JP2004227534A - Server system, and error shielding method for mediating apparatus and client type server system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a highly-reliable server system and mediating apparatus, which is suitable for continuous operation and can be inexpensively structured, and an error shielding method. <P>SOLUTION: In a server-side system in a client server type system, a request from a client is transferred to a server and a plurality of mediating apparatuses provided with a transfer means for transferring an appropriate response among responses to the request received from the server to the requester client. In addition, the server-side system can be accessed as a single server from the client, and an virtual address is predetermined so as to be accessed as a single router from the server. The mediating apparatuses have a control means for controlling so as to allow the transfer means to transfer requests and their responses to the virtual addresses. <P>COPYRIGHT: (C)2004,JPO&NCIPI

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a server system such as an Internet business system such as an EC server, a Web server, and an authentication server in which a service cannot be stopped due to a failure or a node system of a communication network.
[0002]
[Prior art]
In this type of server system, it is necessary to always appropriately provide services to clients. That is, it is necessary to conceal “errors” of the server system and prevent the occurrence of “failure”. Here, “failure” means loss of function of the system. Specific examples of the cause of the “failure” include a bug or a failure in hardware, a bug in software, and the like. The cause of the "failure" is called "fault". Even if a "fault" exists in the system, it does not necessarily immediately cause a "failure" but may be latent. The appearance of an abnormality due to a “fault” is referred to as an “error”. Then, if the system in which the “error” occurs deviates from a normal state, a “failure” occurs. Note that the "error" is not caused only by a "fault" in a narrow sense such as a bug or a failure. For example, an intermittent failure, an operator setting error, or an operation error also causes the "error".
[0003]
As an example of a highly reliable server system capable of concealing this “error”, “Error Concealment Method in Server System, Mediation Device, Client-Server System” (Japanese Patent Application No. 2001-256511) has been proposed. An example of the high reliability server system will be described with reference to FIG. Here, a system that provides a Web service to a client will be exemplified.
[0004]
As shown in FIG. 19, the high reliability server system 1010 receives a request from the client 1041 and intervenes between the high reliability server 1020 and the client 1041 that respond to the request. And an intermediary device 1030.
[0005]
The high reliability server 1020 includes one or more servers. In the example of FIG. 19, the high reliability server 1020 includes three servers 1021, 1022, and 1023. Each server returns the same response to the same request transmitted from the mediation device to each server if the server is operating normally.
[0006]
The mediation device 1030 receives the request from the client 1041 via the network 1050, transfers the request to the high-reliability server 1020 at an appropriate timing described later, checks the validity of the response from the high-reliability server, and sends a correct response. It includes a control unit 1031 for returning to the client 1041, and a server management table 1032 for managing the status of each of the servers 1021, 1022, and 1023 of the high reliability server 1020. The control unit 1031 registers which server is normal or has a failure in the server management table 1032, and refers to the server management table 1032 to cause which server to process a request from the client 1041. Judge. The server management table 1032 includes a server ID for identifying a server, an IP address of the server, and server operation information.
[0007]
The client 1041 requests a Web page from the reliable server 1020 via the network 1050 using a Web browser.
[0008]
Next, the operation of the mediation device 1030 will be described. The client 1041 transmits a Web page request message 1110 to the high reliability server system 1010. The control unit 1031 receives the request message 1110, looks at the server management table 1032, and transfers it to a normal server. In the example of FIG. 19, since the servers 1021, 1022, and 1023 are normal, the control unit 1031 makes copies 1111, 1112, and 1113 of the request message 1110, and transfers them to the servers 1021, 1022, and 1023.
[0009]
The server 1021 receives the request message 1111 and returns a response 1121 to the control unit 1031. Similarly, the server 1022 returns a response 1122 and the server 1023 returns a response 1123 to the control unit 1031. The control unit 1031 makes a majority decision with the responses 1121, 1122, and 1123, determines that the majority response is normal, and returns a response 1120 to the client 1041.
[0010]
If, as a result of the majority decision, there is a response that does not match all but does not match, the control unit 1031 determines that the server that returned the response is a failure, and changes the operating status information of the server management table 1032 from “normal” to “normal”. Change to "failure". Thereafter, the control unit 1031 does not transfer the request from the client to the failed server.
[0011]
In addition to the system described above, a dual system in which two computers perform the same processing in parallel (for example, see Patent Document 1), and a control computer of two systems divided into an online side and a standby side, A control system that synchronizes tasks (for example, see Patent Document 2) is known as a conventional technique.
[0012]
[Patent Document 1]
JP-A-52-60540
[Patent Document 2]
JP-A-1-303562
[0013]
[Problems to be solved by the invention]
As described above, in the conventional high-reliability server system, a control unit that receives a request from a client transfers the request to a plurality of servers, receives a response from the server to the request, and executes a majority vote. Has been adopted. Therefore, even if a server fails and returns an erroneous response to the control unit, the error could be concealed by majority decision.
[0014]
However, there has been a problem that if the control unit itself becomes an obstacle, the entire system goes down. In order to avoid this problem, it is conceivable to simply arrange a plurality of control units, but in this case, a plurality of control units are visible from a client or a server. For this reason, every time the control unit fails, it becomes necessary to manually change the settings on the client and server sides, or if both the settings are to be avoided, the control unit is remodeled into both the client and server for automation There was a problem that necessity arises.
