JPH11126195A - Distributed system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To make it possible to constitute a system optimum for the throughput and reliability of each server for performing work processing. SOLUTION: Respective server managing devices 2a-2n perform the operation management/control of respective PC servers 1a-1n, share constitution information expressing the throughput of the latest respective CPU servers 1a-1n, a fault information expressing the reliability and the constitution information of external input/output units 5a-5m by mutually exchanging information through a communication path 3, determine the respective server managing devices 2a-2n to be the suitable connection destination through autonomous judgment while considering the throughput and reliability of the respective PC servers 1a-1n and start the job processing by starting OS 6a-6m while being connected to an external input/output switching mechanism 4.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a distributed system comprising a plurality of processing units having different business processing capabilities and reliability, and more particularly to a system configuration in a distributed system having no computer for integrally controlling the entire system. Regarding control.

[0002]

2. Description of the Related Art In a system configuration control method in a distributed system using a plurality of PC servers or data processing units thereof, a system using a dedicated system control device for integrally managing and controlling the entire distributed system and an operating system ( There is a method of controlling the system configuration by monitoring the operation status between the distributed system components using the OS (OS) or its extended function. Among them, FIG. 14 is a block diagram showing a conventional distributed system for monitoring the operation status between the distributed system components.

[0003] This distributed system is installed in a plurality (n) of PC servers 101a to 101n that execute business processes, and server management devices 102a to 102n that individually manage the operation of the PC servers 101a to 101n; Operating systems (abbreviated as OS) 106a to 106 that operate on 101a to 101n and perform business processes
n, a plurality (m) of external input / output devices 105a-10 for storing data necessary for business processing and communicating between systems
5m, shared disks 108a to 108m for storing shared data in the external input / output devices 105a to 105m, and redundant Local Area N for communication between the external input / output devices 105a to 105m and other external systems.
network (abbreviated as LAN) 107a, 107b, PC servers 101a to 101n, and external input / output device 105
a to 105 m, and the PC servers 101 a to 101
and an external input / output switching mechanism 104 for enabling data sharing among the n.

Next, an operation when the PC server 101c becomes inoperable due to a hardware failure when the system is started by turning on the power will be described.

[0005] Under the control of the server management devices 102a to 102n, each of the PC servers 101a to 101n performs initialization of internal hardware and self-diagnosis, and executes OSs 106a to 106n.
Start up 106n. During this time, the PC server 101c
Is in a stopped state before the startup of the OS 106c can be completed due to a hardware failure. Remaining OS 106
a, 106b, 106d to 106n normally start up and connect to the corresponding external input / output devices 105a to 105m via the external input / output switching mechanism 104;
The operating status is mutually confirmed through the LANs 107a and 107b. As a result, the operating OSs 106a, 106
b, 106d to 106n know that the OS 106c has not been started, and provide a service to the shared disk 108c on the external input / output device 105c to which the OS 106c is to correspond. For example, the PC server 101b sends the external input / output switching mechanism 104 You can take over by accessing via. Hereinafter, the PC servers 101a, 101b, 1
01d to 101n start business processing as a distributed system using the shared disks 108a to 108m and the LANs 107a and 107b.

[0006]

However, since the conventional distributed system is configured as described above, a special OS for server switching or its extended function is required, and an open system using a PC server is required. It was not suitable for building a high-performance, low-cost system using products.

In order to timely adapt to a PC server whose processing capability and reliability are rapidly improving, the PC server is partially upgraded by adding a processor, etc. to partially improve the capacity of the PC server, and the entire system is improved. When trying to improve the business processing capacity, there is a problem that the OS cannot always allocate an appropriate business according to the processing capacity of each PC server.

In recent years, the use of multiprocessors has been advanced in PC servers, and it is possible to perform a degenerate operation on a processor basis. However, when the degenerate operation is performed, a decrease in processing performance and reliability due to the degeneration is appropriately dealt with. Therefore, there has been a problem that the improvement in the availability of the entire system due to the degeneration cannot be fully utilized.

[0009] Further, regarding the alternative operation by another PC server when the PC server is stopped due to a failure, the processing capacity and reliability of each PC server cannot be properly considered in selecting the PC server to be replaced. It was not always possible to achieve a well-balanced system configuration as a whole system.

