JPH1084377A - Mutual backup method and routing method for a plurality of servers - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide the mutual backup method for a plurality of servers by which the client server system with high reliability is realized at a low cost. SOLUTION: When a client requests update of data to a server A (access server), the server A inquires about a time stamp stored in all other servers B, C. Since the time stamp of the server A is To a time stamp reply from the servers R, C to the server A is To , the server A confirms it that data stored in the server A are not old. Moreover, a time T1 at this point of time is stored in the server A as the time stamp for updated data this time, and the server A makes data update request this time to the other servers B, C and distributes the time stamp T1 to them. The other servers B, C update the data and the time stamp similarly in the case of the server A. The server A receives data update reply from the other servers B, C and the data update reply is returned to the client.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a mutual backup method and a routing method for a plurality of servers in a client server system.

[0002]

2. Description of the Related Art A client server system is a distributed processing system in which a client and a server distributed on a network are integrated to execute one process as a whole. A client requests a predetermined request from a server that owns resources and provides a service, and the server executes a process based on the request and returns a result to the client. Currently, the network connection between the client and server includes Ethernet, TCP / I
LAN (Local Area Network) connection using P is mainly used.

[0003] Here, as an example of a wide area network,
Multiple LANs are scattered over a wide area, and each LAN
A WAN (Wide Area Network) configured such that a relay node exists therein and a client in a certain LAN can communicate with a server in a remote LAN via these relay nodes.
Consider k).

When the number of LANs is small as shown in FIG. 6, the LAN where the server exists (indicated by a double circle で は in the figure,
Hereinafter, the central LAN 1) and other LANs
(Indicated by a circle in the figure) are connected in a star shape. In a network that performs such a star connection, the central LAN
1 and each LAN are directly connected,
When the number of LANs increases, a huge number of trunk lines are required.

[0005] Therefore, as shown in FIG.
N is not directly connected to the central LAN 1, but LANs that are close to each other are grouped into one group, and one base LAN 2 (shown by a black circle in the figure) is set for each group, and a star connection is made for each group. And then the center LA
N1 and these base LANs 2 are connected in a star configuration. As described above, the remote LAN is connected to the central LA via the base LAN 2.
If it is configured to be connected to N1, connection to all LANs can be made with a small number of trunk lines, and an inexpensive network can be constructed.

By constructing such a wide area network,
Conventionally, in a client-server system in which a client uses a server arranged therein, two methods for improving reliability have been considered.

The first is a method of increasing the reliability of the server itself, in which the components of the server are configured in a redundant configuration, and when a specific component of the operating device fails, a standby state is established. Is immediately backed up by the parts. Disk mirroring, in which the same information is simultaneously written to two hard disks or two hard disk partitions, is effective in the fault tolerant system of the server having such a redundant configuration, and is generally widely used.

In the second method, a detour is prepared in advance in the central LAN 1 so that client servers on the network are connected by a plurality of trunk lines.

FIG. 8 shows a network in which four LANs 2 in the WAN shown in FIG. In this network, the base LA
In addition to the trunk line that directly connects N2 and the central LAN1,
A detour relay line 3 that connects a plurality of base LANs 2 is provided. Then, even if a fault occurs in the trunk line of a specific route, communication between each LAN and the central LAN 1 is backed up by the bypass trunk line 3, and the reliability of communication between the client and the server is improved.

[0010]

However, in order to increase the reliability of the client-server system by the first method, conventional mirroring of the disk alone is not sufficient, and all of the CPU, bus, memory, etc. are backed up. There is a need. However, general-purpose workstation, PC
However, if those parts are to be made redundant, it becomes very expensive.

In the second method, as shown in FIG. 8, since a large number of trunk lines are concentrated on the base LAN 2 and the central LAN 1, it is necessary to upgrade the processors constituting the relay nodes. . Therefore, the number of LANs in one group is reduced, and the LAN
It is also conceivable to increase the number of servers or move some of the servers of the central LAN 1 to the base LAN 2 to distribute the load on the relay nodes and reduce the load on the central LAN 1. However, when the load is distributed, there is a problem that a routing function for bypassing the relay node becomes complicated, and costs for network maintenance and the like increase.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-described problems, and a first object of the present invention is to provide a method for mutually backing up a plurality of servers which can realize a client server system having high reliability at low cost. That is.