[0015]
The present invention has been made in view of the above circumstances, and it is an object of the present invention to provide a server system, an intermediary device, and an error concealment method which can be constructed with high reliability, suitable for continuous operation, and inexpensively. is there.
[0016]
[Means for Solving the Problems]
In order to achieve the above object, the present invention provides a server-side system in a client-server type system, in which a request from a client is transferred to a server, and a legitimate response among the requests received from the server is transmitted to a request source. A plurality of intermediary devices having transfer means for transferring to the client are provided, and a virtual address to be shown as a single server from the client and as a single router from the server is determined in advance, and the intermediary device is Control means for controlling the transfer means to transfer the request and response addressed to the virtual address is provided.
[0017]
According to the present invention, even if one intermediary device transferring a request and a response between a client and a server becomes a failure, the transfer can be continued by another normal intermediary device. A more reliable client-server system can be constructed. In addition, virtual addresses are determined in advance so that any intermediary device can process requests and responses addressed to these virtual addresses. Thus, even if any intermediary device goes down due to a failure or the like, the server or client does not need to make any changes. That is, since existing servers and clients can be used without modification, a highly reliable client-server type system can be constructed at low cost.
[0018]
BEST MODE FOR CARRYING OUT THE INVENTION
A highly reliable server system according to a first embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram illustrating the overall configuration of the high reliability server system according to the present embodiment. In the present embodiment, a system that provides an authentication service, that is, a system in which the server is a RADIUS server will be described as an example. In the case of a personal computer or an intermediary device on which a server is running, a system in which Linux (registered trademark) is running as an OS is exemplified.
[0019]
As shown in FIG. 1, the high-reliability server system 10 includes a high-reliability server 20 that returns a response to a request from a client 70, and an intermediary device interposed between the high-reliability server 20 and the client 70. 30. In the example of FIG. 1, two mediation devices 30a and 30b are provided as mediation devices 30. In the following description, when each of the plurality of intermediary devices 30 and its components is individually represented, the mediation device 30a and the intermediary device 30b are denoted by reference numerals “a” and “b”. And
[0020]
In this system, the communication protocol system between the devices does not matter. In this embodiment, the TCP / IP protocol suite is used. The TCP / IP protocol suite does not express TCP and IP in a narrow sense, but means a broadly defined protocol system including UDP, ARP, and the like.
[0021]
The high reliability server 20 includes one or more servers. In the example of FIG. 1, the high reliability server 20 includes three servers 21, 22, and 23. Each server responds to the same request sent from the mediation device to each server if the server is operating normally. The servers 21, 22, and 23 may be implemented by software running on a computer or the like, or may be implemented by hardware. That is, the high reliability server 20 may be realized by a plurality of hardware, or may be realized by mounting a plurality of servers on one piece of hardware. In the example of FIG. 1, the servers 21, 22, and 23 are software running on computers 91, 92, and 93, respectively. Computers 91, 92 and 93 are assigned IP addresses 172.16.3.3, 172.16.3.4 and 172.16.3.5, respectively. The computers 91, 92, and 93 have MAC addresses cc: cc: cc: cc: cc: 03, cc: cc: cc: cc: cc: 04, and cc: cc: cc: cc: cc: 05, respectively. ing.
[0022]
The client 70 is a device that receives a service provided by the high reliability server system 10. In the present embodiment, the client 70 is an NAS (Network Access Server) such as a RAS (Remote Access Server). In the present embodiment, 192.168.3.3.200 is assigned to the client 70 as the IP address. In addition, it has a MAC address, dd: dd: dd: dd: dd: 00: 00.
[0023]
The mediation device 30 includes an NIC (Network Interface Card) 31 connected to a server 21, 22, 23 side network (hereinafter referred to as a Uchi-gun network) 81, and an NIC 32 connected to a client side network (hereinafter, referred to as an external network) 82. It has. The mediation device 30 includes an internal network-side packet receiving unit 33 that receives a packet from the internal network 81 via the NIC 31, an internal network-side packet transmitting unit 34 that transmits a packet to the internal network 81 via the NIC 31, and an NIC 32. An external network-side packet receiving unit 35 that receives a packet from the external network 82 via the external network 82, and an external network-side packet transmitting unit 36 that transmits a packet to the external network 82 via the NIC 32. Further, the mediation device 30 performs a process described below on the packets received from the internal network-side packet receiving unit 33 and the external network-side packet receiving unit 35, and performs an internal network-side packet transmitting unit 34 and an external network-side packet transmitting unit 36. And a packet control unit 40 for transmitting a packet by using the packet control unit.
[0024]
Each intermediary device 30 is connected via the NIC 31 to an internal network 81 connected to computers 91, 92, 93 on which the servers 21, 22, 23 are operating. In the example of FIG. 1, the computers 91, 92, 93 and the NIC 32 are connected to the repeater hub 83. Each intermediary device 30 is connected via the NIC 32 to an external network 82 to which the client 70 is connected. In the example of FIG. 1, the client 70 and the NIC 32 are connected to a repeater hub 84.