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems, and has as its object to perform business processing without depending on a special OS control function or a system management function for integrating the entire system. It is an object of the present invention to provide a distributed system that enables an optimum system configuration for each processing capacity and reliability of a server to be performed, and that enables an optimum system configuration using remaining components even when a failure occurs.

[0011]

In order to achieve the above object, a distributed system according to a first aspect of the present invention comprises a plurality of processing units and a plurality of external input / outputs each storing an operating system and a business application. Device, and external input / output switching means for connecting the processing device and the external input / output device, wherein the processing device activates an operating system of the external input / output device via the external input / output switching device. Thus, in a distributed system that executes a predetermined business process, a plurality of processing device management devices mounted on each of the processing devices and performing operation management of the processing device, and perform communication between the plurality of processing device management devices A communication path, provided in the processing device, and configured to communicate with the processing device in accordance with an instruction from the processing device management device. And a device switching means for connecting, disconnecting the input-output device, wherein the processing device management apparatus,
The configuration information and the failure information of the respective processing devices are shared by transmitting and receiving via the communication path, and the connection destination of the external input / output device is determined by autonomous determination based on the shared configuration information and the failure information. The device switching means connects and disconnects the processing device and the external input / output device based on the connection destination of the external input / output device determined by the processing device management device.

In a distributed system according to a second aspect, in the first aspect, each of the processing unit management devices is configured to optimize the external input / output device in terms of processing capability and reliability based on shared configuration information and fault information. Is determined.

[0013] In a distributed system according to a third aspect of the present invention, in the first aspect, each of the processing unit management devices performs an initialization process with reference to the configuration information and the failure information shared when the processing unit is started up. Thereby, the configuration information and the fault information of the processing device are automatically generated, and the automatically generated configuration information and the fault information are reported to the other processing device management devices.

[0014] In the distributed system according to a fourth aspect, in the first aspect, the configuration information of the external input / output device includes information necessary for executing a business application stored in the external input / output device. An index indicating the processing capability and reliability of the processing device is included.

According to a fifth aspect of the present invention, in the distributed system according to the first aspect, the configuration information of the processing device includes information on a connection destination of each of the processing devices.

According to a sixth aspect of the present invention, in the distributed system according to the first aspect, each of the processing device management devices is configured to indicate an effective processing capability of each of the processing devices based on the shared configuration information and fault information. A performance score and a reliability score representing the reliability of each of the processing devices are calculated, and the connection destination of the processing device is calculated based on the calculated effective performance score, the reliability score, and the configuration information of the external input / output device. It determines the external input / output device.

According to a seventh aspect of the present invention, in the distributed system according to the first aspect, each of the processing device management devices determines a connection order of all the processing devices based on an effective performance score and a reliability score. The connection order of the external input / output devices is determined based on the configuration information of the external input / output devices, and the external input / output device having the same connection order as the processing device is determined as the connection destination.

According to an eighth aspect of the present invention, in the distributed system according to the first or third aspect, each of the processing device management devices responds to a notification from the processing device management device mounted on another of the processing devices. This is to update the configuration information or the failure information held by the system.

According to a ninth aspect of the present invention, in the distributed system according to the eighth aspect, each of the processing device management devices connects each of the processing devices and each of the external input / output devices, and as a result, When it is not connected to the external input / output device, it is made to stand by as a substitute processing device, and when a notification of the alternative operation is received from the other processing device that is executing the business process, instead of the processing device that has sent the report. The external input / output device connected to the processing device that has issued the notification is reconnected to the processing device so that the business process that has been executed by the processing device that has issued the notification can be taken over.

[0020] A distributed system according to a tenth aspect of the present invention comprises a ninth aspect.
In the invention of the above, when there are a plurality of the alternative processing devices, each of the processing device management devices performs the job processing on the alternative processing device having the closest processing capability and reliability at the time of the notification of the processing device that sent the notification. Is to be taken over.

[0021] The distributed system according to the eleventh aspect of the present invention includes a ninth aspect.
In the invention, the processing device management device, when there are a plurality of the alternative processing devices, the replacement processing device that is closest to the processing capability and reliability originally possessed by the processing device that sent the notification. To take over the business process.

