APIs) provided to the data storage layer 10, one is a data write interface and the other is a data update notification callback interface.

The replication layer 11 is provided with a duplicated cache middleware that can interoperate with the OMG DDS standard and has a function of propagating data input from the data storage layer 10 to another node and a function of transmitting data propagated from another node to a data storage layer 10).

The node manager 12 monitors and manages the state of the duplicate cache mounting node joined to the cluster.

That is, the data storage layer 10 performs data input and callback execution functions from the DDS application 300, and the replication layer 11 includes a new node cache synchronization function, a cache replication function, a non- And the node manager 12 performs a new node detection function, a leaving node detection function, and a primary node determination function.

FIG. 2 is an exemplary diagram illustrating a data sharing process according to the present invention.

2, when a data change occurs in the node 2, the replication layer 11 interfaces with the data storage layer 10 to receive the corresponding data from the data storage layer 10, 1, 3), and each replication layer 11 of the other node inputs data received from the node 2 to the data storage layer, thereby performing data sharing among the nodes.

Hereinafter, specific functions and operations of the node manager 12 will be described.

New node  Detection function

3 is an exemplary diagram illustrating a process of adding a new node according to the present invention.

As shown in FIG. 3, the node manager of the duplicate cache multicasts its node ID and startup time to the network after application program startup (1).

When the node ID and startup time of the new subscription node (node 3) are received, the replica cache node manager of the other node (nodes 1 and 2) compares the received node list with the existing node list, 3) in the node list (2).

Next, the node manager of the existing nodes (nodes 1 and 2) transmits its node ID and start time to the new subscriber node (node 3) so that each node of the cluster grasps the existence of each other (③).

Each node (node 1 to node 3) of the cluster periodically exchanges heartbeat messages with each other to maintain cluster affiliation (4).

Detection Node Detection Function

If the node manager does not reach the set number of times (for example, three times) within a predetermined time (for example, one second), the node manager determines that the node has left the node and removes the node from the node list.

Primary node determination function

When all the nodes perform synchronization in the active mode, when the new node joins the cluster, there arises a problem that all existing nodes in the cluster repeatedly transmit the same data to the new subscription node. If the number of nodes is N and the number of data types is M, N × M times of data are replicated. To avoid this problem, there is only one primary node in the cluster, and only this node performs the function of performing synchronization to the new subscriber node.

Therefore, the node manager of each node checks the start time of each node using the node list, and confirms it as the main node if it has the fastest start time.

The specific functions and operations of the copy layer 11 are as follows.

New node cache synchronization function

When detecting a new subscription node, the node manager 12 of the decentralized cache 100 evaluates whether it is the primary node itself, and if the primary node is correct, transmits data to the new subscription node through the replication layer 11. That is, only one of the nodes in the cluster transmits the cache synchronization data to the new subscription node.

That is, in FIG. 3, when the node ID and startup time of the new subscription node (node 3) are received, the replica cache node manager of the other node (nodes 1 and 2) compares the received node list with the received node number, After registering the new node (node 3) in the node list (2), it confirms that it is the main node having the fastest startup time. As a result, if node 1 is the main node, only node manager of node 1 sends its node ID and start time to new subscriber node (node 3) (③).

Cache Replication Method

The application program registers the variable in the data storage layer 10 of the duplication cache 100. The data storage layer 10 detects a data write event in a registered variable and transmits the data to the replication layer 11, and the replication layer 11 immediately transmits the data to other nodes connected to the network .

The replication layer 21 of the node receiving the data transfers the data to the data storage layer 20 of the same node and the data storage layer 20 updates the registered variables to the received value.

Non-replication  Communication with Cache Node

4 is an exemplary diagram illustrating a method of communication between a replicated cache cluster and a non-replicated cache node of the present invention.