[0025]
A different IP address is assigned to each of the intermediary devices 30 to the NIC 32 connected to the external network 82 and the NIC 31 connected to the internal network 81, and each has a different MAC address (hardware address). In the example of FIG. 1, 192.168.3.1 is assigned as the IP address to the NIC 32a of the mediation device 30a, and the MAC address is aa: aa: aa: aa: aa: 01. The NIC 32b of the mediation device 30b is assigned 192.168.3.2 as the IP address, and the MAC address is aa: aa: aa: aa: aa: 02. In the example of FIG. 1, 172.16.3.1 is assigned as an IP address to the NIC 31a of the mediation device 30a, and the MAC address is bb: bb: bb: bb: bb: 01. 172.168.3.2 is assigned as an IP address to the NIC 31b of the mediation device 30b, and the MAC address is bb: bb: bb: bb: bb: 02.
[0026]
Next, a virtual server and a virtual router will be described. In order to realize the object of the present invention, that is, providing the client and the server without modification and providing higher reliability, it is necessary to satisfy the following two conditions.
[0027]
(1) Even if a failure occurs inside the high reliability server system 10, the failure is hidden from the client 70. That is, the client 70 does not need to change the operation or the setting due to the failure, and can receive the service from the high reliability server system 10 as before the failure.
[0028]
(2) Even if a failure occurs in any of the plurality of mediation devices 30, the failure is hidden in each of the servers 21, 22, and 23. In other words, each of the servers 21, 22, and 23 does not need to change the operation or setting due to the failure, and can communicate with the client 70 as before the failure.
[0029]
In the present invention, a “virtual server” is defined to realize the above condition (1), and a “virtual router” is defined to realize the above condition (2).
[0030]
A virtual server is a server that conceals the internal configuration of a high-reliability server system and virtually appears as one server when viewed from a client. For example, in the case of the highly reliable server system 10 illustrated in FIG. 1, as illustrated in FIG. 2, the virtual IP address 192.168.3.3.100 and the virtual MAC address aa: aa: aa: aa: The virtual server 11 having aa: 00.
[0031]
Here, the virtual IP address is an IP address assigned to the virtual server 11, and the client 70 transmits a request packet to this IP address. The virtual MAC address is a MAC address assigned to the virtual server 11, and is a network connected to the virtual server 11, that is, a communication connected to the external network 82 connected to the high-reliability server system 10. This is a MAC address used when the device communicates with the virtual server 11. That is, the virtual MAC address is a MAC (included in the ARP response packet) obtained when a communication device connected to the external network 82 makes an address resolution request (transmits an ARP request packet) for the virtual IP address. An address. The virtual IP address and the virtual MAC address need to be selected so as not to overlap with the IP address and the MAC address of the communication device connected to the external network 82. It should be noted here that no communication device having a virtual IP address and a virtual MAC address actually exists. Therefore, the ARP response to the ARP request is normally executed by the Linux kernel of the communication device having the virtual IP address and the virtual MAC address. However, since there is no communication device having the virtual IP address and the virtual MAC address, The problem of not being solved arises. In order to solve this problem, the intermediary device 30 has a pseudo ARP response unit and implements address resolution (details will be described later).
[0032]
With such a configuration, if one or more mediation devices are normal and one or more servers are normal, the virtual server 11 functions normally. Therefore, with the realization of the virtual server 11, the client 70 only needs to communicate with the virtual server 11 having the IP address 192.168.3.100 regardless of whether or not there is a failure inside the high reliability server system 10. good. That is, the above condition (1) can be satisfied.
[0033]
The virtual router is a router that hides the redundant configuration of the intermediary device and appears as one virtual router when viewed from the server. For example, in the case of the highly reliable server system 10 of FIG. 1, as shown in FIG. 3, the virtual IP address 172.16.3.100 and the virtual MAC address are located on the side of the internal network 81 when viewed from the servers 21, 22, and 23. bb: bb: bb: bb: bb: means the virtual router 37 having 0:00.
[0034]
Here, the virtual IP address is an IP address assigned to the virtual router 37. The communication device connected to the internal network 81 recognizes the packet as the IP address of the router when transferring the packet to the external network 82. On the other hand, the virtual MAC address is a MAC address assigned to the virtual router 37, and is a MAC address used when a communication device connected to the internal network 81 communicates with the virtual router 37. That is, the virtual MAC address is obtained when a communication device connected to the internal network 81 makes an address resolution request for the virtual IP address (that is, transmits an ARP request packet) (included in the ARP response packet). MAC address). The virtual IP address and the virtual MAC address need to be selected so as not to overlap with the IP address and the MAC address of the communication device connected to the internal network 81. It should be noted here that no communication device having a virtual IP address and a virtual MAC address actually exists. Therefore, the ARP response to the ARP request is normally executed by the Linux kernel of the communication device having the virtual IP address and the virtual MAC address. However, since there is no communication device having the virtual IP address and the virtual MAC address, The problem of not being solved arises. In order to solve this problem, the intermediary device 30 has a pseudo ARP response unit and implements address resolution (details will be described later).
