JP2006260059A - Server device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To attain load distribution and averaging in server devices in a group, without making it necessary to provide devices for managing the whole or external specific devices or the like, and to make the whole server devices adjust the loads independently corresponding to load fluctuation, and to make these mutual server devices always have uniform loads. <P>SOLUTION: A server group, including a plurality of servers äa, b, c}, has a common group IP address. Each server measures the load of the processing of own server, stores the load information of each server in the server group, and transmits the load information to the server group according to the load fluctuation. If a processing request is received with a group IP address as the destination of transmission from a PC, that the load of own server is detected as the minimum by referring to the load information, each server will accept the processing. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a server apparatus that receives and processes a request from a client such as a PC (personal computer), and more particularly to a load distribution technique in a plurality of server apparatuses.

  Conventionally, in a configuration in which functions such as mail, Web, DB (database), and application execution are set for each server device, a server device corresponding to a desired function is selected from a client such as a PC, and the corresponding server device is set as a destination. A processing request was issued using the communication address.

  Therefore, when the number of PCs increases, the load on a certain server device increases. Therefore, it is necessary to add an external server device for distributing the load each time and change the setting on the connection target PC side.

  Furthermore, in recent years, in order to facilitate the expansion of server devices, a plurality of server devices called “blade servers” are built in the chassis, and the server functions are improved by inserting and removing in units of blade servers. An easy-to-use method has come out. Each blade server has a CPU, a memory, a hard disk drive (HDD), a network control function, and the like.

  However, in the above case as well, as for how to distribute the load among a plurality of blade servers, load distribution is achieved by changing the connection destination setting on the PC side.

  In order to improve the above problem, in Patent Document 1, in the configuration of the proxy server in parallel operation, when the other party's load is notified to each other and a processing request is received from the PC, first, the main server receives the request once, It is disclosed that load distribution is achieved by comparing with the load of the server, thereby processing the processing request by the own server or by requesting the other server.

  Also, in Patent Document 2, a configuration having a plurality of server nodes, one of which first receives a request from a PC, and if it is within a predetermined load, the request is processed by its own server, In the case of exceeding the load, it is disclosed that load information is acquired from another server and the request is transferred to a server having an appropriate load.

Further, in Patent Document 3, in a configuration in which a management server is provided, the management server receives performance information of each blade server, and a QoS attribute such as throughput of each server is lower, so which server has a job A technique for allocating is disclosed.
Japanese Patent Laid-Open No. 2002-271415 “Proxy server system and communication method thereof” Japanese Unexamined Patent Application Publication No. 2004-287889 “Distributed Processing System, Distributed Processing Device, Method, and Computer Program” Japanese Unexamined Patent Application Publication No. 2004-110791 “Dynamic Adaptive Server Provisioning for Blade Architecture”

  However, in any of the above background arts, in order to distribute the load among a plurality of servers, first, a main server device that manages the whole is required, and the main server device primarily receives processing requests and requests other server devices. Had decided whether or not to do. Alternatively, it is necessary to provide an external load distribution device for a plurality of server devices.

  Therefore, since a relatively high load is applied to an apparatus such as the main server apparatus, a machine having higher performance than other server apparatuses is required. In addition, when the main server device goes down, the preparation of an alternative machine becomes separate, requiring time and man-hours for recovery.

  Also, in some cases, until the load on the main server device reaches a certain value, the remaining server devices may be in an idle state, aiming at complete load balancing among servers It was not.

  The present invention has been made in view of the above problems, and has the following objects (1) and (2) regarding load distribution in a plurality of servers. This provides technologies such as a server device, a control method, a program, and an information processing system that improve facility use efficiency and work efficiency.

  (1) A defined group that does not require a device such as the main server device that manages the whole or primarily receives a processing request, and does not require a special device such as the external load balancer. Load balancing and balancing of processing in the server device (referred to as a server group).

  (2) All server devices in the server group adjust the load autonomously in response to load fluctuations accompanying the increase or decrease in the number of server devices or the number of clients. Share the load. In particular, a technique optimal for the blade server system is provided with respect to the increase or decrease of the number of server devices.

  Of the inventions disclosed in the present application, the outline of typical ones will be briefly described as follows. In order to achieve the above object, the server device of the present invention is each server device in a group including a plurality of server devices, and has the following technical means.

  (1) This server device holds group definition information that defines a group (server group) configured to include a plurality of server devices. The group definition information has a unique address for each server device in the group and a common address for all server devices in the group. This server device has means for measuring the processing load on its own server device, and holds load information of each server device including the own server device of the group. The server device performs processing for transmitting and reporting the load information or load variation information of the server device to another server device of the group in accordance with the variation in the processing load of the server device. Conversely, when the server device receives a report of load information from the other server device of the group, the server device holds the other server device of the group, that is, the server device of the transmission source based on the received information. To update the load information.

  (2) Further, the server device is characterized in the following in (1). All the server devices of the group receive a processing request message using the common address as a transmission destination from clients connected for communication. Then, each server device that has received the processing request refers to the load information of each server device of the group that is held, and when the load information of the own server device is minimum, in response to the processing request All processes are accepted. That is, it is determined whether to accept the process according to the load state of each server device. Alternatively, as another determination criterion, the server device accepts the process when the load information of the server device is within a threshold.

  (3) Further, the present server device is characterized in the following in (1). The server device holds information indicating the operation state of the server device in the group and information indicating the operation state of the other server device. The server apparatus updates the information indicating the operation state in response to a change in the operation state such as power-on or start-up or resume operation from a hibernation state, and includes the group definition information. In this case, information indicating that the own server device is newly added or re-added as a load distribution target, that is, a state in which processing can be accepted, or information indicating the operation state is sent to the other server device of the group. Process to send to. Conversely, when the server device receives the information indicating that it is added from the other server device of the group, the operation of the other server device of the group that is held based on the received information. A process of updating the information indicating the state so that the corresponding server device is added as a load distribution target is performed.

