JP2001092766A - Client server system and naming service method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a dynamic load distribution naming service suitable to a client server system having many clients. SOLUTION: In the client server system consisting of plural server machines 2 to be connected with the clients 1 via a network 3, respective server machines 2 are provided with plural kinds of server objects 40 to offer information service to the clients and naming service objects 50 to specify the server objects to execute the information service to the clients. The respective naming service objects are provided with routines 505 to collect load information of other server machines and executes dynamic naming service according to load states of the respective server machines to client terminals.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a client server system and a naming service method, and more particularly, to a plurality of server programs each of which can selectively provide the same information service from two or more server machines. The present invention relates to a client-server system and a naming service method in which in a client-server system including a plurality of server machines provided with (server objects), a server object can be dynamically allocated to each client terminal so as to distribute the server load. .

[0002]

2. Description of the Related Art A plurality of server machines are provided with a server program (server object) having the same function, and a client terminal (hereinafter, simply referred to as a client) is provided with an information service from one of the server machines. In a client-server system configured to provide services, the server load is distributed as much as possible so that service requests from a large number of clients are not concentrated on a specific server or a small number of servers, and service performance to the clients does not deteriorate. In this way, it is necessary to assign a server or server object to a client (hereinafter referred to as naming).

As a naming service method in consideration of load distribution, for example, Japanese Patent Application Laid-Open No. Hei 7-93238 (Prior Art 1) discloses that a client requests a name server to acquire a server address, There is proposed a server selection method in which a server with a low load is selected from among the above, and the address is notified to the client to dynamically distribute the load. Further, Japanese Unexamined Patent Publication No.
In Japanese Patent No. 319689 (Prior Art 2), a dynamic link library (DLL) is provided on the client side, and when a server name is designated by a client program, the DLL is transmitted from the name server to the server in ascending order of load. By obtaining an address list specifying the selection order, transmitting a probe packet to each server according to the address list, and determining a response, a low-load server providing a desired service specified by the server name can be identified. There has been proposed a server selection method for selecting a server.

[0004]

However, in the above-mentioned prior arts 1 and 2, in addition to a plurality of server machines that provide information services to clients, a naming service dedicated to performing a naming service while monitoring the load state of these servers. Need a server. As described above, in the client server system in which the dedicated name server is installed, since the inquiries are concentrated on the name server, there is a problem that the naming service decreases as the number of clients increases. Further, in the prior art 2, it is necessary to equip an unspecified number of client terminals with DLLs each having a special function.

An object of the present invention is to provide a client server system and a naming service method that can speed up a naming service. It is another object of the present invention to provide a client server system and a naming service method capable of dynamically distributing the load of a server machine and accelerating a naming service.
It is still another object of the present invention to provide a client server system and a naming service method that can efficiently collect server load information and reflect it in a naming service even when the number of server machines and the number of clients increase.

[0006]

In order to achieve the above object, a client server system of the present invention comprises a plurality of server machines connected to a client terminal via a network, and each server machine is connected to a client terminal. A plurality of types of server objects for providing an information service to the client terminal, and a naming service object for designating a server object on which the information service is to be executed on the client terminal. And a means for collecting dynamic load information for client terminals by executing a dynamic naming service according to the load state of each server machine. According to the above configuration, the naming service function is distributed to the plurality of server machines for the information service. Therefore, even when the number of clients connected to the server system increases, the naming service request is transmitted to the plurality of server machines. , A quick response is possible.

More specifically, in the client server of the present invention, each of the naming service objects communicates load information with another server machine and stores the collected load information of a plurality of server machines. The load information table and the load information table are referred to, in order from the server machine in the low load state, it is determined whether there is a server object that matches the request from the client terminal, and the server object to be assigned to the client terminal is determined. And means for performing the operation. According to one embodiment of the present invention, one of the plurality of server machines has a naming service object for master for generating a load information notification message, and each of the communication means includes the naming service object for master. The generated load information notification message is sequentially transferred to the next naming service object so as to be circulated to a plurality of naming service objects, and the load information of each server machine is exchanged via the load information notification message.

According to the above configuration, each naming service object acquires load information of another server machine from the load information notification message, and sets new load information of its own server machine in the load information notification message. By transferring to the next naming service object, it becomes possible to grasp the load state of the other plurality of server machines every time the load information notification message is received. If no response is received within a predetermined time to the transmitted load information notification message, information indicating that the destination naming service object has failed is set in the load information notification message, and the By sending the naming service object to the naming service object located next to the failure naming service object, the naming service object can be notified of the failure information together with the load information of the server machine. in this case,
By storing the failure information in the load information table, when a naming service request is received from a client terminal, each naming service object excludes a server machine corresponding to the failure naming service object and allocates a server object. An action can be taken.

[0009] If the naming service object designated as the master holds the load information notification message that has cycled through the plurality of server machines until the next transmission timing, the load information notification message is circulated at a predetermined cycle. be able to. Also, the naming service object designated as the master acquires the load information of the other server machine from the load information notification message that has cycled through the plurality of server machines, and adds the new load information of the own server machine to the load information notification message. After the setting, the load information notification message can be continuously circulated if the operation is performed so that the load information notification message is immediately transmitted to the naming service object of the next server machine.

In the former case, the number of times the load information notification message circulates is reduced, so that the communication load on the network and the overhead for updating the load information table can be reduced. In the latter case, the load information communication message circulates more frequently, so the load information is updated frequently, and each naming service object
A naming service based on the latest load information of another server machine becomes possible.

