JPH09259096A - Network high reliability method and system - Google Patents

Abstract

(57) [Summary] (Modified) [PROBLEMS] Client consisting of active server that processes processing requests from clients, and standby server that substitutes the active server when a failure occurs in the active server. In the server system, when a failure occurs in the active server, the alternative process of the active server performed by the standby server is realized without the client being aware of it. SOLUTION: A means capable of setting and processing a plurality of addresses other than its own network address on a single network interface for an active server and a standby server, and the network address of the active server when the active server failure is detected. Means for setting on the interface and means for notifying the client that the server holding the address has changed are provided.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a LAN (Local
l Area Network) related to a system in which a client and a server that processes a request from the client are connected, and normally, the active server that processes the client request and the active server in place of the active server perform processing. The present invention relates to a network high reliability method and system in which a server in a standby system is made redundant, and when a failure occurs in the active server, the standby server substitutes the active server.

[0002]

2. Description of the Related Art In a conventional method for achieving high reliability by making a server redundant, as described in Japanese Patent Laid-Open No. 5-244164, a data processing device (server) operates in a main system (active system). In this case, different network addresses (information used by the communication processing unit to uniquely identify a device connected to the LAN) are automatically assigned depending on whether the operation is performed in the slave system (standby system). As a result, the redundant configuration of the server can be easily managed. As a communication protocol used for communication processing between devices connected to a network, a standardization organization IAB (Internet Architecture) is used.
e Board's IETF (Internet En)
RFC (Request for Comment) issued by the Gineering Task Force
TCP / IP registered as a standard protocol in
(Transmission Control Pro
Tocol / Internet Protocol) is widely used.

[0003]

The above-mentioned prior art is based on the LA
This is an effective method for automatically managing the network address assigned to each server in making the server redundant on N. However, because redundant servers are used effectively, it cannot be applied when all servers are operated in the same state (the standby server can also provide the same services as the active server to the clients). There is a problem.

A first object of the present invention is to provide a unique network address to each redundant server so that the network addresses of all the servers on the LAN do not change from the viewpoint of the client. It is to be able to set.

A second object of the present invention is to enable alternative processing of a failed server by an arbitrary server in the redundant server when a failure occurs in one of the redundant servers. is there.

Furthermore, a third object of the present invention is to send a request to a server via a LAN and receive a response to the software on the client which is provided with a certain service from the server and its software. The purpose of this is to enable a user to execute without making them aware that a failure has occurred in any of the redundant servers and the alternative processing has been performed by another server.

[0007]

In order to achieve the above first and second objects, a communication process for performing a communication process with a client for each server having a redundant configuration on a LAN by using one LAN interface of the server A virtual layer that makes it appear as if there are multiple virtual interfaces, means for registering and managing multiple virtual interfaces in the virtualization layer, mutual monitoring of the status between redundant configuration servers, and exchange of its own address with other servers. Mutual monitoring unit is provided.

Further, in order to achieve the above third object,
When a failure occurs in any of the servers in the server group, a means for setting the server that performs the alternative process so that the network address of the failed server can be used in the communication processing unit along with the own network address unique to the alternative server, LA
A means is provided for notifying the communication processing units of all clients on N that the alternative server has taken over the network address of the failed server.

Since each server sets a unique network address for one virtual interface of the virtual layer at the time of startup, the client can uniquely identify each server and send a request. This means that each server can process / provide some service to the client (all server active systems).

Further, since the virtualization layer controls transmission / reception of data according to the state of the virtualized interface, illegal data will not flow on the network. Further, the mutual monitoring unit has a function of simultaneously transmitting the same data to all devices connected to the LAN included in the communication processing unit (broadcast) or a function of simultaneously transmitting the same data to a plurality of specific devices (multicast). Since data is exchanged between each server using, it is not necessary to set all addresses of the redundant configuration server group for each server.

Since the network address takeover notification of the failed server is exchanged between the server and the communication processing section between the client, the software on the client and its user need not be aware.

[0012]

DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below. Note that the standby server referred to in the present invention is not a dedicated standby server that does not function as a server while the active server is operating, but normally performs the same processing as the active server, causing a failure in the active server. In this case, the processing is performed on behalf of the user. The network used in the present embodiment is currently widely used mainly for UNIX workstations Ethernet (X
Erox registered trademark), TCP / IP (Transm)
ision Control Protocol / In
A network using the Internet Protocol is described as an example. In the above network, a protocol processing unit that actually performs communication processing with a physical address assigned to each LAN controller is used as an address that uniquely represents a node (client, server, etc.) on the network for destination identification between nodes. End-to-end communication is performed with a combination of logical addresses. In this embodiment, a system having one active server and one standby server will be described as an example, but the present invention itself does not matter in the number of active / standby servers.