  (4) Further, the present server device is characterized in the following in (2). When the server apparatus accepts a process in response to the process request, the server apparatus transmits a process acceptance message using the unique address of the server as a transmission source to the client. When the reception response message is received, the process corresponding to the process reception is continuously executed.

  (5) Further, the present server device is characterized in the following in (2). Each server device that has received the processing request refers to the load information of each server device of the group that is held, and when the load information of the own server device is not the minimum, the time determined according to the rules, for example, During a random time, etc., the process reception message transmitted from the other server device of the group to the client is monitored. If the message is not intercepted, the load status related to the process reception is monitored. Retry the decision including the decision.

  (6) Further, the server device is characterized in the following in (2). When the server apparatus receives the processing request, and there is the same load information of the own server apparatus of the group and the load information of the other server apparatus, the server apparatus determines the server apparatus included in the group definition information. When the value is the smallest, the server accepts the processing by the server device. That is, when there are a plurality of server devices having the same load, one server device is determined according to a predetermined rule and processing is accepted.

  Among the inventions disclosed in the present application, effects obtained by typical ones will be briefly described as follows.

  (1) It is possible to achieve load distribution and balancing in a plurality of server devices, thereby improving facility use efficiency and work efficiency.

  (2) The number of server devices and client devices can be easily increased and decreased, and optimization can be performed independently by the server device in response to load fluctuations associated therewith. In particular, it is possible to provide a technique optimal for the blade server system.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. Note that components having the same function are denoted by the same reference symbols throughout the drawings for describing the embodiment, and the repetitive description thereof will be omitted. 1 to 13 are diagrams for explaining an embodiment of the present invention.

  The server device according to an embodiment of the present invention receives a request from a client in a server group in a configuration in which a server group is defined and set using a plurality of the server devices, and the load state in each server device is Processing is accepted according to the judgment. This adjusts the load between servers. On the server device, processing according to a control method for load distribution in the server group is executed according to a program, hardware logic, or the like.

<Information processing system>
FIG. 1 shows a configuration example of an information processing system including a server 200 that is a server device according to an embodiment of the present invention. The system includes a network 1, a PC 1, a plurality of servers 200 {a, b, c} mounted in a server aggregate housing 2, and a DB server 3.

  The network 4 is the Internet or LAN based on TCP / IP. In the present embodiment, the network 4 is an Ethernet (registered trademark) LAN, and communication processing is performed using a TCP / IP-based protocol. In addition, as an actual connection form, there can be a multi-drop connection, a star connection via a switching hub or a router, etc., but in any case, all messages entering and exiting each server 200 can be intercepted by other servers 200 as well. Is set to

  The PC 1 is a network terminal device, and requests the server 200 to execute processing such as an application through the network 4 as a client device for the server 200. In this configuration, one or more PCs 1 are connected to the network 4 as necessary.

  The server 200 {a, b, c} is a device that receives a processing request from the client PC 1 and performs processing such as an application corresponding thereto. In this example, each server 200 is a blade server system that can be mounted in the server aggregate housing 2.

  The server collective housing 2 is a housing that allows a plurality of servers 200 to be mounted on one housing. In this example, the server collective housing 2 is in the form of a blade chassis corresponding to the blade server type server 200. In addition, the server collective housing 2 can take various forms such as a rack mount cabinet according to the form of the server 200. Each server 200 {a, b, c} can be mounted on the server collective housing 2 by an insertion / extraction operation as required by a maintenance / administrator. The mounting of the server 200 is practically fixed by wiring, soldering, or the like at the time of manufacture, in addition to methods such as insertion / extraction, mounting, and screwing according to the form of the server 200 and the server aggregate housing 2. You can include things.

  The server collective housing 2 has a configuration such as a slot and a connector for inserting and removing each server 200. For example, when the number of servers 200 is increased, the maintenance / administrator inserts the target server 200 into the slot and connects the connectors to each other, thereby connecting and fixing the server 200 in the server collective housing 2. Each server 200 {a, b, c} is freely mounted in the server aggregate housing 2, and all the servers 200 are connected to the same network 4. Any number of servers 200 can be mounted as long as the mounting space in the server collective housing 2 permits. In this example, three servers 200 {a, b, c} are mounted in the server aggregate housing 2 and can communicate with each other on the network 4.

  In this system, the plurality of servers 200 {a, b, c} are defined as the same group and are handled as targets for processing load distribution on the network 4. Necessary setting information including definition information about the group is held in each server 200 of the group. A plurality of servers 200 {a, b, c} that are defined and communicated as the same group are referred to as a server group.

  The DB server 3 is a server that performs DB processing such as DB reference and update in response to an external request. The DB server 3 performs DB processing in response to a request from the server 200 and returns the result to the server 200. The DB server 200 is one typical example in the case where a server that is shared and accessed from a plurality of servers 200 is provided. In addition, a server for various applications such as Web and mail is also possible.

  In this system, a processing request from the PC 1 is received by the server group using a common communication address, and processing is accepted by the server 200 {a, b, c} according to the load state. The server 200 that has received the process issues a request to the DB server 3 when the DB process is necessary for the execution of the process. Then, the server 200 transmits the processing result from the DB server 3 or the result of processing based on the processing result to the PC 1 as a response. The processing request issued from the PC 1 is a request for specific processing in the server 200 or a DB reference request to the DB server 3. In this example, a case where the server 200 further accesses the DB server 3 will be described. Of course, a case where the process is completed between the PC 1 and the server 200 is also possible.

<PC>
FIG. 2 shows a hardware block configuration diagram of the PC 1. The PC 1 includes a CPU 101, a memory 102, a LAN board 103, an HDD 104, an input control board 105, and an output control board 106.

  The CPU 101 is a processor that controls the entire PC 1. The memory 102 temporarily stores programs and data. The HDD 104 is an external storage device that stores programs, tables, DB information, and the like. The CPU 101 executes a program on the memory 102 to realize the function as the PC 1. The input control board 105 is a board that controls input devices such as a keyboard and a mouse. The output control board 106 is a board that controls a display device such as a monitor and a liquid crystal display and other output devices. The LAN board 103 is a board that controls a LAN interface for the network 4. Various commands and data are exchanged on the network 4 by the LAN board 103. The PC 1 recognizes a group IP address 614 described later.