In the case of a large-scale client server system having a large number of server machines, in the present invention, the server machines are divided into a plurality of groups, and a master naming service object is designated for each group. in this case,
Each communication means sequentially transfers the load information notification message generated by the master naming service object to the next naming service object so as to circulate in the group, and transmits the load information notification message to each server machine via the load information notification message. Communicate load information.

In the large-scale client-server system, when it is desired that each naming service object execute a naming service in consideration of load information of server machines belonging to another group, the present invention provides a plurality of naming services for masters. One of the objects
First, a function of generating an inter-group communication message is provided, and each communication means in the master naming service object sequentially transfers the inter-group communication message to another master naming service object, and transmits the inter-group communication message through the message. Communicates server machine load information.

According to the above configuration, each master naming service object collects the load information of each server machine of another group by the above-mentioned inter-group communication message, and transfers the load information of each server machine of its own group to another group. Can notify. Therefore, the load information of each server machine of another group collected by the inter-group communication message is set in the above-mentioned load information communication message, and circulated within the own group, so that the naming service objects of the own group are assigned to the server of the entire system. Machine load information can be distributed.

A naming service method according to the present invention comprises a plurality of server machines connected to a client terminal via a network, wherein each server machine provides a plurality of types of servers for providing an information service to the client terminal. In a client-server system including an object and a naming service object for designating a server object to execute an information service on a client terminal, (A) the naming service object designated as a master among the naming service objects is Generating a load information notification message that causes the plurality of server machines to make a round, setting new load information of the own server machine in the load information notification message, and transmitting the message to the network;
(B) The naming service object which is the destination of the load information notification message acquires the load information of another server machine from the load information notification message, and adds new load information of the own server machine to the load information notification message. Setting and sending to the naming service object of the next server machine; and (C)
A naming service object that has received a naming service request from a client terminal, sequentially determines whether there is a server object that matches the request from a server machine in a low load state, and allocates a server object to the client terminal. Features.

[0015]

Embodiments of the present invention will be described below in detail with reference to the drawings. As shown in FIG. 1, the client server system of the present invention includes a plurality of client terminals 1 (1A, 1B,...) Interconnected by a high-speed communication link 3.
1N) and a plurality of server machines 2 (2A2, B2,...).
W). The communication between the servers and between the server and the client is performed using a standard networking protocol such as TCP / IP, but the communication between the clusters described later requires special specifications other than TCP / IP. May be applied.

In the client server system of the present invention, each server machine 2 is provided with a server program 4 and a naming service program 5. The server program 4 includes a plurality of programs (hereinafter, referred to as “programs”) for providing various information services in response to requests from clients.
Server object: called SO) 40 (40-1, 4)
0-2, ...). The naming service program 5 includes a load information communication function for periodically updating the load information table 51 while mutually exchanging load information among a plurality of servers, and a naming service program 5 for naming a desired server object from a client. When the service request NRQ is received, the optimum server machine in the light load state is searched from the load information table 51, and the address information (positional information and object reference) of the desired server object 40-i operating on the server machine is obtained. A naming service function for replying with a naming message NRY. In the following description, the naming service program 5 having the load information communication function is referred to as a naming service object (NSO) 50.
I will call it.

In the client server system of the present invention, as described above, each server machine 2 on which the server object 40 operates is provided with a naming service function. Services can be requested, and concentration of naming service requests to a specific name server as in the conventional system can be avoided. In this case, for each client,
A specific naming service object (NSO) as a destination of the naming service request may be specified in advance, or each client randomly selects a destination NSO from a plurality of naming service objects each time a naming service request is made. You may make it.
In the client server system of the present invention, each naming service object includes its own server,
The load state of a plurality of server machines constituting the client server system is periodically collected, and the address information of the server object running on the server machine in the low load state is named to the client a naming message N.
Since the notification is made by RY, as a result, the service processing request SRQ from each client is issued to the low load server machine, and the dynamic load distribution of the server machine is realized. In the client-server system shown in FIG. 1, each server machine 2 relays a received message to an adjacent server in order, as indicated by a broken line, so that a load information communication message M is transmitted between a plurality of server machines.
1 is transmitted cyclically.

FIG. 2 shows the load information notification message M1.
An example of the format is shown below. In the figure, 61 is IP
The header, 62 is a TCP header, 63 is a data field, and the load information notification message M1 is set in the data field 63. Load information notification message M
1 includes an identifier 631 indicating the type of message, a message sequence number 632, an identifier 633 of the NSO that has transmitted the message, an NSO number 634 indicating the number of subsequent load information data blocks, and a plurality of load information items. It comprises data blocks 635-1 to 635 -m and a footer 636 indicating the end of the message (IP packet). The load information data blocks 635-1 to 635
"-M" corresponds to a server machine included on the circulation route of the load information notification message M1, and includes an NSO identifier 635a and a load information field 635b.

The load information notification message M1 is shown in FIG.
As shown in (1), among a plurality of server machines constituting the client-server system, it is generated by an NOS (Restart Master NSO) 50A on a specific server designated in advance as a restart master. When the NOS on each server machine receives the load information notification message M1, the load information data blocks 635-1 to 635-635 are received.
The load information of the other server indicated by m is stored in the load information table 51, and the load information of the data block corresponding to the own NOS identifier in the received message is updated to the load information of the current own server machine. N of data block
Transfer to the server machine indicated by the OS identifier. in this case,
The restart master NSO 50A may immediately relay the cycled load information notification message M1 to the next NSO: 50B so that the load information notification message M1 continuously circulates, or the restart master NS.
The load information notification message M1 may be periodically circulated by the O50A holding the cycled load information notification message M1 until the next transmission timing.

