KR20040029730A - Network management system using software distributed shared memory and method therefor - Google Patents

Abstract

PURPOSE: A network management system using a software distributive shared memory and a method therefor are provided to perform implicit communication for equipment information and state collection information through a shared memory in a network management system layer, thereby easily exchanging information. CONSTITUTION: A central database server(10) assigns equipment information on various equipments, and stores changed equipment information in a shared memory region. Gateways(20,30) store collection information on various equipment states in shared memory regions, and transmit the collection information to the central database server(10). Software distributive shared memories(40) include many memories(41-43), and supply one memory address space. The central database server(10) comprises as follows. The first processes(P0-P3) drive each program. The first software cache(11) stores a memory transmitted from a remote node, and reads the memory. The first shared memory(41) stores the equipment information and the collection information.

Description

Network management system using software distributed shared memory and method therefor}

The present invention relates to a network management system using a software distributed shared memory and a method thereof, and to efficiently and easily communicate between nodes using a software distributed shared memory between a plurality of server or workstation nodes constituting the network management system. The present invention relates to a network management system using software distributed shared memory and a method thereof.

1 is a block diagram of a conventional IP (Internet Protocol) network management system.

The conventional network management system includes a central DB server 1 including a database 2 and a plurality of gateways connected to the central DB server 1. In FIG. 1, gateways 3, 4, and 5 are provided. It will be described in the embodiment.

Each gateway 3 to 5 is connected to the central DB server 1 via a network and to a plurality of devices to be monitored.

The gateway 3 is connected to a router 6 and a NAS (Network Access Server) 7, the gateway 4 is connected to a switch 8, and the gateway 5 is connected to a plurality of APs. It is connected to an access point and communicates through ICMP (Internet Control Message Protocol) and SNMP (Simple Network Management Protocol).

Here, the gateways 3 to 5 receive information on the equipment from the central DB server 1, collect the equipment status, information, and traffic volume through ICMP or SNMP from various equipment to the central DB server 1 send.

However, the conventional network management system requires a network communication module using a socket or the like for communication between the central DB server 1 and the gateways 3 to 5 or for communication between the gateways 3 to 5. Then, the API (Application Program Interface) in the module is called to perform mutual communication.

Therefore, since the communication API is called every time due to the frequent communication between the gateways 3 to 5 and the central DB server 1, the network load increases, thereby reducing the efficiency of network management.

The present invention was created to solve the above problems, and provides a network management system developer by easily providing a single memory address space to a user on a physically distributed memory to easily read and write the memory in a remote node. The goal is to allow explicit communication calls to be omitted.

1 is a block diagram of a general IP network management system.

2 is a block diagram of a network management system using a software distributed shared memory according to the present invention.

3 is a flowchart of a network management method using a software distributed shared memory according to the present invention;

A network management system using the software distributed shared memory of the present invention for achieving the above object comprises: a central database server for allocating equipment information of various equipment to be monitored and storing it in a shared memory area when the equipment information is changed; A gateway for storing the collected information collecting the state of various equipments in a shared memory area and transmitting the collected information to a central database server; And a software distributed shared memory provided in the central database server and the gateway to provide a single memory address space in software including a plurality of shared memories shared through a hardware bus.

In addition, in the network management method using the software distributed shared memory of the present invention, in the network management method using a software distributed shared memory including shared memory between the central database server and the gateway, the central database server selects equipment information from the database. A first step of storing in a shared memory; Transmitting a barrier response signal along with the equipment information to the gateway upon application of the barrier call signal from the gateway; A third step of storing collection information collected from various equipments in a shared memory and transmitting a barrier call signal to a central database; And storing a collection information transmitted from the gateway in a shared memory and transmitting a barrier response signal to the gateway.

The present invention sees a network management system as one symmetric multiprocessing (SMP) cluster and provides one memory address space in the network management system using a software distributed shared memory algorithm for the SMP cluster.

Therefore, the network management system developer can communicate with the remote node by reading and writing to the memory address space directly without explicit communication API call.

Hereinafter, with reference to the accompanying drawings will be described in detail an embodiment of the present invention.

2 is a block diagram of a network management system using a software distributed shared memory according to the present invention.

The present invention includes a central database server (10), a plurality of gateways (20, 30), a router (60), a NAS (Network Access Server; network access server; 70), and a switch through a connection network (50). 80 and equipment such as an AP (Access Point; 90).

Here, the central DB server 10 includes a plurality of processes P0 to P3, a software cache 11 and a shared memory 41.

The gateway 20 includes a plurality of processes P4 and P5, a software cache 21, and a shared memory 42. The gateway 30 includes a plurality of processes P6 and P7 and a software cache 31. And a shared memory 43. In addition, each shared memory 41 to 43 includes one software distributed shared memory 40.

In the embodiment of the present invention, the process is illustrated as P0 to P7, but the present invention is not limited thereto and the process may be configured variably.

Definition of terms used in the present invention having such a configuration is as follows.