  FIG. 3 shows a software block configuration diagram of the PC 1. The PC 1 includes an application unit 111, an OS 112, a communication control unit 113, an input control unit 114, and an output control unit 115.

  The application unit 111 indicates a program of application software such as a Web browser, mail, and data reference. An OS (Operating System) 112 performs schedule management such as input / output control, event notification, and activation / termination of a program of the application unit 111. The application unit 111 and the OS 112 issue processing requests to the server 200 as necessary.

  The communication control unit 113 corresponds to the processing of the LAN board 103 and includes a driver for controlling the LAN interface, TCP / IP protocol control, and the like. The input control unit 114 includes a driver that performs input control of the input device. The output control unit 115 includes a display driver for controlling the display device, video memory management, and the like.

<Server>
FIG. 4 shows a hardware block configuration diagram of the server 200 {ac}. Each server 200 includes a CPU 201, a memory 202, a LAN control unit 203, an HDD 204, and an SVP control unit 205.

  The CPU 201 is a processor that controls the entire server 200, and controls processing on the own server and processing on the DB server 3. The memory 202 is a memory that temporarily stores programs, data, table information, and the like. The HDD 204 is an external storage device that stores programs, tables, DB information, and the like. In this example, the HDD 204 holds tables (6, 7) described later. The CPU 201 executes a program on the memory 202 to realize the function as the server 200.

  The LAN control unit 203 is a part that controls a LAN interface for the network 4. Various commands and data are exchanged on the network 4 by the LAN control unit 203. In addition, the LAN control unit 203 once captures a message using a related IP address in the same server group as a source and destination into the control unit, analyzes it, and enables interception.

  The SVP control unit 205 is a part that controls the SVP interface, and is connected to each part in the server and an external SVP (service processor). The SVP performs maintenance / management processing for each server 200 through the SVP control unit 205. The SVP processing is, for example, configuration or failure management.

  Since this server 200 is a blade server system, it does not have an input device or an output device individually. Instead, an SVP control unit 205 is provided to enable input / output by an SVP that functions as an external monitoring device or a system console. The maintenance / manager can perform maintenance / management of the server 200 by operating the SVP. Although not shown, the blade chassis as the server collective housing 2 is provided with parts such as the SVP, a power supply unit that supplies power to each server 200, and a fan that blows air.

  FIG. 5 shows a software block configuration diagram of the server 200 {a, b, c}. The server 200 includes an application unit 211, an OS 212, a communication control unit 213, a monitoring control unit 214, a load measurement unit 215, a setting table 6, and an operation table 7.

  The application unit 211 indicates a program of application server software such as Web reference, mail control, and data reference. The OS 212 performs schedule management such as input / output control, event notification, and activation / termination of the program of the application unit 211. The server 200 performs processing corresponding to the processing request from the PC 1 in the OS 212 and the application unit 211. In this example, the application unit 211 of the server 200 performs processing using DB processing in the DB server 3.

  The communication control unit 213 corresponds to the processing of the LAN control unit 203, and includes a driver for controlling the LAN interface, TCP / IP protocol control, and the like. Communication with the PC 1, between the servers 200 in the server group, and with the DB server 3 is possible via the communication control unit 213.

  The monitoring control unit 214 performs monitoring of the operation status of the servers 200 configured in the same server group, monitoring of electronic messages on the LAN interface, and the like. The monitoring control unit 214 manages the setting table 6 and the operation table 7 and refers / updates them as necessary.

  The load measurement unit 215 measures the CPU usage rate in the CPU 201 of the server 200 as load information for grasping the load state periodically or according to load fluctuations such as start / end of processing of the application unit 211. . The load information of the own server 200 measured by the load measuring unit 215 is reflected in the operation table 7 via the monitoring control unit 214. When the load fluctuates to some extent, communication for load information notification and exchange to other servers 200 in the same server group is performed via the communication control unit 213.

  In this embodiment, a setting table 6 and an operation table 7 are included as group definition information. The setting table 6 is information for setting the system, and holds necessary setting information such as addresses and identification names of the servers 200 {a, b, c} of the same server group including itself. The operation table 7 is information for managing the state of the system, and manages and holds the operation and load state of each server 200 {a, b, c} of the same server group including itself as information.

  The configuration of the DB server 3 may be a server device similar to the blade server type server 200 shown in FIGS. 4 and 5 or a stand-alone type server device similar to the PC 1 shown in FIGS.

<Table>
FIG. 6 shows an example of the format of the setting table 6. The setting table 6 has the same contents except for its own host name 620 in all servers 200. In the setting table 6, information in units of servers 200 {a, b, c} is shown in columns (601 to 603), and items of setting information are shown in rows (611 to 615). The contents of these setting tables 6 are set from an external SVP through the SVP control unit 205 and stored in the HDD 204.

  As the unique host name 611, a unique host name is assigned to each server 200. In this example, the unique host names 611 of the servers 200 {a, b, c} are assigned as “APSRVRa”, “APSRVRb”, and “APSRVRc”.

  As the unique IP address 612, a unique IP address is assigned to each server 200 {a, b, c} as a communication address. In this example, the unique IP address 612 of each server 200 {a, b, c} is “202.2000.256.1”, “202.2000.256.2”, “202.2000.256.3”. And assigned.

  The group host name 613 is assigned the same host name to all of the servers 200 in the same server group. In this example, “APSRVR” is assigned to the group host name 613 of the server 200 {a, b, c}.

  As the group IP address 614, the same IP address is assigned as a common communication address to all of the servers 200 in the same server group. In general, this address may be defined as the same multicast address used in a video conference or the like. In this example, all servers 200 {a, b, c} are assigned “202.2000.256.0” as the group IP address 614.

  The MAC address 615 is a unique MAC address in each server 200. In this example, the MAC addresses 615 of the servers 200 {a, b, c} are “01.23.45.67.89.AB”, “AB.CD.EF.01.23.45”, “ 67.89.AB.CD.EF.01 ". The LAN control unit 203 has a MAC address 615. When the power is turned on, only the MAC address of the local server can be acquired from the LAN control unit 203 and set in this table.