The distributed system according to the twelfth aspect is provided with
In the invention of the first aspect, the processing device management device may connect the processing device and the external input / output device, and as a result, when the processing device is not connected to any of the external input / output devices, When there are a plurality of the alternative processing devices, the alternative processing device having the highest connection order takes over the business process in place of the other processing device that is performing the business process.

A distributed system according to a thirteenth aspect is provided with a ninth aspect.
In the twelfth to twelfth inventions, the processing device management device mounted on the processing device that has taken over the business process restarts the business process from the takeover destination processing device again after the management target processing device is restored. It is.

The distributed system according to the fourteenth invention has the first
In the invention, the processing device management device, according to the constraints on the connection between the processing device and the external input / output device included in both the configuration information of the processing device and the external input / output device, The external input / output device to which the processing device is connected is determined by autonomous determination for each of the processing device group and the external input / output device group satisfying the constraint condition.

The distributed system according to the fifteenth invention is characterized in that
In the invention, the configuration information of the processing devices includes operation control information for causing the processing devices to be on standby as an alternative processing device when starting up each of the processing devices, and each of the processing device management devices includes When a standby instruction is set in the operation control information, the processing device is made to stand by as an alternative processing device when starting up, regardless of the processing capability and the reliability.

The distributed system according to the sixteenth invention is characterized in that
In the invention of claim 5, a standby instruction is set in advance in the operation control information of the processing device having high processing capability and reliability.

[0027]

Preferred embodiments of the present invention will be described below with reference to the drawings.

Embodiment 1 FIG. 1 is a block diagram showing Embodiment 1 of the distributed system according to the present invention. FIG. 1 shows a plurality (n) of PC servers 1a to 1n,
The external input / output switching mechanism 4 and a plurality (m) of external input / output devices 5a to 5m are shown. External input / output device 5a
5 m store data necessary for business processing such as an OS and a business application. PC server 1a-1n
Is a processing device that executes a predetermined business process by activating the OS of any of the connected external input / output devices 5a to 5m. The external input / output switching mechanism 4 is external input / output switching means for connecting all the PC servers 1a to 1n and the external input / output devices 5a to 5m.

Each of the PC servers 1a to 1n has a server management device 2a to 2n mounted thereon as a processing device management device. Each of the server management devices 2a to 2n includes an installed PC server 1a to 1n (hereinafter, "managed server").
), And mutually exchange information via the communication path 3. In addition, the external input / output devices 5a to 5m store operating systems (OS) 6a to 6m for executing business processes, and have a duplicated Local Area Network.
k (LAN) 7a, 7b, external input / output devices 5a-
Communication is performed for 5 m and with other external systems. Although the external input / output switching mechanism 4 is shown as a single switching device in FIG. 1, in a large-scale distributed system, the input / output switching range is increased, the performance of the data transfer path is improved, and the reliability is improved. In order to cope with this, it may be configured by multi-stage / multi-connection by a plurality of small switching devices.

The PC servers 1a to 1n are connected to the external input / output devices 5a to 5m via the external input / output switching mechanism 4, and the OSs 6a to 5m stored in the external input / output devices 5a to 5m.
6m is activated to perform business processing. Processing data relating to business processing is stored in the external input / output devices 5a to 5
m, and communication with an external system is also performed using the external input / output devices 5a to 5m of the connection destinations.
7b. Note that when the OSs 6a to 6m operating on the PC servers 1a to 1n are started, it is necessary to connect only to the selected external input / output devices 5a to 5m. However, sharing of business process data and expansion of data storage devices are required. For the purpose of, for example, connecting one PC server 1a to 1n to a plurality of external input / output devices 5a to 5m,
A plurality of PC servers 1a to 1n may be connected to a to 5m.

The server management devices 2a to 2n monitor and record the operating conditions of the PC servers 1a to 1n, start and stop power supply of the PC servers 1a to 1n, start initialization and self-diagnosis. Operation control such as connection with the external input / output devices 5a to 5m and start / end of business processing is performed. The server management devices 2a to 2n determine the connection destinations of the PC servers 1a to 1n and the external input / output devices 5a to 5m required for starting the OSs 6a to 6m of the external input / output devices 5a to 5m at the time of system startup. The server management devices 2a to 2n share by transmitting and receiving configuration information based on the specifications, environment, and the like of the PC servers 1a to 1n and fault information based on the operation state via the communication path 3, and always share all the PC servers. The latest configuration information and fault information regarding 1a to 1n are held. The server management devices 2a to 2n also hold configuration information based on the specifications, environment, and the like of the external input / output devices 5a to 5m, so that the PC servers 1a to 1n managed by the server management devices 2a to 2n can connect to the external devices. The input / output devices 5a to 5m are determined by autonomous judgment.