[0035]
With such a configuration, if one or more intermediary devices 30 are normal, the virtual router 37 functions normally. Accordingly, with the realization of the virtual router 37, the communication devices connected to the internal network 81 (the servers 21, 22, and 23 in the example of FIG. 1) have the IP address whether or not there is a failure in each of the intermediary devices 30. Only communication with the virtual router 37 of 172.16.3.100 needs to be performed. That is, the above condition (2) can be satisfied.
[0036]
Next, details for realizing a virtual server / virtual router will be described. Although the virtual server is visible from the client and the virtual router is visible from the server, both the virtual server and the virtual router can be realized by the method described below.
[0037]
First, a detailed configuration of the packet control unit 40 will be described with reference to FIG. As shown in FIG. 4, the packet control unit 40 includes a server management table 42 that manages the state of all servers (the servers 21, 22, and 23 in the example of FIG. 1) of the high reliability server 20, and an internal network. , An internal network-side pseudo-ARP response unit 43 that returns a response to an IP address resolution request of a virtual router, and an external network-side pseudo-ARP response unit 44 that returns a response to an IP address resolution request of a virtual server in an external network And a server control unit 41 which receives a request from the client 70, transfers the request to the high reliability server 20, checks the validity of the response from the high reliability server 20, and returns a correct response to the client 70. ing.
[0038]
The server control unit 41 registers which server is normal or has a failure in the server management table 42 and processes the request from the client 70 to which server with reference to the server management table 42. Judge whether to let.
[0039]
FIG. 5 shows an example of the server management table 42. As shown in FIG. 5, the server management table 42 includes a server ID for identifying each server, an IP address of a computer on which the server is operating, and an operating state of the server. The example in FIG. 5 shows a case where the status of each server is normal.
[0040]
When detecting a server failure from a response packet from each server, the server control unit 41 changes the operating state of the corresponding server in the server management table 42 from “normal” to “failure”. As an example, FIG. 6 shows a server management table 42 in a case where the server 23 has entered the failure state.
[0041]
On the other hand, when detecting that the server in the failed state has become normal, the server control unit 41 changes the operating state of the server management table 42 from “failed” to “normal”. For example, when the server 23 recovers from the failure, the server control unit 41 changes the server management table 42 from that shown in FIG. 6 to that shown in FIG. As a method of recognizing that the server has become normal, for example, a method in which the server control unit 41 throws a packet to the server and determines from the response, or a method in which the server control unit 41 starts the computer when the computer starts up And a method of operating software for notifying the user on a computer.
[0042]
Next, packet transmission / reception processing on the internal network side will be described. Communication performed on the internal network is communication between the servers 21, 22, and 23 and the virtual router 37. In the example of FIG. 1, the computers 91, 92, and 93 have set the default gateway to the virtual router 37.
[0043]
First, since the computers 91, 92, and 93 need to know the MAC address for the virtual IP address 172.16.3.100 before communication, the computers 91, 92, and 93 broadcast an ARP request packet. As described above, since there is no communication device having the virtual IP address 172.16.3.100, the address is not normally resolved, and the computers 91, 92, and 93 cannot communicate. appear. In order to avoid this problem, the internal network-side packet receiving unit 33 sets the NIC 31 to the promiscuous mode (a mode for receiving all packets arriving at the NIC), thereby setting all packets arriving at the repeater hub 83 to the raw socket. , And passes the ARP request packet of the virtual IP address 172.16.3.100 among the received packets to the internal network-side pseudo-ARP response unit 43. The pseudo ARP response unit 43 returns a virtual MAC address selected in advance so as not to overlap with the MAC address of a device connected to the same network. All the pseudo ARP responders 43 must be set to return the same virtual MAC address. In this example, the pseudo-ARP response units 43a and 43b return an ARP response packet including bb: bb: bb: bb: bb: 00 which does not overlap with the MAC addresses of the intermediary device 30, the computers 91, 92, and 93. When a plurality of mediation apparatuses are operating normally, a plurality of ARP response packets are returned to the computers 91, 92, and 93, respectively, but there is no problem because the packets have the same contents. In this way, address resolution for the virtual IP address is realized, and the computers 91, 92, and 93 can transmit packets to the virtual router 37.
[0044]
In addition, in order to set the NIC in the promise mode and perform a filter (receive only necessary packets by raw socket), the data link layer to the application layer is used without using the TCP / IP stack provided by the kernel. It is convenient to use a library libpcap (ftp://ftp.ee.lbl.gov/libpcap.tar.Z) for capturing packets.
[0045]
Next, the response packet receiving process of the mediation device 30 in the internal network 81 will be described. FIG. 7 illustrates a packet format that arrives at the virtual router 37 via the internal network 81. In the example of FIG. 7, the format of a packet transmitted from the computer 91 to the virtual router 37 is shown.
[0046]
As shown in FIG. 7, the transmission destination MAC address of the packet is neither the one of the mediation device 30a nor the one of the mediation device 30b. Further, the destination IP address is neither the one of the mediation device 30a nor the one of the mediation device 30b. Therefore, as described above, the internal network-side packet receiving unit 33 receives the packet using libpcap. When the destination IP address is the client 70, the IP address is passed to the server control unit 41. This operation is performed simultaneously and simultaneously by the internal network side packet receiving unit 33a and the internal network side packet receiving unit 33b.