When a failure occurs in the server machine on which the restart master NOS 50A operates, the circulation of the load information notification message M1 is interrupted, and the other NSOs 50B to 5N operate normally.
Since the load information cannot be collected at 0F, the load information notification message M is substituted for the restart master NOS 50A.
A spare restart master NSO that generates 1 is specified in advance. The spare restart master NSO includes, for example,
The NSO 50B that first receives the load information notification message M1 from the restart master NOS 50A is selected. When the reception of the load information notification message M1 is interrupted for a predetermined time or more, the standby restart master NSO 50B determines that a failure has occurred in the restart master NOS 50A and starts the operation as the restart master.

Upon receiving the load information notification message M1, each server machine or NSO returns a reception response message to the source address (DA) 612 of the IP header. After performing the above-described load information collection and update processing, each NSO rewrites the source address (SA) 611 and the destination address (DA) of the IP header, and transmits the load information notification message M1 to the next NSO. Forward. If the reception response message is not received within a predetermined time after transferring the load information notification message M1,
It is determined that a failure has occurred in the destination NSO. For example, in FIG. 3, after the NSO 50D transmits the load information notification message M1 to the NSO 50E, if the reception response message is not received within a predetermined time, the NSO 50D
D sets a failure code in the load information field corresponding to the NSO 50E in the load information notification message M1, and transfers this to the NSO 50F located next to the failure NSO 50E as a new load information notification message M1 '. The relationship between the failure NSO50E and the next NSO50F is
It is determined by the arrangement of the load information data block 635 in the load information notification message M1.

By transferring the load information notification message M1 'described above, N
The server machine 2E on which the SO 50E operates can be notified that the server machine 2E is in a failure state, and each NSO can execute a naming service excluding the server machine 2E.
When the server machine 2E recovers from the failure, the NSO
50E, for the next NSO 50F, a re-entry request message M including the identifier of the NSO 50E and the load information.
3 in the NSO 50F having received the re-entry request message M3, when receiving the next load information notification message, rewriting the failure code included in the load information data block of the NSO 50E into the load information, and
Operate to transfer to SO50A. by this,
All other NSOs can be notified of the failure recovery of the NSO 50E, and the load information notification message can be relayed from the NSO 50D to the recovery NSD 50E. When a failure occurs immediately after receiving the load information notification message M1 in any of the server machines, the circulation of the load information notification message is interrupted in the failed server machine. in this case,
The restart master 59A generates a new load information notification message by monitoring the time, and in the process of circulating the load information notification message, a failure occurrence location can be detected by a process similar to that of the NSO 50E described above.

When the number of server machines 2 constituting the client server system is relatively small, the load information communication message M1 can be circulated to all NSOs in a relatively short time. However, in a large-scale client-server system including a large number of server machines, it takes time to circulate the load information notification message M1, and the load information notified to each NSO has a value greatly deviated from the actual machine state. there is a possibility. In the present invention, when the number of server machines increases, as shown in FIG. 4, these server machines are divided into a plurality of groups (clusters) G1, G2, G
3, the restart master is designated for each group, and the load information communication message M1 is circulated within each group, thereby localizing the circulating range of the load information communication message M1, and the load information due to the relay delay. Prevent deterioration of freshness.

When the server machines are clustered as described above, the naming service may be performed by each naming service object with load distribution limited to each group. To do so, it is necessary for each naming service object to circulate a load information notification message including load information of all groups. In order to communicate the load information of all groups, for example, a plurality of NSOs (50A, 150B, 15
0C), the inter-group communication message M2 is circulated, and the other clusters are notified of the server load status in each cluster via the inter-group communication message M2.
In this case, in order to generate the inter-group communication message M2, one of the restart masters NSO is specified in advance as the inter-group communication master. According to the above configuration, each restart master reflects the load information of each server machine of the other group collected in the inter-group communication message M2 in the load information communication message M1 circulating in the own group, whereby the entire system is updated. Server load information can be notified to each naming service object.

FIG. 5 shows a configuration of a program provided in each server machine 2. In the server machine 2, an operation system (OS) 60 having a function of communicating with the client 1 via the network 3 and a naming service program (naming service object: NS)
O) 5 and plural types of server objects 40-1 to 40-4
A server program 4 having 0-n is running.

Naming service object (NS
O) 5 is an initialization processing routine 501, a distribution processing routine 502, a naming service processing routine 503, a time activation processing routine 504, a load information communication processing routine 505, a time monitoring processing routine 506, an inter-group communication processing routine 507, a re-entry request. It consists of a processing routine 508, and these plural component routines operate in multi-process and multi-thread. Also, the naming service object is
The table referred to by the component routine includes a load information table 51, an initialization definition file 52, a re-entry request buffer 53, and a server object (SO) position information table 54.

FIG. 6 shows the configuration of the load information table 51. The load information table 51 is a table referred to by the naming service processing routine 503 when a naming service request is received from the client 1.
The contents are updated by the load information communication processing routine 505, the inter-group communication processing routine 507, and the re-entry request processing routine 508. Load information table 51
Are the groups G1 to G1 in the cluster configuration shown in FIG.
Each group ID 510 indicating an identifier of G3 has a plurality of load information records corresponding to naming service objects belonging to each group. Each load information record includes an NSO identifier 511 indicating a naming service object and the naming service object. Includes load information 512 indicating the load state (CPU utilization) of the server machine in which is located, the number of service allocations 65, and a flag 514. It should be noted that the group ID 510 is unnecessary in a relatively small client-server system that does not require clustering.