[0013] A second object of the present invention is to provide a plurality of servers which can reliably connect to another server so that the reliability of a client server system can be improved even if a client cannot communicate with one server. Is to provide a routing method.

[0014]

The invention according to claim 1 is
In a method of mutually backing up a plurality of servers in a client server system in which a plurality of servers are arranged and communicably connect between an access server specified by a client and the client, the access server may update / restore data from the client. In response to the reference request, a determination is made by communication between the servers as to whether or not the data held in each of the servers is the latest one, including the sameness of data.

According to a second aspect of the present invention, the access server instructs the data update only to a server holding data determined to be the latest data among other servers.

According to a third aspect of the present invention, when it is detected that the data stored in the access server is old, the data update / reference request is rejected.

According to a fourth aspect of the present invention, the access server inquires of another server about a time stamp, and the data held in each of the servers is determined based on the response result.

According to a fifth aspect of the present invention, when the data update request detects that the data held in the access server is old, the data update / reference request from the client is made in response to the next data update / reference request from the client. The data update / reference request is rejected without performing communication between the servers.

According to a sixth aspect of the present invention, the server that has failed to update the data loads the latest data from another server based on a return request from the client.

According to a seventh aspect of the present invention, in the method for mutually backing up a plurality of servers in a client-server system in which a plurality of servers are arranged and an access server specified by a client is communicably connected to the client, In each server, when a data update request from the client to the plurality of servers conflicts with each other, the communication between the servers notifies each other that the data update request has conflicted, and the server updates the data update request. It is characterized by refusal.

According to an eighth aspect of the present invention, each of the clients, for which the data update request has been rejected from each of the servers, retries the data update after a lapse of time different from each other.

According to a ninth aspect of the present invention, each of the servers includes:
By mutually inquiring the time stamps, a conflict determination of the data update request is made depending on whether or not the data is being updated.

According to a tenth aspect of the present invention, in the method of routing a plurality of servers in a client-server system in which a plurality of servers are arranged and an access server specified by a client is communicably connected to the client, Servers are arranged on a wide area network connected via a relay node, and different priorities of access servers are set for each of the clients.

According to an eleventh aspect of the present invention, the priority order of the access servers is set in ascending order of the distance from the client to each server.

[0025]

BRIEF DESCRIPTION OF THE DRAWINGS FIG.
An embodiment of the present invention will be described.

The method according to the first embodiment of the first embodiment have the same function, in a client-server system where the server holding the same data are arranged a plurality redundant, when a failure occurs in one server This is a mutual backup method for a plurality of servers, which is applied to a configuration in which a client issues a request to another server again.

In order to allow a plurality of servers to back up each other, it is necessary for each server to hold the same data. However, even if a plurality of servers have the same data in the initial state, the data of the server generally changes by executing a data update (add / change / delete) request from the client. In order for all servers to keep the same data, a certain server receives a data update request from a client, and when the server executes the request, the same data must be updated in other servers.

When one of a plurality of servers fails to update data due to some kind of failure when a data update is requested from a client, the data of this server becomes out-of-date. This server must return a rejection response to both a data update request and a data reference request from a client. This way, the client knows that it can't use the other server
The request can be promptly reissued to another server.

Each server stores an identification code (hereinafter, simply referred to as a time stamp) indicating the time at which the data held therein was last updated as a means for knowing whether or not the data is the latest. By comparing this time stamp between servers, including the data identity,
You can determine if it is the latest one.

Next, a communication procedure when three servers A, B and C mutually back up a client will be described.

FIG. 1 shows a case where all the servers A to C can operate normally, FIG. 2 shows a case where the server A which has received a request from the client has a failure, and FIG. This shows a communication procedure when the server B that has not received a fault has a failure.

In FIG. 1, a client is a server A
When requesting the (access server) to update the data, the server A inquires the time stamps held in all the other servers B and C. Since the time stamp of the server A is T ₀ and the time stamp responses from the servers B and C to the server A are both T ₀ , it is confirmed that the data held in the server A is not old. it can. Further, the time T ₁ at this point is stored in the server A as a time stamp of the current data update, and the time stamp T ₁ is distributed from the server A to the other servers B and C together with the current data update request. . Other server B,
In C, the data and the time stamp are updated as in server A. The server A receives a data update response from the other servers B and C, and then returns a data update response from the server A to the client.

FIG. 2 shows a communication procedure for rejecting a server A which has received a data reference request or a data update request (data reference / update request) from a client and has a failure.