First, the system configuration of the network high reliability system and the internal configuration of the server constituting the system will be described with reference to FIG.

In the figure, 101 receives a request from a client 103 which sends a request to a server,
Server A (hereinafter, active server) that performs the processing, 10
Reference numeral 2 denotes a server B (hereinafter referred to as a standby server) that performs alternative processing when a failure occurs in the active server 101, 104
Is the client 103 and the active / standby system server 101
Is a LAN connecting / 102. In the standby server 102, 105 is a protocol processing unit that processes a TCP / IP protocol, and 106 is a LAN having a physical address from the protocol processing unit 105 so that a plurality of logical addresses can be set on one LAN board 108. A LAN driver virtualization layer that makes a virtual plurality of LAN drivers 107 that control the board 108 exist, and 109 exchanges monitoring information with the active server 101 via the LAN to detect a failure of the active server 101. At this time, the mutual monitoring unit instructs the LAN driver virtualization layer 106 to take over the logical address of the active server 101. Here, the LAN driver virtualization layer 10
6 is a protocol processing unit 105 and a LAN driver 107
By providing an interface that does not change from the existing one, the present invention can be used without changing the protocol processing unit 105 and the LAN driver 107 that are existing software assets. Further, the mutual monitoring unit 109 uses a network address (broadcast address) for transmitting the same data to all devices connected to the LAN of the protocol processing unit 105 at once, or the same data once to a plurality of specific devices. Mutual monitoring data is exchanged using the network address (multicast address) to be transmitted to. Further, by exchanging the physical / logical address of the own server with each other, it is necessary to set the physical / logical address of the server to be mutually monitored. In the processing when a failure occurs in any active server 101, the LAN driver virtualization layer 106 is also instructed to take over a logical address. The internal configuration of the active server 101 is the same as that of the standby server 102.

Next, the internal structure and operation of the LAN driver virtualization layer will be described with reference to FIGS. 2 and 3.

FIG. 2 shows the internal structure of the LAN driver virtualization layer.

In the figure, 201 (-1 to n) are virtual driver groups which make the protocol processing unit 105 appear to have a plurality of LAN interfaces and can set and process a logical address for each. Is a virtualization layer management unit that changes the state of the virtual driver group 201 and manages the virtual driver management table 203. Reference numeral 204 is a virtual driver that corresponds to the data received by the LAN driver 107 based on the virtual driver management table 203. It is a virtual driver control unit that is assigned to. Here, the virtualization layer management unit 202 generates the virtual drivers 201 for the number of servers (including itself) that configure the active / standby redundant server group at the time of server startup, and the active server from the mutual monitoring unit 109. In response to the 101 failure detection notification, the logical address of the failed active server 101 is set in any unused virtual driver 201, the status is registered in the virtual driver management table, and the physical address change notification is sent to the client. Here, in this physical address change notification, ARP (Address Re) used for address resolution processing (corresponding physical address to logical address) in the TCP / IP network.
solution Protocol: RFC8 for details
26) is used. Specifically, the logic of the local server /
An ARP message in which a combination of physical addresses is set is sent to all clients.

FIG. 3 is a diagram showing the contents of a virtual driver management table in which the virtualization layer management unit registers management information on the status of the LAN driver virtualization layer.

In the figure, 301 is a virtual driver number which is information for uniquely identifying each virtual driver 201, 302 is a setting address field for registering a logical address set in the virtual driver 201, and 303 is the virtual address. This is a usage status field for registering the usage status (usable / unusable) of the driver 201. Note that this table is set to an initial value (“0” in the setting address field 302 and “unusable” in the usage status field 303) by the virtualization layer management unit 202 at server startup, and then any virtual driver When the own address is set in (for example, # 1), the logical address is registered in the setting address field 302 and "available" is registered in the usage status field 303. Further, even when an address is taken over when a failure of the active server is detected, any unused virtual driver 201
The same process is performed for. Then, when the active server is restored, "unavailable" is registered in the usage status field 303 of the virtual driver 201 to which the logical address of the active server is assigned. Here, the virtual driver control unit 204 operates so as to pass only the transmission / reception data with the virtual driver whose usage status field 303 is “available”. Furthermore, by providing "use only transmission" or "use only reception" in the usage status field 303, it is possible to virtually have a network interface for only transmission and only for reception.

Next, in the network high-reliability system shown in FIG. 1 below, the process of taking over the standby server due to the failure of the active server and the recovery process to the system by restarting the active server are realized. A network reliability improving method to be described will be described with reference to FIGS. 4 and 5.