The group ID 510 is arranged in the transfer order of the inter-group communication message M2 with the group communication master as a starting point, and the NSO identifier 511 is transferred with the load information communication message M1 starting with the restart master in each group (cluster). They are arranged in order. The service allocation count 65 is prepared as an index for indicating an increase in load within the update cycle of the load information table. The value of the allocation count 65 is cleared each time the load information table is updated. It is incremented each time any server object on the server machine is named. The flag 514 is used to indicate whether or not the record has been checked in the process of the naming service processing routine 503 searching for the optimum server machine.

FIG. 7 shows the contents of the initialization definition file 52 referred to by the initialization processing routine 501 when the server machine is started. The initialization definition file 52 includes a restart master display flag 521 indicating whether the naming service object is a restart master for generating the load information notification message M1, and a restart master NSO identifier indicating a restart master in the own group. 522, a backup restart master display flag 523 indicating whether the naming service object is a backup restart master for generating the load information notification message M1 when a restart master failure occurs, and a transfer destination of the load information notification message M1. Next NS indicating the next naming service object
O identifier 524 and a group contact master display flag 5 indicating whether the naming service object is a group contact master for generating an inter-group contact message.
25, a group identifier 526 indicating a group to which the naming service object belongs, and a next destination to which the inter-group communication message M2 is transferred when the naming service object is designated as a restart master or a standby restart master. Inter-group communication message destination N indicating restart master of group
The SO identifier 527 and the source NSO identifier 5 of the inter-group communication message indicating the restart master adjacent upstream on the circulation route of the inter-group communication message M2
28.

FIG. 8 shows the contents of the re-entry request buffer 53. When the server machine in the stopped state is started or recovers from the failure state, a re-entry request message M3 is issued from the naming service object on the server machine to the next running naming service object. Is done. Upon receiving the re-entry request message M3, the naming service object receives the next received load information notification message.
The other naming service object is notified of the fact that the server machine of the re-entry request source has become active and the load status of the request source server machine. The re-entry request buffer 53 stores the contents of the re-entry request message until the next load information notification message is generated or received. It comprises a plurality of record areas including an identifier 531 and load information 532. Each record of the re-entry request buffer 53 is
It is cleared when the content is reflected in the load information notification message.

FIG. 9 shows the structure of the server object (SO) position information table 54. The SO position information table 54 is for storing definition information of the server objects 40-1 to 40-n operating on each server machine, and for each NSO identifier 541, a server object name 542 and a server object name. Location information 543
And a plurality of records including an object reference 544.

Each naming service object 50
When the naming service request is received from the client, the naming service processing routine 503 selects the NS selected from the load information table 51 in ascending order of load.
The SO position information table 5 based on the O identifier 511
4, the service (server object name) specified by the naming service request is the NSO identifier 51
It is determined whether or not it can be executed by the server machine having No. 1. The NSO identifier 51 is displayed on the SO position information table 54.
If the designated server object name is not registered in the record area corresponding to No. 1, the naming service processing routine 503 reads the next NS from the load information table 51.
The O identifier is selected, a similar search is performed in the SO location information table 54, and if a record having the specified server object name 542 is found, the location information 543 and the object reference 544 defined in the record are transmitted to the requesting client. To answer.

FIG. 10 shows the format of a naming service request message NRQ transmitted from the client 1 to the naming service object 50. The naming service request message NRQ includes a message identifier 631 and location information 637 of the transmission source client.
And the server object name 63 to be named
8 and a footer 636, like the load information notification message shown in FIG.
The packet is transmitted to the communication link 3 in the form of an IP packet to which the CP header 62 has been added.

FIG. 11 shows an inter-group communication message M2.
Shows the format. Group communication message M2
In FIG. 2, the additional information data blocks are arranged for each group. In addition to the items 631 to 636 included in the additional information notification message M1 shown in FIG. 2, the number of groups 639 and the group identifiers 640 (640-1, 640-2) ,…)including. After the group number 639, the group identifier 6
The additional information data blocks 635-1 to 635 -m for each group led by 40 and the NSO number 634 are arranged.

FIG. 12 shows a re-entry request message M3.
Shows the format. Re-entry request message M3
Is a message identifier 631, a source NSO identifier 634 indicating the naming service object that issued the message, and additional information (CPU utilization) of the server machine on which the naming service object is running.
635b and a footer 636.

FIG. 13 shows an operation availability confirmation message M4.
And the inter-group communication message address change message M5
Shows the format. Operation confirmation message M4
Is a message for confirming whether or not another naming service object is operating normally, and is issued, for example, when the standby restart master confirms the state of the restart master. In addition, when a failure occurs in the server machine of the inter-group communication master, the spare restart master to be operated instead of the inter-group communication message destination change message M5 indicates the inter-group communication message transmission source NSO of the initialization definition file.
This is a message transmitted to the naming service object specified by the identifier 528. These messages include a message identifier 631, a source NSO identifier 634, and a footer 636.

FIG. 14 shows an initialization processing routine 5
01 shows a flowchart. In the initialization processing routine 501, first, the initialization definition table 5
It is determined whether the own NSO is designated as the restart master with reference to the restart master display flag 521 of Step 2 (Step 100). If it is not designated as the restart master, a re-entry request message M3 is generated, transmitted to the naming service object indicated by the next NSO identifier 524 (step 101), and the distribution processing routine 502 is executed.