As shown in FIG. 4, the data transmission from the replication cache node to the external non-replication cache node is performed when a data change is made in one node (node 1) in the replication cache cluster, And issues the change contents to the registered Topic (1). At this time, the publisher and the subscriber in the replication layer 11 form the middleware according to the present invention.

The subscriber of node 3 in the external non-replicated cache node receives the changed data from the topic (2). At this time, in order for the external non-replicated cache node to receive the data change of the replicated cache cluster, the subscriber of the node 3 must be subscribed to the same domain (main) and topic as the publisher of the nodes 1 and 2.

On the other hand, data transmission from the external non-replicated cache node to the replicated cache node begins by issuing a message to the Topic to which the external node's Publisher subscribes (③). Therefore, node 1 and node 2 in the replica cache cluster receive and record the corresponding message through the subscriber, respectively (④). At this time, in order to write the data transmitted from the non-replicated cache node to the node of the distributed cache cluster, the Publisher of the node 3 is supposed to be subscribed to the same domain and topic as the subscriber of the nodes 1 and 2.

As described above, according to the present invention, clusters can be seamlessly extended to a legacy that performs communication with an OMG DDS by introducing a replicated cache into a cluster. In addition, even if a node that does not need or can not mount a duplicate cache is introduced into a cluster that is newly operating a duplicate cache, smooth expansion is also possible if the new node communicates with OMG DDS. Therefore, the design / administrator of the cluster can obtain the freedom of the system configuration whether or not the copy cache is mounted.

Meanwhile, since the present invention includes a main node determination function between duplicate caches, a duplication system including 1 + 1 or 1 + N can be implemented.

It is to be understood that the configurations and methods of the embodiments described above are not limitedly applied, but that the embodiments may be embodied in other specific forms without departing from the spirit or essential characteristics thereof. Therefore, it should be understood that the above-described embodiments are to be considered in all respects as illustrative and not restrictive.

10: Data storage layer 11: Replication layer
12: Node manager 100: Replication cache
200: Legacy node

Claims (7)

In a cluster structure in which a plurality of nodes are connected via a network,
A plurality of nodes having a replication cache including a data storage layer, a replication layer, and a node manager to support an OMG standard DDS (Data Distribution Service); And
A legacy node that does not have a replication cache but supports an OMG standard DDS,
When a legacy node joins a cluster structure, the replication layer propagates data input from the data storage layer to other nodes and legacy nodes through middleware interoperable with the OMG standard DDS predetermined standard, And the data is transmitted to the data storage layer. The cluster structure based on OMG standard middleware. The system according to claim 1, wherein the middleware
And a subscriber for receiving the data issued by the other node and the other node to the topic when the data is changed in the data storage layer, and a cluster structure based on the OMG standard middleware. The method of claim 1, wherein the node manager of each node
When the legacy node is newly joined and the node ID and the startup time are received from the node, the node list and the received value are compared to register the legacy node as a new node if the node ID and startup time are not duplicated. . The method of claim 1, wherein the node manager of each node
If the heartbeat message is not received from the specific node a predetermined number of times within a predetermined time, it is determined that the node has withdrawn and the corresponding node is removed from the node list, based on the OMG standard middleware. The method of claim 1, wherein the node manager of each node
And the node manager of the node having the fastest startup time only transmits the node ID and the startup time to the legacy node, and the cluster structure based on the OMG standard middleware. 2. The method of claim 1, wherein each node
A legacy node receives data changed from the topic through a subscriber to perform communication between each node and a legacy node,
Wherein the subscriber of each of the nodes is subscribed to the same domain and topic as the publisher of the legacy node, and the cluster structure based on the OMG standard middleware. 2. The method of claim 1, wherein the legacy node publishes a message to a subscribed topic, the subscriber of each node receives the message and writes the message to the data storage layer to perform communication between the legacy node and each node,
Wherein the publisher of the legacy node is subscribed to the same domain and topic as the subscriber of each node, and the cluster structure based on the OMG standard middleware.

Images