[0047]
Next, a request packet transmission process of the mediation device 30 in the internal network 81 will be described. FIG. 8 illustrates a format of a packet transmitted from the virtual router 37 to the server via the internal network 81. In the example of FIG. 8, the format of the packet transmitted from the mediation device 30 (both the mediation device 30a and the mediation device 30b) to the computer 91 is shown.
[0048]
As shown in FIG. 8, the source IP address of the packet is the client 70, not the mediation device 30a or the mediation device 30b. This is because the mediation device 30 appears to the server 21 as a simple router. Also, the source MAC address is neither the mediation device 30a nor the mediation device 30b. Therefore, the packet cannot be transmitted using a normal socket using the standard TCP / IP stack provided by the kernel. Therefore, the internal network-side packet transmission unit 34 creates an Ethernet header, an IP header, and a UDP header, and transmits them to the server 21 using the raw socket.
[0049]
When a plurality of mediation devices are operating normally, the server receives the packets of the number of mediation devices that are operating normally as shown in FIG. Since all packets are exactly the same, only the first one needs to be processed and the remaining packets must be discarded at the server. For example, in the case of an application using UDP such as RADIUS, there is a possibility that a plurality of identical packets are received repeatedly, and therefore, the function of discarding the second and subsequent duplicated packets usually has originally been provided. Therefore, the function of discarding the unnecessary packets has already been implemented, and there is no need to modify the server for the present invention.
[0050]
Next, packet transmission / reception processing on the external network side will be described. Communication performed on the external network 82 is communication between the client 70 and the virtual server 11.
[0051]
First, before communication, the client 70 broadcasts an ARP request packet because the client 70 needs to know the MAC address for the virtual IP address 192.168.3.3.100. As described above, since there is no communication device having the virtual IP address 192.168.3.100, the address is not normally resolved, and the client 70 cannot perform communication. In order to avoid this problem, the external network-side packet receiving unit 35 sets the NIC 32 to the promiscuous mode (mode for receiving all packets arriving at the NIC) to receive all packets arriving at the repeater hub 84 in a raw socket. The ARP request packet of the virtual IP address 192.168.3.100 among the received packets is passed to the external network-side pseudo ARP response unit 44. The pseudo-ARP response unit 44 returns a virtual MAC address selected in advance so as not to overlap with the MAC address of a device connected to the same network. All pseudo ARP responders 44 must be set to return the same virtual MAC address. In this example, the pseudo ARP response units 44a and 44b return an ARP response packet including aa: aa: aa: aa: aa: 00 that does not overlap with the mediation device 30 and the MAC address of the client 70. When a plurality of mediation devices are operating normally, a plurality of ARP response packets are returned to the client 70, but there is no problem because the packets have the same contents. In this way, address resolution for the virtual IP address is realized, and the client 70 can transmit packets to the virtual server 11.
[0052]
In addition, in order to set the NIC in the promise mode and perform a filter (receive only necessary packets by raw socket), the data link layer to the application layer is used without using the TCP / IP stack provided by the kernel. It is convenient to use a library libpcap (ftp://ftp.ee.lbl.gov/libpcap.tar.Z) for capturing packets.
[0053]
Next, a request packet receiving process of the mediation device 30 in the external network 82 will be described. FIG. 9 illustrates a packet format coming to the virtual server 11 via the external network 82. In the example of FIG. 9, the format of the packet transmitted from the client 70 to the virtual server 11 is shown.
[0054]
As shown in FIG. 9, the destination MAC address of the packet is neither the one of the mediation device 30a nor the one of the mediation device 30b. Therefore, as described above, the external network-side packet receiving unit 35 receives a packet using libpcap. When the destination IP address is the virtual server 11, the address is passed to the server control unit 41. This operation is performed simultaneously and simultaneously by the external network-side packet receiving unit 35a and the external network-side packet receiving unit 35b.
[0055]
Next, a response packet transmission process of the mediation device 30 in the external network 82 will be described. FIG. 10 illustrates a format of a packet transmitted from the virtual server 30 to the client 70 via the external network 82. In the example of FIG. 10, the format of a packet transmitted from the mediation device 30 (both the mediation device 30a and the extension device 30b) to the client 70 is shown.
[0056]
As shown in FIG. 10, the source MAC address is neither the mediation device 30a nor the mediation device 30b. Therefore, the packet cannot be transmitted using a normal socket using the standard TCP / IP stack provided by the kernel. Therefore, the external network-side packet transmitting unit 36 creates an Ethernet header, an IP header, and a UDP header, and transmits them to the client 70 using the raw socket.
[0057]
When a plurality of mediation devices are operating normally, the client receives the packets of the number of mediation devices that are operating normally as shown in FIG. Since all packets are exactly the same, only the first one needs to be processed and the remaining packets need to be discarded at the client. For example, in the case of an application using UDP such as RADIUS, there is a possibility that a plurality of identical packets may be received repeatedly, and therefore, the function of discarding the second and subsequent duplicated packets usually has originally been provided. Therefore, the function of discarding the unnecessary packets has already been implemented, and there is no need to modify the client for the present invention.