When the own NSO is designated as the restart master, the load information 512, the service allocation count 513, and the flag 514 are cleared in each record of the load information table 51, and then the load information of the server machine (CPU utilization rate) is cleared. ) And load information table 5
1 in the record corresponding to the own NSO identifier (1
02). Next, a new message sequence number (MSN) 632 including the own NSO identifier and the load information is provided.
And generates a load information notification message M1 having
It is transmitted to the next naming service object indicated by the O identifier 524 (103). Waiting for a response message from the next NSO, if there is no response after a certain period of time, it is determined whether the number of transmission retries of the load information notification message M1 has reached a specified number (106), and the specified number has been reached. If not, the process returns to step 104 to repeat the transmission operation of the load information notification message M1. If there is a response from the next NSO, or if the number of transmission retries has reached the specified number, the inter-group communication master display flag 525 in the initialization definition table 52 is checked (1).
10) If the own NSO is not designated as the inter-group communication master, this initialization processing routine is terminated, and the distribution processing routine 502 is executed.

When the own NSO is designated as the inter-group communication master, a new message sequence number (MSN) 632 including the own NSO identifier and the load information is provided.
Is generated and transmitted to the naming service object indicated by the inter-group communication message destination NSO identifier 527 (111).
As in the case of the load information notification message M1, the destination NSO
(112), and if there is no response after a certain period of time, the transmission operation of the inter-group communication message M2 is repeated until the number of retries reaches the specified number (113, 114). If a response is received from the destination NSO, or if the number of transmission retries has reached the specified number, the initialization processing routine is terminated and the distribution processing routine 502 is executed.

FIG. 15 shows a flowchart of the distribution processing routine 502. In the distribution processing routine 502, the time monitoring processing routine 506 shown in FIG. 16 is started (step 120), and the monitoring timer is started. Thereafter, an input event is waited for (121), and when the naming service request message NRQ is received (1)
22) Naming service processing routine 503, when a time monitoring timer interrupt occurs (123) Time activation processing routine 504, and when load information notification message M1 is received (124), load information communication processing routine 50
5. When the inter-group communication message M2 is received (1
25) executes the inter-group communication processing routine 507 and, if the re-entry request message M3 is received (126), executes the re-entry request processing routine 508, and waits for the next input event.

When the inter-group contact address change message M5 is received (127), the initialization definition file 52 is updated according to the contents of the received message (13).
1) When the operation availability confirmation message M4 is received,
A reception response message is returned (132). If a command for ending the naming service is input (129), this routine ends.

FIG. 16 shows a flowchart of the time monitoring processing routine 506. When the time monitoring processing routine 506 is started, a monitoring timer is started (step 14).
0), and waits for generation of a time monitoring reset request (14).
1). If a reset request has occurred, the monitoring timer is reset (142), and this routine ends. If the monitoring timer times out without a time monitoring reset request (143), a timer interrupt for executing the time activation processing routine 504 is generated (144), and this routine is terminated.

FIG. 17 shows a load information communication processing routine 50.
5 shows the flowchart of FIG. The load information communication processing routine 505 compares the message sequence number (MSN) 631 of the received load information communication message M1 with the MSN of the previously received message (step 150). If the received message is not a new message, the received message is Ignore and end this routine.

When the received message is a new message, after giving a reset request to the time monitoring processing routine 506 (151), the parameter i for counting data blocks is set to an initial value "1" (152). next,
It is determined whether or not the NSO identifier included in the i-th data block 635-i in the load information notification message has already been registered in the load information table 51 (153). Data block 63
A new record including the NSO identifier of 5-i and load information is added (154), and if registered, the data block 635-i indicates the value of the load information 512 of the record in the load information table. Replace with the new load information (155). Thereafter, the value of the parameter i is incremented (156), and the parameter i is set to the NSO number 631.
(157). If parameter i is NS
If the number does not exceed O, the process returns to step 153, and the above processing is repeated for the next data block in the load information notification message.

When the parameter i exceeds the number of NSOs,
It is checked whether or not the re-entry request data is stored in the re-entry buffer 53 (158). If the re-entry request data has been stored, a data block including the NSO identifier 531 read from the re-entry buffer 53 and the load information 532 is set in the load information notification message M1 (159). At this time, if the data block 635-j corresponding to the NSO identifier 531 already exists in the load information notification message M1, the content of the load information 635-j of the data block is updated and the data block corresponding to the NSO identifier 531 is updated. If there is no NSO, the re-entry requested NSO is placed before the data block having the own NSO identifier in the load information notification message M1.
Add a data block with an identifier. The re-entry request data that has been set in the load information notification message M1 is deleted from the re-entry request buffer (16).
0).

Next, the load information of the own server machine is acquired and set in the record corresponding to the own NSO identifier of the load information table 51 and the corresponding data block of the load information notification message (161). The restart master display flag 521 is checked (162). If the own NSO is designated as the restart master, the message sequence number (MS
N) is updated (163), and if the own NSO is not designated as the restart master, the load information notification message M1 is transmitted to the next NSO without changing the MSN at the time of reception.
Transfer to At this time, the NSO located next to the own NSO identifier in the data block array of the message M1
An identifier is selected as a destination NSO, an IP header addressed to the NSO is generated, and a load information notification message M1 is transmitted (1
64, 165, 166), and waits for a response from the destination NSO (167). If there is no response within a certain time, the destination NSO
Is determined to have failed. In this case, a failure code is set in the load information field corresponding to the destination NSO in the load information notification message M1 and the load information table 51 (168), and the process returns to step 164.

An NSO identifier for which a failure code has already been set in the load information field is excluded from the destination NSO, and the next NSO identifier is selected from the data block array of the message M1 as the destination NSO. The transmission operation of the message M1 is repeated. Thus, the collection and notification of the load information of the active server can be repeated while dynamically changing the circulation path of the load information notification message M1 according to the occurrence of a failure in the server machine. If there is a response from the destination NSO within a certain period of time, or if there is no NSO identifier to be selected in the data block array of the message M1, the time monitoring processing routine 506 is started (169), and this routine ends. I do.