When a data reference / update request is issued from the client to the server A, the server A inquires the servers B and C about the time stamp. When the time stamp responses from the servers B and C return to the server A as T ₁ , respectively, the time stamp of the server A becomes T ₀ (<T
_{From 1} ), it is known that the server A has a failure. That is, the time stamp T ₀ of the server A is compared with T _1, and since T ₀ <T _1, it is determined that the data held by the server A is not the latest. Then, in response to the data reference / update request from the client, the server A returns a rejection response to the client. The client whose data reference / change request is rejected by the server A requests the server B for data reference / change. Even on server B,
The same time stamp inquiry and response procedures as described above are repeated (omitted in the figure). When it is detected that the data of server B is the latest, a data reference / update response is sent from server B to the client. Will be issued.

FIG. 3 shows a communication procedure when the server B is excluded from the distribution target of the data update request.

When the client requests the server A to update the data, the server A inquires the servers B and C about the time stamp. Timestamp Reply is T ₀ from the server B to server A, when the timestamp reply from server C to Server A is returned as T ₁ (> T _0), since the timestamp of the server A is T ₁ It can be seen that the server B has a failure. That is, when the data of the server B is not the latest, the server A can detect the failure of the server B by inquiring the time stamps of the other servers B and C and responding to them. As a result, server A becomes server B
As distribution excluded, there without distributing the data, only the server C, and distributes the time stamp T ₂ together with the current data update request.

After each of the servers A and C performs data update, the server A receives a data update response only from the server C, and then returns a data update response to the client. Therefore, as shown in FIG. 3, the data in server B is not updated, and its time stamp is also T
It is kept at ₀ .

As described above, in the method of the first embodiment, a plurality of servers are arranged, and a plurality of servers in a client-server system for communicably connecting an access server specified by a client and the client. In the mutual backup method, is the access server up-to-date with respect to the data update / reference request from the client, including the identity of the data held in each server, through communication between the servers? Is determined, the CPU, bus,
Even if a memory or the like is not specially configured in a redundant manner, an inexpensive server having high reliability can be configured using a general workstation or PC, and mutual backup of the plurality of servers becomes possible.

Embodiment 2 The method of the second embodiment is a mutual backup method of a plurality of servers which is applied to a recovery process for updating data between servers mutually backing up when a server fails to update data. .

As described in the first embodiment, in response to a data update request from a client, the data of the server that failed to update the data becomes older than the data of the other servers that are backed up with each other. Therefore, in order to respond to the request from the client, it is necessary to immediately rewrite the data content so as to match the latest data content.

Until the data of the server is restored, every time an access is made from the client or every time an access is made to another server, the inquiry and response of the time stamp are repeated. Decrease network processing efficiency. Therefore, in the server that failed in data update, when the next data update / reference request is made, communication between the servers is not performed, and the data update / reference request is immediately rejected to the client. Does not lower the processing efficiency.

FIG. 4 shows a procedure in which three servers A, B and C mutually back up each other and always rewrite data to the latest one.

When the server A finds that its own data is older than the data of the other servers B and C, its time stamp is rewritten to "-1". Then, when the server A receives a request from the client after that, the server A becomes the other server B,
The time stamp is not inquired to C, and a refusal to the client is immediately returned based on the time stamp "-1". by this,
Useless inquiries and responses between servers can be prevented, and network processing efficiency can be improved.

Also, a server A having out-of-date data
Sends a time stamp inquiry to the other servers B and C when there is a "return request" from the client. Then, based on the time stamps returned from both servers B and C, only one server holding the latest data is selected. In Figure 4, the server B, and the time stamp from C is T ₀ Both server B, and may be either a C, relative now for example, a server B, and performs data loading request from the server A, the data in the server B received a loading request and sends a time stamp T ₀ with the latest data to the server a (data loading response). As a result, the data content of the server A is rewritten to the latest one, and a "return response" is sent to the client at that time.

Here, while the server A is updating the data, the updating of the data by the other servers B and C must be prohibited. Therefore, when the server A receives the time stamp inquiry from the other servers B and C during the period from the reception of the return request from the client to the return of the “return response”, the server A will be described below. As in the case of the third embodiment, the effect that "data is being updated" is returned.