First, with reference to FIG. 4, description will be given of the takeover process in the standby system server due to the failure of the active system server.
Here, initially, the client 103 holds a combination of a physical address and a logical address of the server A (active system) 101 (410). The client 103 exchanges a request and a response thereto with the server A (active system) 101 according to the content of the address combination 410. Meanwhile, between the server A (active system) 101 and the server B (standby system) 102, the mutual monitoring units 109 mutually monitor the operation via the LAN 104 (401). Then, it is assumed that a failure occurs in the server A (active system) 101 for some reason (402). At this time, the client 103
Has no way of knowing the failure of server A (active system) 101,
Since there is no response from the server to the send request, it is resent. On the other hand, server B (standby system)
The mutual monitoring unit 109 of 102 monitors the server A by mutual monitoring.
When a failure of (active system) 101 is detected (403), LA
Virtualization layer management unit 2 of N driver virtualization layer 106
02 is notified of the occurrence of a failure in the server A (active system) 101 and its logical address. Upon receiving the notification, the virtualization layer management unit 202 sets the takeover logical address for any free virtual driver 201 in the LAN driver virtualization layer 106 to the protocol processing unit 105 (40
4) Register the takeover logical address in the setting address field 302 of the virtual driver 201 entry of the virtual driver management table 203 and the "available" state in the usage state field 303 (405). After that, the logical address of the combination of the logical address of server A and the physical address of server B is
The physical address change notification is notified to all nodes on the LAN 104 (406). Upon receiving the address change notification, the client 103 updates the logical / physical address combination held as notified (411). As a result, when the logical address takeover process from the server A (active system) 101 is completed and then the response waiting for the request resending request times out, the client sends to the server A (active system) 101 based on the address combination 411. Resend request of server B (standby system) 1
02, the request process is processed by the server B (standby system) 102, and a response is returned. Here, the mutual monitoring unit 109 considers that the failed server A (active system) 101 does not adversely affect the system in some way, and provides a dedicated communication line between the active and standby servers to provide the server A (active server). Current system) 1
01 may be instructed to completely stop operation. This operation stop command is executed by directly controlling the control signal line of the hardware so that the server A (active system) 101 is executed in any state. Also, server B
The mutual monitoring unit 109 of the (standby system) 102 detects the failure of the server A (active system) 101, and then the server A (active system) 101.
Server A (active system) until the failure recovery notification from
It operates by removing 101 from the mutual monitoring target.

As described above, according to the present embodiment, when the processing is handed over to the standby server due to a failure of the active server, the process on the client and the user receiving the service from the server are transferred to the standby server. There is no need to be aware of the succession of request processing, that is, the failure of the active server.

Next, the fault recovery processing of the faulty active server will be described with reference to FIG. Here, the area between the client and the server is in the state after the takeover processing described with reference to FIG. That is, the server A (active system) 101 is in a failure state, and the server B (standby system) 102 originally has the server A (active system) 10 in addition to its own logical address.
The client 103 has two virtual drivers 201 in the “available” state having a logical address of 1, and the client 103 is the server B.
Physical address of (standby system) 102 and server A (active system)
It holds the combination with the logical address of 101 (51
0). Therefore, the client 103 uses the address combination 5
According to the content of 10, a request to the server A (active system) 101 is transmitted to the server B (standby system) 102, and a response to the request is received. Where server A
If the operation recovery of the (active system) 101 is performed manually by the user or by some means, the mutual monitoring unit 109 of the server A (active system) 101 determines that the server B (standby system) 1
02 to the LAN via the LAN 104 (50)
1). When the mutual monitoring unit 109 of the server B (standby system) 102 which has received the failure recovery notification from the server A (active system) 101 detects the recovery of the server A (active system) 101 from the notification (502), the LAN driver virtualization layer The virtualization layer management unit 202 of 106 is notified of the failure recovery of the server A (active system) 101 and its logical address. Upon receiving the notification, the virtualization layer management unit 202 sets the setting address field 302 of the virtual driver 201 entry in which the logical address in the virtual driver management table 203 is registered to zero, and the usage status field 303 to "unavailable" status. Is registered (503), and processing for invalidating the setting of the logical address for the virtual driver 201 is performed on the protocol processing unit 105 (504). Here, after the process 503, since the virtual driver control unit 204 does not pass the data transmitted / received to / from the virtual driver except for the use status field 303 of the virtual driver management table 203 of “available”, the server B (standby system) 102 Does not respond to the request from the client 103. At this time, the client 103 has no way of knowing the failure recovery of the server A (active system) 101, and since there is no response from the server to the transmission request, the client 103 retransmits it. After that, the server B (standby system) 102 issues an LA / physical address change notification for the combination of the physical / logical address of the server A to the LA.
Notifying all the nodes on N104 (505), the client 103 having received the address change notification updates the logical / physical address combination held as notified (51).
1). As a result, when the logical address takeover process from the server B (standby system) 102 is completed, and then the waiting for the response of the request resending request times out, the client is based on the address combination 511. ) 102, the request retransmitted to the server A (active system) 101, which has been transmitted to the server A (active system) 101, is transmitted to the server A (active system) 101, and the request processing is processed by the server A (active system) 101. The response will be sent back. The mutual monitoring unit 109 of the server A (active system) 101 and the server B (standby system) 102 restarts the mutual monitoring operation thereafter (5
06).