In the above-described flowchart, the naming service object designated as the restart master immediately transfers the received message to the next naming service object, similarly to other general naming service objects. If it is desired to periodically circulate the load information notification message M1 at regular time intervals, the naming service object designated as the restart master waits for the next transmission timing, so that the naming service object waits for the next transmission timing instead of step 163. The monitoring processing routine 506 may be started (169), and this routine may be terminated.

FIG. 18 shows a flowchart of the naming service processing routine 503. In the naming service process, the load information table 51 is referred to, and the flag 51
4 is in the “0” state, and the record (the NSO identifier 511) having the minimum load information 512 is selected (step 170). Records in which a failure code is set in the load information 512 are not selected, and the load information table 51
If there is no NSO identifier 511 corresponding to the above selection condition (171), an error message is transmitted to the client (172), and this routine ends.

When a record meeting the selection condition is found, the flag 514 is set to "1" (173), and the number of service assignments 513 is compared with a predetermined reference value (174). If the number of service assignments exceeds the reference value, it is determined that the NSO identifier of the record is busy, and the process returns to step 170 to return to the next record (N
(SO identifier). If the service allocation count is equal to or less than the reference value, the SO location information table 54 is referred to,
It is checked whether a record having the service object name (SO name) for which the naming service is requested is registered in correspondence with the NSO identifier 511 selected in the load information table 51 (175). If there is no corresponding record in the SO position information table 54, step 17
Returning to 0, the next record (NSO identifier) is selected.

When the record of the requested SO name is registered in the SO position information table 54, the position information 543 and the object reference 544 defined by the record are answered to the client by a response message NRY (177), and the load information table The value of the service allocation count 513 of the selected record 51 is incremented (178). Thereafter, all the flags 514 of the load information table 51 are cleared (179) in preparation for the next search processing, and this routine is terminated.

FIG. 19 shows a flowchart of the time start processing routine 504. Time start processing routine 504
Is executed when the load information notification message cannot be received within a predetermined time after the previous transmission of the load information notification message M1. Further, when a method of periodically circulating the load information notification message is adopted, the time activation processing routine 504 is executed when the next transmission timing of the load information notification message M1 is reached for the restart master.

In the time monitoring processing routine, first, load information of the server machine is acquired, and the load information 512 in the record having the own NSO identifier of the load information table 51 is obtained.
Is updated (step 180). Next, by referring to the restart master display flag in the initialization definition file 52, it is determined whether or not the own NSO is designated as the restart master (181). If the own NSO has been designated as the restart master, in step 187, the value of the message sequence number (MSN) 632 of the load information notification message previously received and the load information 635 of the data block corresponding to the own NSO identifier. Are updated, thereby generating a new load information notification message (187).

If the own NSO has not been designated as the restart master, it is determined whether or not the own NSO has been designated as the backup restart master (182). After transmitting the re-entry request message M3 destined for the naming service object of the next NSO identifier (183), the time monitoring processing routine 506 is activated (192), and this routine ends. If the own NSO is designated as the backup restart master, the restart master NSO identifier 5
The status of the restart master is confirmed by the operation availability confirmation message M4 addressed to the naming service object indicated by 22 (184). If there is a response from the restart master, that is, if the restart master is operating normally, the operation from step 183 onward is executed as in other naming service objects.

If there is no response from the restart master, the standby restart master sets the restart master display flag 521 of the initialization definition file 52 to the ON state (“1”), and sets the NSO identifier for the inter-group communication message to the NSO identifier. Inter-group contact address change message M addressed to the naming service object indicated by 528
5 is transmitted (186). Thereafter, the standby restart master performs the same processing operation as the restart master described above. That is, a new load information notification message is generated in step 187, and the load information notification message is corrected according to the contents of the re-entry request buffer 53 (18).
8), and sends the load information notification message to the naming service object indicated by the next NSO identifier (18).
9).

The restart master further checks the inter-group communication master display flag 525 (190), and if not specified as the inter-group communication master, starts the time monitoring processing routine (192) and ends this routine. I do. If it is designated as the inter-group communication master, the inter-group communication message destination NSO identifier 52
A new inter-group communication message M2 is transmitted to the naming service object of the other group indicated by No. 7 (191), and the time monitoring processing routine is started (19).
2). The inter-group communication message M2 updates the message sequence number (MSN) 632 of the previously received inter-group communication message M2, and stores the load information 635b of the data block 635 (635-1, 635-2,...) Of the own group. This is replaced with new load information indicated by the load information table 51.

FIG. 20 shows a flowchart 508 of the re-entry request processing routine. In the re-entry request processing, the source NSO identifier 634 and the load information 635b notified by the re-entry request message M3 are added to the re-entry request buffer 53 (step 200), and these information are also reflected in the load information table 51. Yes (20
1). If the record corresponding to the transmission source NSO identifier 634 already exists in the load information table, the load information is rewritten. If the record corresponding to the source NSO identifier 634 does not exist in the load information table, the record is immediately preceding the record having the own NSO identifier. Add a new record to.

Next, the re-entry request message M3
The transmission source NSO identifier 634 of the restart master NSO identifier 5 defined in the initialization definition file 52
Compare with 222 (202), if the transmission source of the request message M3 is not the restart master, this routine ends. If the transmission source of the request message M3 is the restart master, the restart master display flag 521 of the initialization definition file 52 is checked (203). If the flag is on, the flag is returned to the off state ( 204), this routine ends. The destination change notification to the source NSO of the inter-group communication message may be performed by the preliminary restart master in step 204, but is performed by the restart master when the reception response to the re-entry request message M3 is received. Is also good.