  The unique IP address 612, the group IP address 614, and the MAC address 615 are used as addresses of a transmission source and a transmission destination in communication on the network 4.

  In addition, the host name of the server 200 is set in the host name 620. For example, “APSRVRa” is set in the own host name 620 of the server a (200). By linking the own host name 620 with the unique host name 611, it is possible to know which of the columns (601 to 603) is the content of the own server 200.

  In the present embodiment, the group IP address 614 is used as a transmission destination for processing requests from the PC 1 to the server 200. In addition, the unique IP address 612 is used as a transmission source for the response from the server 200 that has received the processing request to the PC 1.

  Although it is not necessary to identify the server 200 that actually receives and executes the process corresponding to the process request from the PC 1, the server 200 can be identified by the unique IP address 612 or the like.

  The group IP address 614 and the unique IP address 612 are also used in communications such as notification and exchange of load information between the servers 200.

  Even if the group IP address 614 and the unique IP address 612 are set in the IP address part in communication, the MAC of the LAN control unit 203 that actually performs transmission / reception processing at the data link level of the Ethernet (registered trademark) as the network 4 Address 615 is set. However, hereinafter, the technology for converting the IP address to the MAC address 615 is performed by using the ARP (Address Resolution Protocol) or the like and the server corresponding to the condition returns a response to convert it to the MAC address 615. . Hereinafter, although the detailed process of this process is omitted, the operation performed with the IP address can be similarly applied to the MAC address 615 as well.

  FIG. 7 shows an embodiment of the format of the operation table 7. Columns (701 to 703) of the operation table 7 indicate information in units of servers 200 {a, b, c}, and rows (711, 712) indicate items such as the operation and load status of each server 200. In this example, the state of each server 200 is managed by the operation state 711 and the CPU usage rate 712. Based on these values, it is determined that the server 200 accepts processing, that is, the load between the servers 200 is adjusted.

  The operation state 711 is an item indicating an operation state related to processing in each server 200, and is set according to a change in state. If the operation state 711 is “ON”, it indicates that the server 200 is powered on and is in an operating state, that is, a state in which a process is in operation or a process can be provided. Further, if the operation state 711 is “OFF”, it indicates that the server 200 is in a power-off state or a hibernation state (a state in which processing cannot be provided).

  The CPU usage rate 712 is a value representing a load state related to processing in the server 200, and is indicated by a numerical value (unit: [%]) of the usage rate of the CPU 201.

  Further, the information in the operation table 7 is also linked to the own host name 620 to be arranged in the same column as in FIG. 6, so that which of the above columns (701 to 703) is the content of the own server 200. I can know.

  The contents of these operation tables 7 are used to maintain the consistency of the information held in each server 200 by communicating with each other between the plurality of servers 200 by the method described later each time the state of each server 200 changes. ing.

<Telegram>
FIG. 8 shows an embodiment of a message format used on the LAN interface in the network 4. Various types of information set in the message corresponding to the type of message are shown. In this example, commands 811 to 818 are used as the types of messages exchanged between the PC 1, the server group server 200, and the DB server 3. Each command 811 to 818 has, for example, a format having a header and content, and includes cases such as status and simple data. Each message includes areas of a command 801, a source IP address 801, a destination IP address 802, a session number 804, and a content 805.

  The command 801 includes activation completion 811, load information 812, processing request 813, processing reception 814, processing result 815, DB access request 816, DB access result 817, and reception response 818. Each command 811 to 818 indicates the type of message, and what is included as the content 805 is determined.

  For the source IP address 802 and the destination IP address 803, a source address and a destination address in the TCP / IP protocol are designated, respectively. The session number 804 is set with identification information for preventing interference with other similar requests and processes.

  The start completion 811 notifies the other servers 200 in the same server group that the own server 200 is powered on and is in an operating state (“ON”). In this example, “202.200.256.i” (i is one of 1, 2, 3) is designated as the unique IP address 612 of each server 200 in the source IP address 802. A group IP address 614 is designated as the destination IP address 803. The content 805 includes “APSRVRj = ON” (j is one of a, b, and c) as information indicating which server 200 is “ON”.

  The load information 812 is for reporting load information of each server 200 including itself to other servers 200 in the server group when the load of each server 200 varies. The load fluctuation includes starting of the server 200 by power-on. The source IP address 802 and the destination IP address 803 are the same as in the case of the start completion 811. In addition, the load information about all the servers 200 in the server group held in the operation table 7 in the own server 200 is reported in the content 805, including confirmation to other servers 200. In this example, the content 805 includes information on the operation state 711, and includes, for example, “APSRVRa = ON, 70%”, “APSRVRb = ON, 80%”, “APSRVRc = OFF”, and the like.

  The processing request 813 is a message representing a request from the PC 1 to the server 200 such as DB reference, URL reference, mail transmission, data reference. The source IP address 802 is the IP address of the PC 1, and the destination IP address 803 is the group IP address 614. In addition, “aa” is assigned to the session number 804 in this example. The content 805 includes processing request details and data such as parameters.

  The process reception 814 is a message representing the reception of the process in the server 200 for the process request 813 from the PC 1. The PC 1 or the like that has received this message can recognize that the server 200 accepts the process. As the source IP address 802, the unique IP address of the server 200 that has received processing corresponding to the processing request 813 in the server group is used. The destination IP address 803 is the IP address of PC1. As the session number 804, the same value (“aa”) as the session number 804 corresponding to the processing request 813 from the PC 1 is given.

  The reception response 818 is a response for confirmation with respect to the processing reception 814. The source IP address 802 is the IP address of the PC 1, and the destination IP address 803 is the unique IP address 612 of the corresponding server 200 that has transmitted the process reception 814. When the PC 1 receives the message of the process reception 814, the PC 1 returns a message of the reception response 818 in which the same session number 804 (“aa”) as the session number 804 assigned thereto is returned to the corresponding server 200.