First, a node is defined as a single server or workstation with independent processors, memory, and I / O buses.

In addition, the gateways 20 and 30 may provide equipment information through ICMP (Internet Control Message Protocol) or SNMP (Simple Network Management Protocol) from each device among the nodes constituting the network management system. It is defined as a node that serves to collect and deliver to the central DB server (10).

In addition, a symmetric multiprocessor (SMP) node is defined as a node in which several processors share a physical memory and an input / output bus, and one operating system (OS) manages all the processors.

Also, a local node is defined as its own node to which a process belongs and is executed in the context of a particular process. A remote node is any node other than the local node and is defined as a node that must traverse the network to communicate with any process in the node.

In the present invention, the entire network management system is viewed as one cluster computer, and the central DB server 10 and each gateway 20 and 30 shown in FIG. 2 become an SMP node.

In addition, although a plurality of processes P0, P1 .. P7 may be considered as a processor (CPU) in a general cluster system, the processes are represented in the present invention. That is, the number of processors and the number of processes do not match, and there are usually more processes in a node than the number of processors.

In addition, each node has shared memories 41 to 43 shared between processes P0 to P7, and the shared memories 41 to 43 are shared in hardware and connected through a hardware bus. In addition, the shared memories 41 to 43 in the same node are maintained consistent by a hardware cache consistency model.

It is assumed that all nodes are connected to the connection network 50, and the network management system model of the present invention is connected to a wide area network such as a wide area network (WAN) or the Internet. Therefore, each shared memory 41 to 43 in each node is physically distributed and cannot be shared unless special work is performed.

In addition, the software distributed shared memory 40 virtually provides a single memory address space between the shared memories 41 to 43.

In addition, the present invention has a form of non-uniform memory access (NUMA), in which the time of accessing the memory of the local node is much shorter than the time of accessing the memory of the remote node.

Thus, each node has software caches 11, 21, and 31 for the purpose of resolving access time mismatches. The software caches 11, 21, and 31 are physically part of memory. When the contents of the memory in the remote node are read, the software caches 11, 21, and 31 are stored in their software cache and continuously read.

Therefore, the present invention allocates IP addresses, interfaces, and SNMP OIDs (Object IDs) of devices to which the central DB server 10 should collect information by dividing each gateway 20 and 30. When the equipment is added, the added equipment is allocated to a specific gateway 20, 30, which uses the software distributed shared memory 40 without using a network communication API such as a socket.

That is, the information required for the gateways 20 and 30 can be delivered by simply writing to the shared memory 41 to 43 address space as well as simply reading and writing to the local memory.

On the other hand, Figure 3 is a flow chart for consistency in the network management method using a software distributed shared memory according to the present invention.

The present invention uses a Lazy Release Consistency (LRC) memory coherency model that is commonly used in software distributed shared memory to reduce network load.

That is, because the memory is shared, a synchronization method is needed to prevent the same memory area from being accessed by different processes at the same time. To this end, the software distributed shared memory uses a synchronization method of lock and barrier.

In the LRC, locks and barriers are used as a means of transmitting and receiving memory coherence information as well as synchronization. In the present invention, an operation process for matching memory coherence is as follows.

First, the central DB server T0 selects the device information from the database when the distributed shared memory is initialized and the shared memory area is allocated (step 100). Here, the device information is a router, a switch, etc., which is a management target for network management. IP address, interface, OID, SNMP community, etc.

Thereafter, the central DB server T0 stores the selected monitoring target device information in the shared memory 41 area (step 110). Then, the central DB server T0 stores the device information in the shared memory 41 area and then stores the device information in the shared memory 41 area. Delete the previous monitoring result.

A barrier is a synchronization method for synchronizing the central DB server T0 with the gateways T1, T2 and T3 while matching the contents of the shared memory 41 to 43 of the central DB server T0 with the gateways T1 to T3.

Here, the central DB server T0 waits until all other processes call the barrier Barrier () even if the barrier Barrier () is called first.

Accordingly, the gateways T1 to T3 transmit the barrier call state to the central DB server T0 through the barrier call signal BARR (step 120), and the central DB server T0 indicates that all processes reach the barrier state through the barrier response signal BARRGRANT. A message is sent to the gateways T1-T3 to continue the process according to synchronization (step 130).

In this case, each process has time information called a timestamp. The timestamp is used as a time meaning, and compares a timestamp to determine a predecessor relationship such as how far other processes have progressed. do. This time stamp is incremented by 1 in the barrier, lock, and unlock.

In addition, all shared memory pages have a vector time stamp. When a memory access occurs, the cache time is determined by comparing the vector time stamp of the current process with the vector time stamp of the shared memory page.

If new contents are needed, memory coherency is maintained by updating the cache of the local node by executing a shared memory page request from the corresponding process of the remote node.

In addition, a write-notice transmitted along with the barrier response signal BARRGRANT is (T0,0, {equipment information}), and transmits the instrument information to each gateway T1 to T3.