As described above, according to the method of the second embodiment, when it is detected that the data held in the access server is old by the data update request, the next data update / reference request from the client is performed. In this case, the server instructs to reject the data update / reference request without performing communication between the servers, or, based on the return request from the client,
By loading the latest data from other servers, the server can have a mutual backup function without changing the application server facilities on the existing network and without reducing the efficiency of the client server system, Reliability can be improved.

Embodiment 3 The method according to the third embodiment provides a client-server system in which a plurality of servers having the same function and holding the same data are redundantly arranged.
This is a mutual backup method for a plurality of servers which is applied to a configuration in which requests from a plurality of clients are ranked for each server that is backing up each other.

For example, when two servers A and B hold the same data and back up each other, two or more clients access the servers A and B simultaneously and refer to the respective data. It is possible to do. However, while a data change request from a client is being executed on server A,
If the same data is deleted in response to a request from another client, there is no longer any data to be changed in the server B. In such a case, even if the server A issues a data change request to the server B, the data cannot be updated on the server B. Therefore, in a client-server system including two or more servers, when data update requests compete in this way, it is not possible to maintain data identity between servers.

Here, when a data update request conflicts, the fact is notified by communication between the servers, and the request is rejected to each client, so that the data between the servers A and B is identical. Is secured.
Further, when the same request is made again to the servers A and B from each client whose request has been rejected, the timing of the request is made different between the clients so that access conflict does not occur again.

FIG. 5 shows a communication procedure when the clients a and b make data update requests to the servers A and B, respectively. Here, for example, it is assumed that the client a makes a request in the order of the server A and the server B, and the other client b makes a request in the order of the server B and the server A.

In the figure, when the client a makes a data update request to the server A and the client b makes a data update request to the server B at the same time, each of the servers A and B inquires about the time stamp held by the other server. Server A receives a time stamp inquiry from server B before returning a response to client a. That is, the server A receives the inquiry from the server B before receiving the response to the inquiry about the time stamp from the server B. Then, instead of the time stamp response, the server A responds to the server B that "data is being updated". The server B receives the "data update in progress" response and similarly responds "data update in progress" to the client b, thereby rejecting the data update request from the client b. Similarly, server B waits until server A returns a response to client b.
Receives an inquiry about the time stamp from the server A, and operates in the same manner as the server A, so that the server A rejects the data update request from the client a in response to the response of "data updating".

[0052] Thereafter, each client a request is rejected, the b, in arbitrary time T _a, which is determined by a random number circuit is respectively incorporated, after a T _b, the server again A, the same request to B retry Is done. As described above, by making the access timings to the server different between the clients a and b, the data update requests from the two clients a and b are executed in order without causing access conflict again. For example, if T _a <T _b , a data update request from client a is first executed on server A, and the time stamp of each of servers A and B is changed from T ₀ to T
Rewritten to _1, then the data update request from the client b is performed by the server B, and each server A, the time stamp of B is rewritten to T _2.

Further, even in a client server system in which three or more servers back up each other, the conflict processing is similarly executed. If the request from any one of the clients is a data reference request or the like that does not involve addition / change / deletion of data, the server does not indicate “data updating”.

As described above, in the method according to the third embodiment, a plurality of servers are arranged, and a plurality of servers in a client server system for communicably connecting an access server specified by a client and the client. In the mutual backup method, each server
When a data update request from a client to a plurality of servers conflicts with each other, communication between the servers informs each other that the data update request has conflicted, and rejects the data update request. Since each client whose data update request was rejected by has been configured to retry data update after a different time elapses,
Competition for data update requests can be properly handled. Therefore, when data referencing requests are simultaneously made from a plurality of clients, by distributing the requests to a plurality of servers, the processing capacity of the system can be improved in proportion to the number of servers.

Embodiment 4 The method according to the fourth embodiment is a routing method for a plurality of servers applied to a client-server system configured to simultaneously execute data reference requests from a plurality of clients in a distributed manner.

Here, a plurality of servers having exactly the same function and data are distributed and arranged in each region on the network, for example, at each base LAN 2 in FIG. Typically, a client accesses a server located closest in the network. At that time, a failure of the trunk line in the network or a failure of the server itself occurs,
If the server cannot respond, the client accesses the next nearby server. Further, if this server cannot respond, it accesses the next server.

As described above, if the priority order of the access server is set for each client according to the distance to each server, each client sequentially and efficiently switches servers for all servers, and performs routing. Can be performed. The priority of the access server can be set based on other factors regardless of the distance from the client to each server.