  In this example, in the normal sequence, an acceptance response 818 is returned from the PC 1 to the process acceptance 814 from the server 200. Based on the process reception 814 and the reception response 818, the reception of the process in the corresponding server 200 is confirmed. Note that the processing reception 814 and the reception response 818 may be omitted, and the processing may be performed without confirmation.

  The DB access request 816 and the DB access result 817 are shown as examples in the case where a process is requested to the external DB server 3 according to the process in the server 200 in the server group.

  The DB access request 816 is a processing request from the server 200 to the DB server 3. In this example, the source IP address 802 is the unique IP address 612 of the corresponding server 200, and the destination IP address 803 is the IP address of the DB server 3. Further, the case where the same value (“aa”) as the processing request 813 or the like is used as the session number 804 is shown. The content 805 is search key information for DB reference, for example.

  The DB access result 817 is a response from the DB server 3 to the server 200. In this example, the source IP address 802 is the IP address of the DB server 3, and the destination IP address 803 is the unique IP address 612 of the server 200. In addition, a case where the same value (“aa”) as the DB access request 816 is used as the session number 804 is shown. In communication between the server 200 and the DB server 3, a session number 804 (for example, “bb”) different from the session number 804 (for example, “aa”) used with the PC 1 may be used. The content 805 is, for example, a DB search result and its data.

  A processing result 815 indicates a response to the PC 1 that is a processing result corresponding to the processing reception 814 in the server 200. This source IP address 802 is the unique IP address of the corresponding server 200 that performed the process, and the destination IP address 803 is the IP address of the PC 1. Further, the same value (“aa”) as the session number 804 of the process reception 814 is set as the session number 804. The content 805 includes the processing result and details of the data.

  The reason why the unique IP address 612 of the server 200 that received the process is used in the process reception 814, the reception response 818, or the like is that a plurality of servers 200 are caused by delays or omissions in the exchange of load information between the servers 200. This is because the server 200 can be identified in response to the case where the process acceptance 814 is returned to the PC 1. In this case, the plurality of servers 200 that accept the process transmit a message of the process acceptance 814 to the PC 1 using the unique IP address. Then, a reception response 818 is transmitted using the unique IP address of the corresponding server 200 to indicate which processing reception 814 message has been received from the PC 1. The server 200 that has received the acceptance response 818 from the PC 1 continues the process because the process accepted by the server 200 has been confirmed from the PC 1 side. The server 200 that has not received the acceptance response 818 stops the continuation of processing.

  In this example, it is possible to know which server 200 has specifically accepted the process from the unique IP address 612 or the like on the messages such as the process reception 814, the process result 815, the DB access request 816, and the DB access result 817. . The server 200 determines whether or not it should be processed according to the load state, and performs it at its own risk.

  Here, instead of using the process reception 814 and the reception response 818, each server 200 may determine and continue to issue a DB access request 816 in this example. At this time, there are chances that a plurality of servers 200 will be in the same condition and issue a request based on the same judgment. However, even in this case, the server 200 that has returned the processing result 815 early will eventually Data will be adopted. The necessity of the process reception 814 and the reception response 818 is merely reducing the frequency at which the same request is issued from the plurality of servers 200 by the other server 200 intercepting. Even when a plurality of servers 200 receive a message including the case where the processing reception 814 and the reception response 818 are missed due to noise or the like, the processing is performed as described above.

<Processing sequence and flow>
Next, a processing sequence and flow in this system will be described. FIG. 9 shows an example of a series of processing and control sequences among the PC 1, the servers 200 {a, b, c} in the same server group, and the DB server 3. Between the PC 1, the server 200, and the DB server 3, communication and processing corresponding to this sequence and flow are performed using the message.

  In the server group, for example, when the server c (200) is in an operating state due to power-on, a message of start completion 811 for notifying a change in the operating state of itself is sent to other servers a and b ( 200) (S1030). The servers a and b (200) that have received the start completion 811 message update the contents of their own operation table 7 based on the information. In the server group, for example, when the server a (200) detects a certain amount of load fluctuation in itself, the server a (200) transmits a message of the load information 812 to the other servers b and c (200) in the server group (S1011). The servers b and c (200) that have received the load information 812 message update the contents of their own operation table 7 based on the information (S1021, S1031). The server (200) in which the load state or the operation state fluctuates updates the contents of its own operation table 7. In the server group, any server 200 communicates with each other by notifying or exchanging information associated with the occurrence of fluctuations at any time within the server group.

  Next, in S1001, a message of the processing request 813 is transmitted from the PC 1 to the server group. This processing request 813 is issued to a server group using the group IP address 614 as a transmission destination.

  On the other hand, in S1300, each server 200 in the server group receives the processing request 813, and determines whether or not the server 200 accepts processing corresponding to the processing request 813 based on the reference to the operation table 7. And so on. In accordance with the processing and control described in FIG. 13, the processing is received by one of the servers 200, and the message of the processing reception 814 is transmitted from the server 200 that has received the processing to the PC 1. In this example, a case is shown in which processing is received by the server b (200) in which the load is minimum in the server group.

  In S1300, since each server 200 makes a determination regarding reception of a process independently, as a result, depending on the situation, a plurality of servers 200 receive a process other than when only one server 200 receives a process. In some cases, a retry is made without accepting any of them. In either case, processing is actually performed with one server 200.

  In step S1300, in step S1300b, the server b (200) determines whether or not to accept processing on its own server based on the comparison of the CPU usage rate 712 for each server 200 {a, b, c} in the operation table 7. When the server decides to accept the process, it sends a message of the process acceptance 814 to the PC 1. The PC 1 that has received the message of the process reception 814 transmits a message of the reception response 818 to the server b (200). The server b (200) confirms the acceptance of the process by receiving the message of the acceptance response 818, and continues the process.

  When the server b (200) in the server group accepts the process, a load change occurs in the server b (200) and is detected. Therefore, similarly to the processing of S1011, S1021, and S1031, in the processing of S1022, S1012 and S1032, the load information 812 is reported from the server b (200) to the other servers a and c (200), and each operation table 7 is stored. Update.