FIG. 2 shows a plurality of PC servers 1a to 1 shown in FIG.
FIG. 3 is a block diagram showing a PC server 1i which is one of n components. In the PC server 1i, a plurality of (n) processors 8a to 8n for performing data processing, a memory 9 formed by a RAM or the like for temporarily storing processing data, and a bus bridge 10 are connected by a processor bus 24. I have. The bus bridge 10 absorbs the difference between the operating speeds of the processor bus 24 operating at a high speed and the operating speed of the input / output bus 23 at a low speed, so that data transfer between any devices connected to the buses 23 and 24 can be performed smoothly. I can do it. The input / output bus 23 is connected to the external input / output devices 5a to 5m and the PC server 1 via the external input / output switching mechanism 4.
i, an external input / output device connecting device 11 for connecting
ROM 12 for storing a program for performing initialization and self-diagnosis of C server 1i, keyboard 14 for directly inputting / outputting between PC server 1i and the outside, mouse 1
6, display 18 and floppy disk drive (F
DD) 20, each of the control devices 13, 15, 17, 19, and P
A control device 21 of a hard disk device (HDD) 22 used for storing a diagnostic program of the C server 1i and temporary data is connected. This input / output bus 23 is
It is a bus based on an industry standard such as a CI bus or an EISA bus, and includes an external input / output device connection device 11 and control devices 13, 1
5, 17, 19, and 21 are configured by a PCI bus-compliant expansion board or the like. The server management device 2i that manages and controls the operation of the PC server 1i has a built-in nonvolatile memory 25 for recording management information of the server management device 2i. Note that the non-volatile memory 25 has a dedicated H for the server management device 2i without being built in the server management device 2i.
An external storage device such as a DD or a memory card may be used.

The external input / output device connection device 11 is device switching means for connecting / disconnecting the PC server 1i to / from the external input / output devices 5a to 5m, and the external input / output device selected by the instruction of the server management device 2i. By connecting the 5a to 5m to the input / output bus 23 via the external input / output switching mechanism 4, the plurality of processors 8a to 8n, the memory 9, and the control devices 13, 15, 17, 19, 21 and the selected external device are connected. Data transfer between the input / output devices 5a to 5m is performed. The external input / output device connection device 11 is connected to the external input / output device 5a.
After the initialization process of the PC server 1i is completed, the server management device 2i determines a connection destination and outputs a connection instruction to the external input / output device connection device 11 after the initialization process of the PC server 1i is completed. The external input / output device connection device 11 maintains the load device offline state, and keeps the plurality of processors 8a to 8n waiting for system startup.

The plurality of processors 8a to 8n execute initialization and self-diagnosis programs of the PC server 1i stored in the ROM 12 when the PC server 1i is started up, and store the memory 9, the external input / output device connection device 11, the keyboard 14,
Mouse 16, display 18, FDD 20, HDD2
2 and their control devices 13, 15, 17, 19, 21 are initialized and diagnosed. After completion of the initialization and the self-diagnosis, the plurality of processors 8a to 8n
OS stored in the external input / output devices 5a to 5m via the
The business process is started by reading 6a to 6m into the memory 9. The operator has a keyboard 14, a mouse 16,
The display 18 is used to give instructions to the OSs 6a to 6m. Further, even if a failure of the processor 8c is found by the self-diagnosis, for example, the plurality of processors 8a to 8n perform the degenerate operation in which the remaining processors 8a, 8b, and 8d to 8n perform data processing while the processor 8c is stopped. Has functions.