As described above, according to the present embodiment, when the failed active server is restored, when the service processing for the client is taken over from the standby server to the active server, the service on the client is received from the server. The process and the user are not aware that the request processing has been taken over by the active server, that is, the failure recovery of the active server.

An application example of the system to which the present invention is applied will be described below with reference to FIGS. 6, 7 and 8.

First, with reference to FIG. 6, an example of a system in which the network high reliability method according to the present invention is applied to a remote file system for sharing a file on a disk of a server via a LAN will be described.

In the figure, reference numeral 601 denotes a shared HDD which is a disk for storing an access file from the client 103 when the active server 101 is operating normally.
In (A) and 602, the standby server 102 is the client 1
The shared HDD (B) is a disk for storing the access file from the server 03. Where shared HDD
The (A) 601 and the shared HDD (B) 602 are connected to both the active server 101 and the standby server 102. Then, when a failure occurs in the active server 101, the standby server 102 takes over the address of the active server 101 and then the client 10 for the address.
By processing the request from the shared HDD (A) 601, the file on the shared HDD (A) 601 can be accessed even if the active server 101 has a failure.

Next, referring to FIG. 7, a network high reliability system according to the present invention is applied to a host access system centering on a large-scale general-purpose computer (host) for performing routine / online (batch) processing of routine work in a company. An example of the applied system will be described.

In the figure, reference numeral 701 is a large-scale general-purpose computer for performing routine / non-online (batch) processing of routine work in a company. Here, the large general-purpose computer 70
1 is connected to the active server 101 and the standby server 102, and when a failure occurs in the active server 101,
Although the access to the large-scale general-purpose computer 701 is once disconnected, the client 103 requests the active-system server 101 to access the large-scale general-purpose computer 701 again, and the standby-system server 102 performs the substitute process for the request. By doing so, even if there is a failure in the active server 101, the large general-purpose computer 701 can be accessed again.

Further, referring to FIG. 8, a network high reliability system according to the present invention is applied to an Internet connection system for connecting to the Internet, which is a huge network covering the entire world, and for providing various services on the Internet to client users. An example of a system to which is applied will be described.

In the figure, reference numeral 801 denotes a shared HDD (A) which is a disk for storing Internet information such as mail and news accessed from the client 103 when the active server 101 is operating normally, and 802 is a standby. A shared HDD (B) 803, which is a disk for storing Internet information such as mails and news accessed by the system server 102 from the client 103, is a huge network Internet for exchanging the Internet information with the entire world. . Where shared HDD
The (A) 801 and the shared HDD (B) 802 are connected to both the active server 101 and the standby server 102. Then, when a failure occurs in the active server 101, the standby server 102 takes over the address of the active server 101 and then the client 10 for the address.
Access request from 3 is shared HDD (A) 801
By performing the processing for the above, even if there is a failure in the active server 101, it becomes possible to access the Internet information on the shared HDD (A) 801. Further, the active server 101 and the standby server 102 are respectively connected to the Internet 803 to exchange the Internet information. However, when the active server 101 fails, the standby server 102 also performs the processing instead. As a result, there is no omission of the Internet information.

As described above, the present invention can be applied to various client-server type systems, and the reliability of the system can be improved.

[0033]

By using the present invention in a client server system consisting of an active server and a standby server, a client can be connected to a LAN and operate normally without being aware of the active / standby server. Since various servers can be requested to provide various services, there is an effect that the processing capacity equivalent to the system investment amount can be sufficiently obtained.

Further, the present invention is realized only by the software of the server without using special hardware, and only by changing the software of the server of the existing system, in other words, without changing the client. Since it can be realized, there is an effect that the high reliability of the existing system can be achieved with a small investment.