FIG. 21 shows an inter-group communication processing routine 5.
07 is a flowchart. In the inter-group communication process, the source N of the received inter-group communication message M2
The SO identifier 632 is compared with the NSO identifier 528 of the source of the inter-group communication message defined in the initialization definition file 52 (step 210). If they do not match, the NSO identifier 528 of the initialization definition file is used.
Is replaced with the source NSO identifier 632 of the received message (211). Next, the value of a parameter j for counting the number of groups is set to an initial value (= 1) (2
12) Compare the j-th group identifier 640-j of the received message M2 with the own group identifier 526 defined in the initialization definition file 52 (21)
3). If the group identifier 640-j is the own group identifier, after updating each load information of the j-th group of the received message with the latest load information of the own group in the load information table 51 (214), the value of the parameter j is updated. Is incremented (218).

If the group identifier 640-j is not the own group identifier, it is determined whether the group identifier 640-j has been registered in the load information table 51 (21).
5). If the group identifier 640-j has been registered, the load information 512 of the record corresponding to the j-th group in the load information table 51 is replaced with the load information indicated by the received message (216), and the group identifier 640-j is set. Is not registered, the content of each data block of the j-th group of the received message M2 is added to the load information table 51 (217), and then the value of the parameter j is incremented (218). The value of the incremented parameter j is compared with the number of groups 639 (21
9) If the value of the parameter j does not exceed the number of groups, the process returns to step 213 to repeat the same operation for the next group. When the value of the parameter j exceeds the number of groups, that is, when all the groups of the received message have been reflected in the load information table, the local NSO is referred to by referring to the inter-group communication master display flag 525 in the initialization definition file. Is determined as an inter-group communication master (220).

If it is not the inter-group communication master, the received message M2 is transferred to the naming service object of another group indicated by the inter-group communication message destination NSO identifier 527 (222). If it is designated as the inter-group communication master, the message sequence number 632 is updated (221), and then the received message M2 is transferred (222). If there is no response within a predetermined time (223), the transfer operation of the received message M2 is repeated, and if there is a response from the destination naming service object, this routine ends.

If there is no response within a predetermined time, the message transmission operation is repeated by changing the destination NSO according to the data block arrangement in the received message M2, as in steps 164 to 168 of FIG. Is also good. In this case, according to the inter-group communication message M2 shown in FIG. 11, since the NSO identifier of the restart master is set in the load information block 635-1 at the head of each group, the inter-group communication message destination NSO identifier 527 The NSO identifier of the restart master is sequentially extracted from the group next to the group including the in the first block, and the transmission of the message M2 is repeated. In the data block arrangement in the message M2, if the own NSO identifier belongs to a group other than the first group,
Returning from the last group to the first group, the extraction of the restart master NSO identifier is repeated, and the retry operation ends when the self-group is reached. If any restart master fails in a client / server system in a cluster configuration, load distribution is temporarily disrupted.
If the standby restart master operates, normal load distribution can be restored.

According to the client server system of the present invention described above, a plurality of server machines have a naming service function and can update their load information tables periodically. By making a naming service request to any of the plurality of server machines, the load of the naming service itself can be distributed, and the dynamic load distribution of the server machines can be realized. Further, as shown in the embodiment, as one method of exchanging load information of a server machine among a plurality of objects having a naming service function,
When a method of sequentially circulating load information notification messages to a plurality of naming service objects is employed, there is an advantage that load information required for server load distribution control can be efficiently collected and distributed with a relatively small communication amount. Further, since the load information notification message can transfer not only load information but also server failure information at the same time, a naming service excluding the failure server can be easily realized.

When a plurality of naming service objects execute a naming service in parallel as in the present invention, new service allocation may temporarily concentrate on a server machine with the lightest load. As shown in the embodiment, the number of times of service allocation is counted for each server, and when the count value reaches a predetermined threshold, it is excluded from allocation targets, and the count value is cleared each time load information is updated. By doing so, a naming service that avoids load concentration on a specific server can be provided.

In the embodiment, for convenience of explanation, the load information notification message M1 transferred within the cluster and the inter-group notification message M2 transferred between the clusters are shown in different formats.
The format 1 is effective for a relatively small client-server system that does not need to adopt the cluster configuration. In the client server system having the cluster configuration as shown in the embodiment, the same message format as the inter-group communication message M2 can be applied to both the load information communication in each cluster and the load information communication between clusters.

In the embodiment, a logical loop is formed between a plurality of server machines, and the load information communication message M1 is continuously circulated along the loop. The contact message is defined as a first message for collecting load information and a second message for distributing load information, and when the first message goes through a loop, the restart master receives the received first message. The naming service object may be notified of the load information of the other server by retransmitting the message as the second message. In this case, the second message does not need to be sequentially transferred along the loop, but may be transmitted as a broadcast message so that all server machines connected to the communication link can receive the second message at the same time. According to the above-described modification, even when the first message is generated intermittently at a predetermined cycle, the collected load information can be immediately notified to all the naming service objects by the second message, so that the communication amount for contacting the load information is reduced. In this way, load information that matches the actual situation can be distributed to all naming service objects, and appropriate load distribution can be realized.

[0067]

As is apparent from the above description, according to the client-server system and the naming service method of the present invention, the naming service can be distributed and executed by equipping the server machine with the naming service function. Therefore, even when the number of client terminals increases, a quick naming response is possible.