The server management device 2i manages and controls the operation of the PC server 1i, and exchanges configuration information and fault information with the external server management devices 2a to 2n using the communication path 3. As a result, all the PC servers 1a to 1n
Is recorded in the nonvolatile memory 25. At the time of initialization and self-diagnosis of the PC server 1i, the server management device 2i automatically disconnects the failed processors 8a to 8n, or stores the specific processors 8a to 8n and memory 9
Degenerate operation such as disconnecting a part of in this case,
The server management device 2i sets the communication path 3
To the other server management devices 2a to 2n via the server. When the initialization process of the PC server 1i is completed, the server management device 2i selects an appropriate external input / output device 5a to 5m based on the configuration information and the failure information,
The PC server 1i and the selected external input / output device 5a are designated by instructing the external input / output device connection device 11 to connect.
To 5 m.

As described above, each of the server management devices 2a to 2n
Has a configuration information management function, a failure information management function, and a configuration control function. Each of the server management devices 2a to 2n performs a configuration information management function, and exchanges configuration information with each other via a communication path 3 by a communication unit (not shown). Share configuration information. Also, each server management device 2a
~ 2n, by exhibiting the fault information management function,
By transmitting and receiving the reliability score obtained based on the failure occurrence histories for all the PC servers 1a to 1n and failure information such as stop due to a failure or degraded operation via the communication path 3 by communication means (not shown), All PC servers 1a
Share the latest fault information about .about.1n. In addition, when a failure has been monitored in the managed server, and the failure has been detected, the failure information is reported to the other server management devices 2a to 2n, whereby the other PC servers 1a to 1n are notified.
Update the failure information of the managed server held by the server. Further, each of the server management devices 2a to 2n exerts a configuration control function, so that the optimum external input / output devices 5a to 5n are determined based on the shared configuration information and fault information.
m is selected by autonomous judgment, the external input / output switching mechanism 4 couples the server to be managed with one of the selected external input / output devices 5a to 5m, and
Activate the OS stored in 5 m to start business processing.

FIG. 3 shows a plurality of external input / output devices 5a of FIG.
It is a block diagram which shows the structure of ~ 5m. The external input / output device 5i includes a plurality of (n) disk devices 27a to 27n and 28a to record and read data by a plurality of (n) disk controllers 26a to 26n and a plurality (n) of disk controllers 26a to 26n. 28n, a plurality (n) of line controllers 29a to 29n for controlling communication with the LAN 7,
PC server 1a through external input / output switching mechanism 4
1n is connected to the server connection device 30. Here, it is assumed that the OS 6i for executing the business process is stored in the disk device 27a.

When the system is started, the server connection device 30 of the external input / output device 5i connects to any one of the PC servers 1a to 1n, and receives an instruction from the connection destination PC server 1i.
The disk device 27 is controlled by the disk controller 26a.
OSi 6i stored in a is read, and the business process is started by PC server 1i. When the business process is started, a plurality of disk devices 27a to 27n and 28a to 28n are used for storing data for business process under the instruction of the PC server 1i of the connection destination, and the PC server 1i of the connection destination is used. Is L using the line controllers 29a to 29n.
The AN 7 exchanges data with the other PC servers 1a to 1n and external systems.

The PC server 1i loads the OSi 6i, and the disk devices 27a to 27n and 28a to 2 in the external input / output device 5i in which the OSi 6i is stored.
8n and the line control devices 29a to 29n are designed so that business processing can be started only. The processing power and reliability required for business processing provided by the OSi6i are equivalent to all PCs.
The configuration information of the external input / output devices 5a to 5m is stored in the non-volatile memory 25 of the server management devices 2a to 2n in the servers 1a to 1n.
External input / output devices 5a-5m to which servers 1a-1n are connected
Is a criterion when selecting. After the business process starts, the disk device 27a in the external input / output devices 5a to 5m is set.
To 27n and 28a to 28n can be shared by the PC servers 1a to 1n under the control of the OSs 6a to 6m.