[0058]
Next, the operation of the server control unit 41 will be described with reference to the drawings. It is assumed that the configuration and each device of the high reliability server system taken as an example are in a state as shown in FIG. FIG. 11 is a flowchart illustrating the operation of the server control unit 41. FIGS. 12 and 13 are diagrams illustrating the operation of the high reliability server system 10.
[0059]
The server control unit 41 checks whether the external network-side packet receiving unit 35 has received a request packet from the client 70 (step SA1). If there is no request packet, the server control unit 41 again checks the reception of the request packet. If there is a request packet, the request packet is transmitted to the server whose operation status is "normal" in the server management table 42 by using the internal network side packet transmission unit 34 (step SA2). The flow of the request from the client by the above processing is indicated by the thick arrow in FIG.
[0060]
Next, a timer is set to check a response packet from each server (step SA3). If it is not time-out, it is checked whether or not the internal network side packet receiving unit 33 has received the response packet from the server (step SA4). If no response packet has been received from the server, a timeout check is performed again (step SA3). If a response packet has been received from the server, it is checked whether response packets have been received from all of the destination servers of the request packet (step SA5). If not all response packets are received, a timeout check is performed again (step SA3).
[0061]
When all the response packets have been collected, it is checked whether or not there is a failed server (step SA6). Methods for determining whether there is a failed server include, for example, a method of performing a majority decision with a plurality of responses, a method of checking the validity of data included in the responses, and the like. As a method of checking the data validity of the latter, for example, if an improbable code is included, it is determined to be a failure, and if a code that should be present is not included, it is determined to be a failure. Is mentioned. If there is a failed server, the operation state of the server is registered as "failure" in the server management table 42 (step SA7). Thereafter, a normal response packet is transmitted to the client 70 by using the external network-side packet transmission unit 36 (step SA8).
[0062]
If there is no failed server in step SA6, a normal response packet is transmitted to the client 70 using the external network-side packet transmission unit 36 (step SA8).
[0063]
If a timeout occurs in step SA3, that is, if there is a server that does not return a response packet, the server is determined to be faulty, and the server management table 42 indicates the operating status of the server as “failure”. (Step SA9), and checks whether there is another failed server (step SA6). Then, a normal response packet is transmitted to the client 70 using the external network-side packet transmission unit 36 (step SA8).
[0064]
After transmitting the response packet to the client 70 (step SA8), the request from the client 70 is checked again (step SA1). The flow of the response packet from the server to the client 70 by the above processing is indicated by the thick line arrow in FIG. In addition, the server management table 42 in the case where the server 23 has failed becomes as shown in FIG. 6 as described above.
[0065]
By the above processing, the request packet from the client 70 is not transferred to the server which has become the “failure” among the servers 21, 22, 23 constituting the high reliability server 20, and the high reliability server system That is, they are separated from 10.
[0066]
The mediation devices 30a and 30b operate independently without being aware of each other. Therefore, even if one of the mediation devices fails, the other mediation devices are not affected. For example, if the mediation device 30a fails and goes down, the mediation device 30b continues to operate the above-described virtual server and virtual router without knowing whether the mediation device 30a is alive or dead.
[0067]
Similarly, even if the mediation device recovers from the failure, it does not affect the other mediation device. For example, even if the intermediary device 30a, which has been down due to a failure, recovers and starts operation of the above-described virtual server and virtual router, the intermediary device 30b does not know whether the intermediary device 30a is alive or not. The operation of is continued.
[0068]
In order for the mediation devices 30a and 30b to execute the same processing independently in parallel as described above, a mechanism that satisfies the condition that completely the same packet must be sent to the mediation devices 30a and 30b is required. However, since the client 70 transmits only one packet to the virtual server 11 and the servers 21, 22, and 23 transmit only one packet to the virtual router 37, the above condition cannot be normally satisfied. In the present embodiment, the above conditions are satisfied by combining the repeater hubs 83 and 84 with the NICs 31 and 32 in the promiscuous mode. In other words, the above condition is realized by combining the feature that the repeater hub transmits all packets to all ports and the feature that the NIC in the promised mode receives all packets that have reached the ports.
[0069]
Therefore, in the present embodiment, for example, instead of the repeater hub, for example, the switching hub cannot be simply replaced. Furthermore, when the repeater hub cannot be used, a means for separately copying the packet and transferring it to all the intermediary devices is required.
[0070]
A highly reliable server system according to a second embodiment of the present invention will be described with reference to the drawings. FIG. 14 is a block diagram illustrating the entire configuration of the high reliability server system according to the second embodiment.
[0071]
This embodiment is different from the first embodiment in that three intermediate devices 30 are provided. However, since the intermediary devices 30 operate without being aware of each other, the operation does not change regardless of whether there are two, three, or more.
[0072]
The difference is in the number of packets received by the server and the number of packets received by the client. In the first embodiment, since the number of the intermediary devices 30 is two, the same two packets are received at the maximum, so that the second one needs to be discarded. In the second embodiment, since there are three mediation devices 30, a maximum of three identical packets are received. Therefore, the server and the client need to discard the second and third packets.