  Next, it is assumed that the server b (200) that has received the process corresponding to the process request 813 needs to refer to the DB server 3. In S1023, the corresponding server b (200) performs DB reference processing. The server b (200) transmits a DB access request 816 message to the DB server 3. In S1041, the DB server 3 performs DB processing in response to the DB access request 816, and transmits a message of the DB access result 817 as a result to the server b (200).

  In step S1023, the corresponding server b (200) transmits the DB access result 817 or the processing result using the DB access result 817 to the PC 1 as a message of the processing result 815. In step S <b> 1002, the PC 1 performs processing for receiving the processing result 815.

  Further, it is assumed that one process corresponding to the process request 813 is completed in the server b (200) as described above. As a result, a load fluctuation occurs and is detected in the server b (200). Therefore, similarly, in the processing of S1024, S1013, and S1033, the load information 812 is reported from the server b (200) to the other servers a and c (200), and each operation table 7 is updated.

  FIG. 10 is a flowchart showing the details of the processing during the power-on operation in the server group server 200 {a, b, c}. In step S <b> 901, the server 200 activated by power-on (for example, server a) changes the operation state 711 in the column corresponding to the server 200 in the operation table 7 from “OFF” to “ON”. In S902, the server 200 sets the CPU usage rate 712 in the same column to “0”.

  Next, in S903, the server 200 activates the load measurement unit 215, the monitoring control unit 214, and the communication control unit 213, thereby entering a state in which the server group is monitored.

  In step S <b> 904, the server 200 transmits a start completion 811 message to the other servers 200 (servers b and c) in the same server group. As a result, the server 200 (server a) is newly added in a state (“ON”) that can provide processing to the server group.

  On the other hand, as shown in S1221 of FIG. 12, when each server 200 (servers b and c) of the server group receives the message of the start completion 811, the server 200 (server a) newly added thereto is thereby received. The contents of the own operation table 7 are updated, and the load information of each server 200 (servers b and c) including itself is notified by a message of the load information 812.

  In step S <b> 905, the server 200 (server b) that transmitted the activation completion 811 confirms reception of a message of the load information 812 from the other server 200. When the message of the load information 812 is received (S905-YES), in S906, the server 200 loads the servers 200 (servers b and c) included in the content 805 in the received message of the load information 812. According to the information, all the columns (702, 703) other than the column (701) of the own server in the own operation table 7 corresponding thereto are updated.

  FIG. 11 is a flowchart showing details of the load information transmission process in the server 200 {a, b, c} of the server group. This process corresponds to the process (S1011 or the like) in FIG. 9 for checking the load variation and transmitting the message of the load information 812 to the other server 200.

  First, in S <b> 1101, each server 200 measures the CPU usage rate 712 as load information of the server 200 by the load measuring unit 215. There are various known techniques for measuring this load. For example, by using an interval timer, the OS 212 is checked at regular intervals to determine whether it is in a wait state (that is, whether it is waiting for processing or being processed). Or by monitoring a specific signal such as a wait signal or a HOLD signal generated from the CPU itself, and calculating the usage rate of the CPU 201.

  In S1102, the server 200 compares the CPU usage rate 712 of the operation table 7 of the server 200 at the time of the last measurement with the value currently measured in S1101. Based on the comparison, in step S1103, the load fluctuation is checked by determining whether the measured value of the CPU usage rate 712 falls within a certain range or fluctuates more than a threshold value. It should be noted that it may be possible to set a threshold value that serves as a reference for the determination.

  In the case of the fluctuation within the threshold (S1103-YES), in S1105, the corresponding server 200 waits for a predetermined time, moves to S1101, and repeats the same processing again. The value of the predetermined time is a random value or a constant value.

  If there is a variation equal to or greater than the threshold (S1103-NO), in S1104, the corresponding server 200 sets the measured value in S1101 in the column corresponding to the server 200 in the item of CPU usage 712 in the operation table 7. And update.

  Next, in S1106, the corresponding server 200 creates a telegram of the load information 812 by detecting the load fluctuation. That is, the corresponding server 200 sets values corresponding to the columns (701 to 703) of all servers in the items of the operation state 711 and the CPU usage rate 712 in its operation table 7 to the message content 805 of the load information 812. . At this time, only the value of the CPU usage rate 712 may be used.

  In step S <b> 1107, the corresponding server 200 transmits the created message of the load information 812 to the other server 200 in the same server group.

  In addition, regarding the information in the operation table 7 of each server 200, there may be a difference between the servers 200 due to a state change, a delay in notification of information, and the like. In determining the information to be reflected in the operation table 7 when the load information 812 is received from the other server 200, for example, the received latest value is always used, or information indicating a heavy load is not limited to the recent value. It may be determined by using the method with priority. Further, for example, as information stored in the operation table 7, time stamp information indicating a load at a certain time may be added, and a new value may be used in comparison with the present time stamp.

  In addition, regarding the exchange of the load information 812 between the servers 200, information about only the own server 200 may be notified from a certain server 200 to the server group. In this example, the load information about all the servers 200 including the own server 200 in the operation table 7 is notified to the other servers 200, thereby reducing the difference in information between the servers 200.

  FIG. 12 is a flowchart showing details of the message reception processing of various commands 801 in the server group server 200 {a, b, c}.

  First, in step S1201, the server 200 that has received the message compares and determines whether or not the destination IP address 803 of the received message is the group IP address 614. If the addresses do not match, the process ends.

  If the addresses match, in step S1202, the server 200 determines whether the message command 801 is load information 812. In the case of the load information 812, in S1211, the corresponding server 200 displays the load information for all servers in the content 805 of the corresponding message in columns (701 to 703) other than its own server in the corresponding row (711, 712) of the operation table 7. Set to.

  In step S <b> 1203, the server 200 determines whether the command 801 is a start completion 811. In the case of activation completion 811, the server 200 sets the information on the operation state of the target server in the content 805 of the corresponding message in the target column of the operation state 711 in its own operation table 7 in S1221. In other words, information (such as “ON”) indicating the operating state of the server 200 of the message transmission source is set.