Accordingly, each of the gateways T1 to T3 collects equipment information through ICMP and SNMP, and records the collected information in each shared memory 42 to 43 (step 140).

The central DB server T0 barrier call status is transmitted through the barrier call signal BARR. (Step 150) The central DB server T0 transmits a PING (Packet Internet Groper; packet Internet Groper) to the database when the barrier call signal BARR is transmitted from the gateways T0 to T3. And information collected such as equipment failure, traffic information, session information, and the like, collected through SNMP (stepper 160), and the barrier response signal BARRGRANT is transmitted to the gateways T1 to T3 (step 170).

At this time, the central DB server T0 includes the write notices T1,1, {collection information}, (T2,1, {collection information}), and (T3,1, {collection information}) together with the barrier response signal BARRGRANT. It transmits to the gateways T1 to T3, and the write notice (T1,1, {collection information}) indicates that the process gateway T1 stored {collection information} in the time stamp 1 in the shared memory.

Here, the method of representing the actual value to Lightnotes is called an update protocol, and only the stored information is sent to Lightnotes, and the actual value is transmitted when a process (T2, T3, etc.) attempts to read the value. This method is called an invalidation protocol.

As described above, in the development of the network management system, the shared memory is not considered in the form of maintaining the memory coherency in the software distributed shared memory 40 and referring to the remote node's memory or the local node's cache. When reading and writing information to shared memory by allocation, it synchronizes only depending on how the lock and barrier are synchronized.

In this manner, communication between the central DB server T0 and the gateways T1 to T3 is performed by implicit communication by the software distributed shared memory 40 without using a network module such as a socket API.

On the other hand, in a general cluster computer, nodes are connected by using a high-speed fast network, whereas the network management system hardware of the present invention is a broadband communication network in which each node is geographically dispersed and is not a high-speed LAN. (WAN), connected to the Internet.

In such an environment, the application of software distributed shared memory should use as little network bandwidth as possible and the smallest amount of message passing through the network.

Therefore, network messages using a relaxed memory model such as lazy release consistency (LRC) or home-based lazy release consistency (HLRC), among the memory consistency models for matching the consistency of software distributed shared memory. The delivery volume should be reduced.

In addition, it reduces network load by providing distributed shared memory through hardware memory sharing on its own, rather than through a network.

As described above, the present invention reduces the number of explicit communication using software distributed shared memory and enables easy network management system development because it is intuitively easy to understand compared to a programming model for delivering messages over a network. . In addition, even if the number of gateways is variable, it provides an effect that can easily communicate between nodes.

Claims (11)

A central database server for allocating device information of various devices to be monitored and storing the device information in a shared memory area when the device information is changed; A gateway for storing the collected information collecting the state of the various equipments in a shared memory area and transmitting the collected information to the central database server; And Software distributed shared memory, characterized in that provided in the central database server and the gateway, each software distributed shared memory to provide a single memory address space in software including a plurality of shared memory shared through a hardware bus Network management system using. The method of claim 1, wherein the central database server A plurality of first processes for driving respective programs; A first software cache for storing and reading the memory transmitted from the remote node; And And a first shared memory configured to store the device information selected from a database and the collection information transmitted from the gateway. The method of claim 1, wherein the gateway is A plurality of second processes for driving respective programs; A second software cache for storing and reading the memory transmitted from the remote node; And And a second shared memory configured to store the equipment information transmitted from the central database and the collected information. The method of claim 1, wherein the software distributed shared memory A network management system using software distributed shared memory, characterized by maintaining consistency by hardware cache coherence model. The method of claim 1, wherein the gateway is The network management system using a software distributed shared memory, characterized in that for collecting the device information through at least one of an Internet control message protocol, a simple network management protocol, a packet Internet gripper to transmit the collected information to the central database server. . The method of claim 1 or 5, wherein the collected information is Network management system using a software distributed shared memory, characterized in that the failure of the various equipment, traffic information, session information. The method of claim 1, wherein the equipment information Network management system using a software distributed shared memory, characterized in that the Internet protocol address, interface, object ID and simple network management protocol of the various equipment to be managed for network management. The method of claim 1, And a shared memory of the central database server and the gateway is synchronized through a barrier and lock method. In the network management method using software distributed shared memory including shared memory between central database server and gateway, A first step of the central database server selecting equipment information from a database and storing the equipment information in the shared memory; A second step of transmitting a barrier response signal together with the equipment information to the gateway when the barrier call signal is applied from the gateway; Storing the collected information collected from various equipments in the shared memory and transmitting the barrier call signal to the central database; And And storing a collection information transmitted from the gateway in the shared memory and transmitting the barrier response signal to the gateway. The method of claim 9, wherein the second step And transmitting a vector time stamp signal including a time information together with the barrier call signal to indicate a prior relationship of the process progression. The method of claim 9, wherein the second step And transmitting a write notice signal together with the barrier response signal.