Further, the present invention is applicable only to a server having a redundant configuration, and can be realized by changing the server of the existing system, in other words, without changing the client, so that the existing assets can be saved. There is an effect that it can be effectively utilized.

[Brief description of drawings]

FIG. 1 is a block diagram showing an example of the overall configuration of a network high reliability system of the present invention and an internal structure of a server.

FIG. 2 is an explanatory diagram showing a configuration of a LAN driver virtualization layer of the present invention.

FIG. 3 is an explanatory diagram showing a configuration of a virtual driver management table for registering management information of a LAN driver virtualization layer of the present invention.

FIG. 4 is an explanatory diagram showing a process of taking over to a standby server due to a failure of the active server.

FIG. 5 is an explanatory diagram showing a process of taking over from a standby server due to failure recovery of the active server.

FIG. 6 is a block diagram showing an overall configuration of a remote file system which is an example of an application system of the network high reliability system of the present invention.

FIG. 7 is a block diagram showing an example of the overall configuration of a host access system which is an example of an application system of the network high reliability system of the present invention.

FIG. 8 is a block diagram showing an example of the overall configuration of an Internet connection system which is an example of an application system of the network high reliability system of the present invention.

[Explanation of symbols]

 Reference numeral 101 ... Active server, 102 ... Standby server, 103 ... Client, 104 ... LAN, 105 ... Protocol processing unit, 106 ... LAN driver virtualization layer, 107 ... LAN driver, 108 ... LAN board, 109 ... Mutual monitoring unit.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Yoshifumi Shin, 810 Shimoimaizumi, Ebina, Kanagawa Prefecture Office Systems Division, Hitachi, Ltd.

Claims (7)

[Claims] 1. A LAN is connected with a client and a server for processing a request from the client, and normally, an active server for processing the request from the client, and a substitute for the active server when a failure occurs in the active server. In a system in which the standby server that performs processing is configured to perform server redundancy, and when the active server fails, the standby server performs a substitute process for the active server, and a redundant configuration is provided on the LAN. In each server, a mutual monitoring unit that mutually monitors the status between redundant servers and exchanges its own address with another server, and a communication processing with a client through one LAN interface that is a connection point between the server and the LAN. In the redundant configuration server group, a virtualization layer is provided to make the communication processing unit that performs When a failure occurs in an arbitrary server, the standby server that has detected the server failure in the mutual monitoring unit sets the network address of the failed server so that it can be used in the communication processing unit, and A method and system for improving network reliability, which notify a communication processing unit that the network address of a failed server has been taken over by the alternative server. 2. The virtual layer according to claim 1, wherein the virtual layer is a virtual interface that makes a virtual appearance of a plurality of LAN interfaces, and a virtual layer management unit that sets a network address to the virtual interface and manages a state. , And a virtual interface control unit that controls data exchange between the virtual interface and an actual LAN interface, and a communication processing unit of the server and L
It is placed between LAN drivers that control the AN interface, and by assigning one of the virtual interfaces to the network address of the own server, the request from the client to the own server can be processed by the communication processing unit. When a failure occurs in an arbitrary server, a request from a client to the failed server can be processed by the communication processing unit of the own server by assigning the network address of the failed server to one of the remaining arbitrary virtual interfaces. Highly reliable network system and system. 3. The means according to claim 1, wherein the mutual monitoring unit transmits the same data to all the devices connected to the LAN, which are included in the communication processing between the client and the server on the LAN, or By exchanging mutual monitoring data using a method of transmitting the same data to a plurality of specific devices at once, and by exchanging the address of the own server with each other, it has the address of the mutual monitoring target server A network high reliability method and system for instructing a virtualization layer of an address of a failed server in processing when a system server failure occurs. 4. A network using TCP / IP is used for a communication processing procedure between a client and a server via a LAN performed by the communication processing unit, and the address change notification is the physical address. A network high reliability method and system using ARP used for address resolution processing (associating a physical address with a logical address in the network) in a TCP / IP network for change notification. 5. A network high reliability system according to claim 1, wherein the server having a plurality of redundant LANs on the LAN is a remote file system server which shares files on the server's disk via the LAN. . 6. An access relay process in a host access system in which a plurality of servers provided on a LAN and having a redundant configuration can access from a client on the LAN to a large-scale general-purpose computer that performs routine / online processing of routine business in a company. The network high reliability system according to claim 1, which is a server for performing the above. 7. A server having a plurality of LANs and having a redundant configuration provides a client user with a connection from a client on the LAN to the Internet, which is a huge network covering the entire world, and various services on the Internet. The network high reliability system according to claim 1, which is a server of an Internet connection system.