  In step S1204, the server 200 determines whether the command 801 is a processing request 813. In the case of the processing request 813, in S1300, the reception processing for the message of the processing request 813 as shown in detail in FIG. 13 is performed. Then, the process ends after this acceptance process.

  Next, in step S1205, the corresponding server 200 determines whether the command 801 is the DB access result 817. In the case of the DB access result 817, in S1231, the corresponding server 200 sets the message content 805 of the DB access result 817 to the message content 805 of the process result 815, and in S1232, the message of the process result 815 is set to PC1. Send to. If the received message command 801 does not correspond to the above, nothing is done and the message is discarded and the process ends.

  FIG. 13 is a flowchart showing details of the reception process (S1300) for the process request 813 in the server 200 {a, b, c} of the server group. In this example, control is performed so that one of the servers 200 accepts the process through the processes of S1303 to S1305.

  First, in S <b> 1301, each server 200 that has received the message of the processing request 813 using the group IP address 614 as a transmission destination reads the contents of the operation table 7. In step S <b> 1302, the corresponding server 200 compares the value of the CPU usage rate 712 in the operation table 7 with the value of the own server column and the other server column. Next, in S1303, the corresponding server 200 determines whether or not the CPU usage rate 712 of the row of the server itself is the smallest in the server group.

  In S1303, if the CPU usage rate 712 of the server itself is not the minimum (S1303-NO), the server 200 moves to S1310.

  When the CPU usage rate 712 of the server itself is minimum (S1303-YES), the corresponding server 200 matches the CPU usage rate 712 of the corresponding column of the other server in the operation table 7 in S1304 at this minimum value. Judge whether there is. That is, the case where the load is the same minimum value in the plurality of servers 200 in the server group is detected.

  When there are a plurality of servers 200 having the same minimum value, the server 200 that accepts the process is selected according to a predetermined rule that is set in advance, and it is determined whether or not the server accepts the process. As an example of the rule at this time, in S1305, the corresponding server 200 refers to the information of the MAC address 615 of each server 200 in the setting table 6 to determine whether the value of the MAC address 615 of the own server is the minimum, The server 200 having the smallest value is determined to accept the process.

  Through the above processing, if there is one server 200 with the lowest CPU usage rate 712 (S1304-NO) or a plurality of servers 200 with the lowest CPU usage rate 712, the server 200 with the smallest MAC address 615 (S1305-S1305- YES) will be obliged to accept the process.

  In the case of S1303-NO, or in S1305, when the value of the MAC address 615 of the server 200 is not the minimum value and the condition for accepting the processing according to the rule is not satisfied (S1305-NO), then in S1310 The server 200 waits for a random time, and determines whether or not the other server 200 has accepted the process by intercepting and monitoring the message of the process acceptance 814 in S1311. If the server 200 intercepts the message of the process reception 814 (S1311-YES), the server 200 ends. When the message of the process reception 814 is not intercepted (S1311-NO), it returns to the start again and starts again from S1301.

  In the case of S1311-NO, the process returns to the start again and a retry relating to the determination of accepting the process is performed. This is because all the servers 200 in the server group do not accept processing in consideration of the fact that the same value is not guaranteed at the same time due to delays or omissions in the reporting of load information regarding the state between the servers 200. This is a measure to prevent the case where nobody responds due to deadlock.

  When the condition for responding to the message of the process reception 814 to the PC 1 is prepared with the process of S1303 as the center, the corresponding server 200 prevents the response of the process reception 814 from the plurality of servers 200 from occurring simultaneously. Based on the same idea as the collision prevention in (Registered Trademark), in S1306, weighting is performed for a random time. In step S1307, the server 200 intercepts the message of the process reception 814 from the other server 200 in the server group during the wait period, so that the other server 200 sends a response of the process reception 814 to the PC 1. Make sure there is no. When the server 200 confirms that the other server 200 is responding by intercepting the electronic message (S1307—YES), the server 200 ends without accepting the process.

  When the server 200 confirms that the other server 200 is not responding because it has not intercepted the message (S1307—NO), the server 200 creates a message for the process reception 814 in S1308. Here, the server 200 sets the unique IP address 612 of its own server as the transmission source in the message of the process reception 814, and sets “aa” as the session number 804 for identifying the process. In step S <b> 1309, the corresponding server 200 transmits the created message of the process reception 814 to the PC 1.

  After S1309, the message of the process reception 814 is received by the PC 1, and the message of the reception response 818 is returned to the server 200 that has received the corresponding process. Then, the process is executed by the corresponding server 200.

  In the processing of S1303 and the like, in addition to determining the load state of the server 200 based on the minimum value, it may be determined by selecting a load information that falls within a certain threshold or less. Further, as the rule used in S1305, a value determined for each server 200 can be used other than the MAC address 615. The wait time in S1306 and S1310 may be a fixed value that is determined based on, for example, the MAC address 615 and is different for each server 200. Control is performed so that one server 200 accepts processing in accordance with predetermined determination criteria and rules related to the decision of acceptance of processing in such a server 200. In addition, depending on the determination criteria, the information held by each server 200, and the like, processing may be performed by a plurality of servers 200. In this case, one server 200 can actually receive the response 818 from the PC 1 to one of the servers 200 and retry the judgment regarding the process acceptance by monitoring the waits and messages in S1310 and S1311. Control is performed so that processing is performed.

  As described above, according to the present embodiment, there is no need for a device that manages the whole or that primarily accepts a processing request, and no special device such as an external load balancer is required. Thus, it is possible to achieve load balancing and balancing of processing in the servers 200 in the server group. Further, all the servers 200 in the server group autonomously adjust the load in response to the load fluctuation accompanying the increase / decrease in the number of servers 200 and PCs 1 so that all the servers 200 always have the load as evenly as possible. Can be.

  In addition to the above, in this embodiment, the CPU usage rate 712 is used as the load information of the server 200. In addition, the number of processes accepted from the PC 1 by the server 200, the so-called traffic amount (transfer rate, etc.), and the execution Other parameters such as the number of jobs (units obtained by dividing the processing in the server 200) may be used.