[0073]
A highly reliable server system according to a third embodiment of the present invention will be described with reference to the drawings. FIG. 15 is a block diagram illustrating a packet control unit 45 according to the third embodiment.
[0074]
This embodiment differs from the first and second embodiments in that the packet control unit includes a server management table correction unit 46. Accordingly, the following operations are added.
[0075]
The operation of the server management table correction unit 46 will be described with reference to FIG. FIG. 16 is a flowchart illustrating the operation of the server management table correction unit 46.
[0076]
First, the server management table correction unit 46 transmits the contents of the server management table 42 managed by itself to the virtual server 11 using the external network-side packet transmission unit 36 (step SB1). Next, the server management table correction unit 46 receives the transferred server management table 42 using the external network side packet receiving unit 35 (step SB2). The information of the server management table 42 thus received is compared to determine whether they are the same (step SB3).
[0077]
If they are the same (YES in step SB3), after sleeping for a certain period of time without doing anything (step SB5), the contents of the server management table 42 managed by itself are again stored in the virtual server using the external network side packet transmitting unit 36. 11 (step SB1).
[0078]
If they are not the same (NO in step SB3), the server management table 42 managed by itself is corrected by some algorithm (step SB4). For example, when the operating states are different such as “normal” and “failure”, the “failure” is determined to be correct and corrected from “normal” to “failure”. Alternatively, it may be determined by a majority decision. The server changed to “failure” is disconnected from the system, and does not forward the request from the client thereafter. After the correction, after sleeping for a certain period of time (step SB5), the content of the server management table 42 managed by itself is transmitted to the virtual server 11 again using the external network side packet transmission unit 36 (step SB1).
[0079]
In this way, if the states of the intermediary devices 30 operating independently of each other become inconsistent for some reason, the server management table 42 can be corrected.
[0080]
A highly reliable server system according to a fourth embodiment of the present invention will be described with reference to the drawings. FIG. 17 is a block diagram illustrating an overall configuration of a high reliability server system 100 according to the fourth embodiment, and FIG. 18 is a block diagram illustrating a packet control unit 400 according to the fourth embodiment.
[0081]
This embodiment is different from the first embodiment in that the internal network 81 does not exist, and the high reliability server 20 and the mediation device 300 are both connected to the same network 820.
[0082]
Accordingly, the intermediary device 300 includes a packet receiving unit 350 having an internal network-side packet receiving unit 33 and an external network-side packet receiving unit 35. A combined packet transmitting unit 360 is provided. Further, the packet control unit 400 includes a pseudo ARP response unit 440 having a combination of the internal network side pseudo ARP response unit 43 and the external network side pseudo ARP response unit 44.
[0083]
As described above, elimination of the internal network 81 changes the configuration block diagram of the high-reliability server system. However, if the network 820 is logically divided into the external network 82 and the internal network 81, it is the same as the first embodiment. It is.
[0084]
Note that the servers do not need to have the same implementation if they return the same response. That is, versions, specifications, programming languages, types of compilers, compiler options, hardware or software, and the like may be different. Further, the server may be composed of one component or a plurality of components.
[0085]
In the above embodiment, the packet transmitting / receiving unit, the control unit, the server management table, the pseudo-ARP response unit, and the server management table correction unit are mounted on one mediation device, and the server is mounted on another device. The physical location of each component of the reliability server system is arbitrary. That is, for example, each component may be mounted on one computer, or may be separately mounted on different computers. Further, a part thereof may be mounted on the same computer.
[0086]
Further, in the above-described embodiment, an example in which there are three servers is shown, but there may be three or more servers. In addition, the number of servers may be three or less as long as the reliability against the server failure is allowed.
[0087]
Further, in the above embodiment, the case where there is one client is assumed, but there may be a plurality of clients.
[0088]
Further, in the first embodiment, the virtual IP address and the virtual MAC address of the virtual server 11 are selected from those which the communication device (including the mediation device 30) connected to the external network 82 does not have. The IP address of the NIC 32 of one mediation device 30 may be selected as the virtual IP address, or the MAC address of the NIC 32 of any one mediation device 30 may be selected as the virtual MAC address. However, it is important that all mediation devices 30 return the same ARP response to the ARP request for the virtual IP address.
[0089]
Further, in the first embodiment, the virtual IP address and the virtual MAC address of the virtual router 37 are selected from those which the communication device (including the mediation device 30) connected to the internal network 81 does not have. The IP address of the NIC 31 of one intermediary device 30 may be selected as the virtual IP address, or the MAC address of the NIC 31 of one intermediary device 30 may be selected as the virtual MAC address. However, it is important that all mediation devices 30 return the same ARP response to the ARP request for the virtual IP address.
[0090]
It should be noted that the above embodiment is illustrative, and the present invention is not limited to this. The scope of the present invention is defined by the appended claims, and all modifications that fall within the meaning of the claims are included in the present invention.
[0091]
【The invention's effect】
As described above, according to the present invention, even if one intermediary device transferring a request and a response between a client and a server fails, the transfer can be continued by another normal intermediary device. Prevents the entire system from being down, and enables a more reliable client-server system to be constructed. Also, since the virtual addresses are determined in advance, any intermediary device can process requests and responses addressed to these virtual addresses. Therefore, even if the intermediary device goes down due to a failure or the like, the server or client does not need to make any changes. That is, since existing servers and clients can be used without modification, a highly reliable client-server type system can be constructed at low cost.