In the routing method according to the fourth embodiment, a plurality of servers are arranged, and in a client-server system for communicatively connecting an access server specified by a client and the client, a plurality of servers are arranged. Servers are located on a wide area network connected via a relay node, and different access server priorities are set for each client, so that even if a client cannot communicate with one server, The communication with the server can be performed efficiently, so that the reliability of the communication between the client and the server can be improved. In particular, if servers are dispersedly arranged for each base LAN, it is more effective in increasing reliability.

Also, since the client normally communicates with the nearest server, it is effective in distributing the communication between the client and server to reduce the load on the relay node on the network.

In the above description of the first to fourth embodiments, it has been described that the client has the routing function of a plurality of servers and each server has the mutual backup function. If the entity is added as a client server system, the reliability of the server can be improved on the existing network without changing the existing application server equipment itself.

[0061]

According to the method for mutually backing up a plurality of servers according to the present invention, a highly reliable client server system can be realized at a low cost in realizing an application server function over a wide area network.

Further, according to the routing method of the present invention, communication between the client and the server can be distributed, and a highly reliable client-server system can be realized at low cost.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating a server A according to a first embodiment of the present invention.
FIG. 11 is a diagram showing a backup procedure in a case where .about.C can operate normally.

FIG. 2 is a diagram illustrating a server A according to the first embodiment of the present invention;
FIG. 6 is a diagram showing a backup procedure when a device has a failure.

FIG. 3 is a diagram illustrating a server B according to the first embodiment of the present invention;
FIG. 6 is a diagram showing a backup procedure when a device has a failure.

FIG. 4 is a diagram showing a data rewriting procedure according to Embodiment 2 of the present invention.

FIG. 5 is a diagram showing a procedure for preventing contention of a data update request according to the third embodiment of the present invention.

FIG. 6 is a diagram showing a network composed of four LANs.

FIG. 7 is a diagram showing a network including a plurality of base LANs.

FIG. 8 is a diagram illustrating a network in which four base LANs are directly connected by a trunk line.

[Explanation of symbols]

1 Central LAN, 2 LANs, 3 Detour trunk line.

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁶ Identification code Agency reference number FI Technical display location H04L 12/00

Claims (11)

[Claims] 1. A method for mutually backing up a plurality of servers in a client server system in which a plurality of servers are arranged and communicably connect between an access server specified by a client and the client, wherein the access server includes the client In response to a data update / reference request from the server, it is determined by communication between the servers whether or not the data held in each server is the latest data including the sameness. Characteristic mutual backup method of multiple servers. 2. The multiple server according to claim 1, wherein the access server instructs the data update only to a server holding data determined to be the latest data among other servers. Mutual backup method. 3. The mutual backup method for a plurality of servers according to claim 1, wherein when detecting that the data held in the access server is old, the data update / reference request is rejected. . 4. The multiple server according to claim 1, wherein the access server inquires of another server about a time stamp, and determines the data held in each of the servers based on a response result. Mutual backup method. 5. The method according to claim 1, wherein when the data held in the access server is detected to be old by the data update request, the next data update from the client is performed. A method of mutually backing up a plurality of servers, wherein the server does not perform communication between the servers in response to the request / reference request and rejects the data update / reference request. 6. The method of claim 5, wherein the server that failed to update the data loads the latest data from another server based on a return request from the client. . 7. A method for mutually backing up a plurality of servers in a client server system in which a plurality of servers are arranged and communicably connect between an access server specified by a client and the client, wherein each server includes the client When the data update requests to the plurality of servers conflict from each other, communication between the servers notifies each other that the data update requests have conflicted, and rejects the data update request. Backup method for multiple servers. 8. The multiple server according to claim 7, wherein each of the clients, for which the data update request is rejected from each of the servers, retries the data update after a lapse of time different from each other. Mutual backup method. 9. The server according to claim 7, wherein the servers inquire about each other's time stamps and determine whether or not the data update request is in conflict according to whether data is being updated. Mutual backup method for multiple servers. 10. A method of routing a plurality of servers in a client server system in which a plurality of servers are arranged and communicably connects an access server specified by a client and the client, wherein the plurality of servers are relay nodes. A plurality of servers arranged on a wide area network connected via a network, wherein different priorities of access servers are set for each of the clients. 11. The method according to claim 10, wherein the priority order of the access servers is set in ascending order of the distance from the client to each server.