FIG. 4 is a table showing examples of configuration information and failure information of the PC servers 1a to 1n stored in the nonvolatile memory 25 by the server management devices 2a to 2n. Furthermore, when the initialization processing of the PC servers 1a to 1n is completed, the server management devices 2a to 2n connect the external input / output devices 5a to 5
The server connection order calculated when determining m is also shown. The configuration information includes a server group serving as a constraint when connecting the PC servers 1a to 1n and the external input / output devices 5a to 5m, a processor model serving as a performance index that affects performance, an operating frequency, the number of processors, a cache capacity, The server management devices 2a to 2a to 2c include a basic capacity score calculated by a system designer based on the memory capacity, a processor model, and the like, and a predetermined formula described later based on the configuration information and a degeneration condition described later. 2n includes an effective performance score indicating an effective processing capacity calculated by 2n. The failure information includes the number of degraded processors indicating a degraded state that affects processing capacity, the number of failures as a reliability index,
It includes the operating status indicating the result of the initialization processing of the servers 1a to 1n, and further includes the reliability score calculated by the server management devices 2a to 2n based on these failure information by a predetermined formula described later.

The configuration information and the failure information of each of the PC servers 1a to 1n are determined by the system designer at the time of system construction and set in the server management devices 2a to 2n. However, a server management device 2i manages its own PC server 1i.
The configuration information and the failure information regarding the other PC servers 1a to 1n are stored, but only the server group, the effective performance score, and the reliability score are retained. For example, when the configuration of the PC server 1j is changed or added, the server management device 2j that manages the PC server 1j communicates with all other server management devices 2 via the communication path 3.
The configuration information and the failure information of all the server management devices 2a to 2n are updated by reporting the new configuration information and the failure information to a to 2n. The new configuration information and fault information reported by the server management device 2j include a server group, an effective performance score, and a reliability score. Thus, the configuration information and the failure information unique to each of the PC servers 1a to 1n can be closed and managed in each of the server management devices 2a to 2n, and the formulas of the effective performance score and the reliability score can be calculated. 1n can have its own formula. The details of each piece of information constituting the configuration information and the failure information will be described below.

The server group includes each of the PC servers 1a to 1
n is one of the restrictive conditions when connecting to any of the external input / output devices 5a to 5m. Here, for example, I denotes an x86 architecture which is a microprocessor architecture of Intel Corporation, and M denotes a multiprocessor configuration. Indicates that the configuration is a single processor by S. Based on this information, the x86 architecture multiprocessor system and the x86 architecture single processor system are defined as different server groups, and the PC server 1 that meets the specifications of the OSs 6a to 6m stored in the external input / output devices 5a to 5m.
a to 1n can be selected. That is, the external input / output devices 5a to 5m to which the PC servers 1a to 1n are connected are selected and determined in units of the PC servers 1a to 1n and the external input / output devices 5a to 5m satisfying the constraint condition. Although only the x86 architecture is shown here, by defining a server of another architecture, for example, even if a PC server of an alpha architecture of DEC is mixed with the PC servers 1a to 1n,
It can be connected to the external input / output devices 5a to 5m storing the corresponding OS.

The basic performance score is calculated for each of the PC servers 1a to 1n.
Is a score indicating the inherent processing capability, and is a score indicating the processing capability in a state where the configuration is not degenerated. The basic performance score is based on the processor model, operating frequency,
A system designer is determined at the time of system construction based on performance indices that affect processing capabilities such as the number of processors, cache capacity, and memory capacity. The basic performance score is weighted for each performance index. For example, (processor model score) + (operating frequency × weight F) + (number of processors × weight P) + (cache capacity × weight C) + ( It is calculated by a calculation formula such as (memory capacity × weight M). Therefore, the larger the value of the basic performance score, the higher the processing capability.

The effective performance score is a score of the processing capacity corrected due to a reduction in the processing capacity as a result of the operation in the state in which the configuration has been degenerated. When the initialization processing of the PC servers 1a to 1n is completed, The server management devices 2a to 2n calculate based on the basic performance score, the degeneration mode of the degenerated part, and the weight value. For example, it is calculated by a calculation formula such as basic performance score− (number of degenerate processors × weight P). According to the example of FIG. 4, since the number of degenerated processors of the PC server 1 b is 2, if the weight P is 1, the effective performance score of the PC server 1 b is (8−2 × 1 = )
It becomes 6.

The number of degenerate processors is the number of PC servers 1a to 1
n is a value updated at the time of completion of the initialization processing, and represents the number of degenerated processors 8a to 8n.

The number of failures indicates the number of failures of the PC servers 1a to 1n that occurred during the initialization processing of the PC servers 1a to 1n and business processing.
It is used as one of the indexes indicating the reliability of the servers 1a to 1n. In the present embodiment, the number of failure occurrences since the system was constructed is accumulated. Each time the number of failure occurrences exceeds a predetermined number, it is added to the reliability score so that a decrease in reliability can be represented.