  Further, in this example, the blade server system having the same performance is targeted for the server 200 that is the load distribution target. In addition to this, in another embodiment of the present invention, the system is configured using a plurality of stand-alone server devices and a plurality of servers having different performances instead of the blade server system. When the system is configured by a plurality of stand-alone server devices, the configuration is the same except for the server collective housing 2 in FIG. In the case of a plurality of servers having different performances, for example, a plurality of servers having different processing speeds [Hz] of the CPU 201, when the CPU usage rate 712 or the like is used as load information, less work is required for a low performance server. Balance the load balancing, which imposes a lot of work on high performance servers.

  Further, regarding the identification of the server 200 that has accepted the process, in addition to using the unique IP address 612, a combination of the group IP address 614, the session number 804, and the ID of the server 200 (such as the unique IP address 612) may be used. Good.

  Moreover, although the case where only one server group is set in the present system has been shown, for example, a plurality of server groups are set by combining the servers 200 in the server collective housing 2, and load distribution is similarly performed in each server group It is also possible to do.

  In this example, the connection / wiring format of each server 200 to the network 4 is a multi-drop type, that is, a format in which each server 200 is independently connected to the network 4 by the LAN control unit 203. However, the present invention is not limited to this, and each server 200 may be connected and wired to the network 4 via a relay unit having a data transfer role such as a switch, hub, router, or gateway. That is, each server 200 is connected to the relay unit, and the relay unit is connected to the network 4. For example, a configuration in which such a relay unit is provided in the server collective housing 2 is also possible. Even in this case, when the group IP address 614 is used in a communication message between the servers 200 and between each server 200 and the PC 1, the message in the same server group is transmitted by the relay unit. The settings relating to the transfer are performed so that the data is transferred to 200. As a result, load distribution can be similarly realized.

  As mentioned above, the invention made by the present inventor has been specifically described based on the embodiment. However, the present invention is not limited to the embodiment, and various modifications can be made without departing from the scope of the invention. Needless to say.

  The present invention can be used for a server or an information processing system including the server.

It is a figure which shows the structural example of the information processing system comprised including the server apparatus in one embodiment of this invention. It is a figure which shows the hardware block structure of PC which is a client connected by communication with the server apparatus of one embodiment of this invention. It is a figure which shows the software block structure of PC which is a client connected by communication with the server apparatus of one embodiment of this invention. It is a figure which shows the hardware block structure of the server apparatus in one embodiment of this invention. It is a figure which shows the software block structure of the server apparatus in one embodiment of this invention. It is a figure which shows one Example of the format of the setting table hold | maintained with the server apparatus in one embodiment of this invention. It is a figure which shows one Example of the format of the operation | movement table hold | maintained with the server apparatus in one embodiment of this invention. It is a figure which shows one Example of the format of the message | telegram used on the LAN interface in a network in the information processing system comprised including the server apparatus in one embodiment of this invention. In the information processing system comprised including the server apparatus in one embodiment of this invention, it is a figure which shows the example of a sequence of a series of processes and control between PC, the server of the same server group, and DB server. is there. It is a flowchart which shows the detail of the process at the time of power-on operation | movement in the server apparatus of one embodiment of this invention. It is a flowchart which shows the detail of the transmission process of load information in the server apparatus of one embodiment of this invention. It is a flowchart which shows the detail about the reception process of the message | telegram of various commands in the server apparatus of one embodiment of this invention. It is a flowchart which shows the detail of the process at the time of power-on operation. It is a flowchart which shows the detail of the reception process regarding a process request in the server apparatus of one embodiment of this invention.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 ... PC, 2 ... Server collective housing, 3 ... DB server, 4 ... Network, 6 ... Setting table, 7 ... Operation table, 101, 201 ... CPU, 102, 202 ... Memory, 103 ... LAN board, 104, 204 ... HDD, 105 ... Input control board, 106 ... Output control board, 111, 211 ... Application part, 112,212 ... OS, 113,213 ... Communication control part, 114 ... Input control part, 115 ... Output control part, 200 ... Server 203 ... LAN control unit 205 ... SVP control unit 214 ... Monitoring control unit 215 ... Load measurement unit

Claims (6)

Holds group definition information that defines a group that includes multiple server devices.
The group definition information has a unique address of each server device in the group and a common address in all server devices of the group,
A means for measuring the processing load on the server device;
Holding load information of the own server device and load information of other server devices in the group,
A process of transmitting the load information of the own server device to another server device of the group in accordance with a change in the processing load of the own server device;
When load information is received from another server device of the group, the server device performs processing for updating load information of the other server device of the group that is held based on the received information . The server device according to claim 1,
The processing request message using the common address as the transmission destination is received from the client, and the load information of the own server device is minimized by referring to the load information of the own server device and the load information of the other server device in the group. If so, a server apparatus that accepts processing in response to the processing request. The server device according to claim 1,
Holding information indicating the operating state of the server device in the group and information indicating the operating state of the other server device,
In response to power-on or activation, if the group definition information is included, information indicating that the own server device is added as a load distribution target to the group is transmitted to the other server device of the group Processing to
A process of updating the information indicating the operating state of the other server device of the group that is held based on the received information when the information indicating the addition is received from the other server device of the group; The server apparatus characterized by performing. The server device according to claim 2,
When accepting a process in response to the process request, a process acceptance message using the unique address of the server as a transmission source is transmitted to the client, and a receipt response message is sent from the client to the client. A server device that, when received, continuously executes a process corresponding to the process reception. The server device according to claim 2,
When receiving the processing request message, if the load information of the local server device is not minimum, the processing acceptance message transmitted from the other server device of the group to the client for a predetermined time. A server device characterized in that when the message is intercepted and the electronic message is not intercepted, the determination regarding the reception of the process is retried. The server device according to claim 2,
When the load information of the local server device of the group and the load information of the other server device are the same when the processing request is received, the unique value determined for each server device included in the group definition information is unique. A server apparatus characterized in that the server server apparatus accepts the process based on the result of comparison.

Images