[Brief description of the drawings]
FIG. 1 is a block diagram illustrating an overall configuration of a highly reliable server system according to a first embodiment.
FIG. 2 is a diagram showing how a system looks when a client sees an intermediary device;
FIG. 3 is a diagram showing how the system looks when the mediation device is viewed from the server.
FIG. 4 is a block diagram illustrating a configuration of an intermediary device.
FIG. 5 illustrates an example of a server management table.
FIG. 6 illustrates another example of a server management table.
FIG. 7 illustrates a format of a packet.
FIG. 8 illustrates a format of a packet.
FIG. 9 illustrates a format of a packet.
FIG. 10 illustrates a format of a packet.
FIG. 11 is a diagram illustrating the operation of the mediation device.
FIG. 12 is a diagram illustrating a flow of a request from a client.
FIG. 13 is a diagram for explaining a flow of a response from a server.
FIG. 14 is a block diagram illustrating an overall configuration of a highly reliable server system according to a second embodiment.
FIG. 15 is a block diagram illustrating a packet control unit 45 according to the third embodiment.
FIG. 16 is a view for explaining the operation of a server management table correction unit according to the third embodiment;
FIG. 17 is a block diagram illustrating the overall configuration of a highly reliable server system according to a fourth embodiment.
FIG. 18 is a block diagram illustrating a packet control unit according to a fourth embodiment.
FIG. 19 is a configuration diagram of a conventional high reliability server system.
[Explanation of symbols]
10 high reliability server system, 11 virtual server, 20 high reliability server, 21, 22, 23 server, 30a, 30b, 30c, 300a, 300b mediation device, 31a, 31b, 32a, 32b NIC , 33a, 33b: internal network side packet receiving section, 34a, 34b: internal network side packet transmitting section, 35a, 35b: external network side packet receiving section, 350a, 350b: packet receiving section, 36a, 36b: external network side packet Transmission section, 360a, 360b Packet transmission section, 37 Virtual router, 40a, 40b, 400a, 400b Packet control section, 41 Server control section, 42 Server management table, 43 Internal network side pseudo ARP response receiving section , 44... The external network side pseudo ARP response receiving unit, 4 .., A server management table correction unit, 70, a client, 81, an internal network, 82, an external network, 820, a network, 83, 84, a repeater HUB, 91, 92, 93, a computer, 1010, a highly reliable server system, 1111, 1112, 1113: copy, 1020: high reliability server, 1021, 1022, 1023: server, 1030: mediation device, 1031: control unit, 1032: server management table, 1041: client, 1050: network.

Claims (11)

In the server side system in the client server type system,
Along with transferring a request from a client to a server, and providing a plurality of intermediary devices including transfer means for transferring a valid response to the request received from the server to the requesting client,
A virtual address that appears as a single server from the client and as a single router from the server is determined in advance,
The server system, further comprising: a control unit configured to control the transfer unit to transfer the request and response addressed to the virtual address. The server system according to claim 1, wherein a TCP / IP protocol suite is used for communication between the client, the server, and the mediation device, and a virtual IP address in a network layer is assigned as the virtual address. 2. The server system according to claim 1, wherein Ethernet is used for communication between the client, the server, and the mediation device, and a virtual MAC address in a data link layer is assigned as the virtual address. 4. The server system according to claim 1, wherein the intermediary device includes a response unit that returns a response to a request for inquiring an address of a lower layer of the virtual address. 5. 5. The server system according to claim 4, wherein said response means responds to an ARP request. 6. The server according to claim 1, wherein the server includes a packet discarding unit that discards second and subsequent duplicate packets among a plurality of identical packets received from the mediation device. 7. Server system. 7. The client according to claim 1, wherein the client includes a packet discarding unit that discards the second and subsequent duplicate packets among a plurality of identical packets received from the mediation device. 8. Server system. 8. The mediation device according to claim 1, further comprising a correction unit configured to confirm the server management information with each other, and to correct the server management information to be in the same internal state when a mismatch occurs. The server system as described. The intermediary device, when receiving a request and a response addressed to the virtual address, includes copy creating / transmitting means for creating copies by the number of the intermediary devices and transmitting the copies to all the intermediary devices. Item 9. The server system according to any one of Items 1 to 8. An intermediary device interposed between a client and a server in a server-side system in a client-server type system,
Transfer means for transferring a request from the client to the server, and transferring a valid response to the request received from the server to the requesting client;
A virtual address that appears as a single server from the client and as a single router from the server is determined in advance,
Control means for controlling the transfer means to transfer the request and response addressed to the virtual address. In a method of concealing an error occurring in a server in a client-server type system from a client,
Between the client and the server, a plurality of intermediary devices having transfer means for transferring a request from the client to the server and transferring a valid response to the request received from the server to the requesting client are interposed. ,
A virtual address to be shown as a single server from a client and a server to be seen as a single router is determined in advance, and the request and response addressed to the virtual address are controlled so that the transfer means can transfer them. Error concealment method in a client-server type system.

Images