[Brief description of the drawings]

FIG. 1 is a diagram showing a configuration of a client server system of the present invention.

FIG. 2 is a diagram showing an example of a format of a load information notification message M1.

FIG. 3 is a diagram for explaining transfer control of a load information notification message M1.

FIG. 4 shows a load information notification message M1 and an inter-group notification message M2 in a server system having a cluster configuration.
FIG. 7 is a diagram for explaining transfer control of FIG.

FIG. 5 is a diagram showing a configuration of a program provided in the server machine of the present invention.

FIG. 6 is a diagram showing a configuration of a load information table 51.

FIG. 7 is a diagram showing a configuration of an initialization definition file 52.

FIG. 8 is a diagram showing a configuration of a re-entry request buffer 53;

FIG. 9 is a diagram showing a configuration of a server object position information table 54.

FIG. 10 is a diagram showing a format of a naming service request message.

FIG. 11 is a diagram showing a format of an inter-group communication message M2.

FIG. 12 is a diagram showing a format of a re-entry request message M3.

FIG. 13 is a diagram showing a format of an operation availability confirmation message M4 and an inter-group contact address change message M5.

FIG. 14 is a flowchart showing an initialization processing routine 501;

FIG. 15 is a flowchart illustrating a distribution processing routine 502;

FIG. 16 is a flowchart showing a time monitoring processing routine 506;

FIG. 17 is a flowchart showing a load information communication processing routine 505;

FIG. 18 is a flowchart showing a naming service processing routine 503.

FIG. 19 is a flowchart showing a time activation processing routine 504.

FIG. 20 is a flowchart showing a re-entry processing routine 508;

FIG. 21 is a flowchart showing an inter-group communication processing routine 507;

[Explanation of symbols]

1: client terminal, 2: server machine, 3: communication link, 4: server program, 40: server object, 5: naming service program, 50: naming server object, 51: load information table, 52: initialization definition table, 53 : Re-entry definition buffer, 54: server object position information table, M1: load information report message, M2: inter-group report message, M3: re-entry request message, M4: operation availability confirmation message, M5: inter-group contact address change message NRQ: Naming service request message, NRY:
Naming message, SRQ: service processing request.

Claims (10)

[Claims] 1. A client server system comprising a plurality of server machines connected to a client terminal via a network, wherein each server machine includes a plurality of types of server objects for providing an information service to the client terminal. A naming service object for designating a server object on which an information service is to be executed in the client terminal, wherein each of the naming service objects has means for collecting load information of another server machine, and the client terminal A dynamic naming service according to the load state of each server machine for the client server system. 2. A means for each naming service object to communicate load information with another server machine, a load information table for storing collected load information of a plurality of server machines, and the load information table. Means for judging the presence or absence of a server object that matches the request from the client terminal in order from the server machine in a low load state with reference to the above, and determining a server object to be assigned to the client terminal. The client server system according to claim 1, wherein 3. The load information table stores load information of each of the server machines and the presence or absence of a failure, and the server object determination unit refers to the load information table to exclude a server machine in a failure state. 3. The client server system according to claim 2, wherein a server object to be assigned to the client terminal is determined. 4. One of the plurality of server machines includes:
A master naming service object for generating a load information communication message, wherein each of the communication means sequentially circulates the load information communication message generated by the master naming service object to the plurality of naming service objects. 4. The client server system according to claim 2, wherein the client server system transfers the load information of each server machine via the load information notification message. 5. The load information generated by the master naming service object, wherein each of the plurality of server machines is divided into a plurality of groups, and each group has the master naming service object. Forward the contact message to the next naming service object in order to cycle through the group,
4. The client server system according to claim 2, wherein load information of each server machine is exchanged via the load information notification message. 6. One of the plurality of master naming service objects has means for generating an inter-group communication message, and each communication means in the master naming service object has the inter-group communication message. 6. The client server system according to claim 5, wherein the client server system sequentially transfers the information to another master naming service object, and communicates load information of the server machine between the groups via the message. 7. A plurality of server machines connected to a client terminal via a network, each server machine comprising: a plurality of types of server objects for providing an information service to the client terminal; A naming service method in a client-server system, comprising: a naming service object for specifying a server object on which an information service is to be executed. In the naming service object, a naming service object designated as a master among the plurality of servers is included in the naming service object. Generating a load information notification message for causing the machine to make a circuit, setting new load information of the own server machine in the load information notification message, and sending the load information notification message to the network; The naming service object serving as the destination obtains the load information of the other server machine from the load information notification message, sets new load information of the own server machine in the load information notification message, and Sending the naming service object to the naming service object of the machine; and determining whether or not the naming service object that has received the naming service request from the client terminal has a server object that matches the request in order from the server machine in a low load state. Allocating a server object to a terminal. 8. The naming service object designated as the master sends a load information notification message that has cycled through the plurality of server machines until the next transmission timing so that the load information notification message circulates in a predetermined cycle. The naming service method according to claim 7, wherein the naming service is retained. 9. The naming service object designated as the master receives load information of another server machine from a load information notification message that has cycled through the plurality of server machines so that the load information notification message circulates continuously. 8. The naming service method according to claim 7, further comprising: acquiring the new load information of the own server machine in the load information notification message, and transmitting the new load information to the naming service object of the next server machine. 10. If each of the naming service objects does not receive a response to the transmitted load information notification message within a predetermined time, the load information notification message indicates that the destination naming service object has failed. The naming service object, which receives the naming service request from the client terminal, is notified by the load information notification message. The naming service method according to any one of claims 7 to 9, wherein the server object allocation operation is performed excluding a server machine provided with the failed naming service object.