The operating status is updated when the initialization process of the PC servers 1a to 1n is completed and when the operating status changes due to a failure. For example, if the PC server 1i is stopped due to a failure as a result of the update at the time of completion of the initialization processing, the operation status is set to stop. By updating in this manner, the server management devices 2a to 2n use the PC server 1i for control to remove the PC server 1i from the connection destination candidates when determining the connection destination of the external input / output devices 5a to 5m.

The reliability score is calculated by the server management devices 2a to 2n from the number of degraded processors, the number of failures, the operation status, and the like by a predetermined formula when the initialization process of the PC servers 1a to 1n is completed. It is a total score representing the reliability of 1n. In the present embodiment, when the operation status is “operation”, the number of failures is used as it is as the reliability score, and when the operation status is “stop”, 100 is added to the number of failures. Therefore, the higher the reliability score, the lower the reliability. It should be noted that a larger value such as 1000 may be used as the added value when the operation status is stopped.

The server connection order is determined by the PC servers 1a to 1n.
Server management devices 2a to 2n
Is an order calculated based on the effective performance score and the reliability score for each server group. Based on the order, the server management devices 2a to 2n are connected to the PC servers 1a to 1n. Determine 5m. In the present embodiment, the first order is performed based on the effective performance score, and the second order is performed based on the reliability score for those having the same effective performance score. Furthermore, the final ordering is:
The ordering is performed in ascending order of server numbers unique to the PC servers 1a to 1n. In addition, the order 0 is the external input / output devices 5a to 5a.
5m is not connected. According to the example illustrated in FIG. 4, the PC server 1c is configured to use the external input / output devices 5a to 5m.
Is not a candidate for connection. Further, since only the PC server 1d exists in the server group IS, the server connection order of the PC server 1d is 1.

FIG. 5 shows the external input / output device 5a stored in the nonvolatile memory 25 by all the server management devices 2a to 2n.
FIG. 8 is a diagram showing an example of configuration information of up to 5 m in a table format. The configuration information includes an OS name indicating the type of the OSs 6a to 6m stored in the external input / output devices 5a to 5m, a performance index indicating processing capability and reliability required for business processing by the installed OSs 6a to 6m, and Reliability index, connection source PC server 1
A server group as a constraint on a to 1n, and a device connection order, which is determined for each server group based on the performance index and the reliability index, and indicates a connection order with the PC servers 1a to 1n. All the configuration information of the external input / output devices 5a to 5m is determined by the system designer except for the device connection order.
It is recorded in all server management devices 2a to 2n. The device connection order is calculated by the server management devices 2a to 2n based on the performance index and the reliability index when the initialization processing of the PC servers 1a to 1n is completed. The external input / output devices 5a to 5
Details of each piece of information constituting the m configuration information will be described.

The server groups are the PC servers 1a to 1n
Has the same meaning as the server group included in the configuration information, and permits connection only between the same server groups. Here, depending on the type of OS 6a to 6m,
The PC server 1a is divided into groups according to those dedicated to a single processor or those compatible with a multiprocessor.
1 to 1n.

In this example, the performance index and the reliability index are relative indexes representing, as ranks, the processing performance and reliability required for business processing by the OSs 6a to 6m mounted on the external input / output devices 5a to 5m. is there. The primary ordering is performed by the performance index for each server group, and the external input / output devices 5a to 5m having the same performance index are further ordered by the reliability index.
For example, when an external input / output device is newly added, the system designer first assigns one of the existing ranks in terms of performance using the performance index, and changes the ordering within the same performance index using the reliability index. . In this example, based on the performance index and the reliability index, the device connection order that represents the connection order to the PC servers 1a to 1n in a unified manner is stored in the server management device 2a.
To 2n at the time of completion of the initialization processing of the PC servers 1a to 1n. The server management devices 2a to 2n compare the device connection order with the server connection order, and
The connection relationship between the servers 1a to 1n and the external input / output devices 5a to 5m is determined.

FIG. 6 shows the server management devices 2a-2 until the connection destinations of the PC servers 1a-1n and the external input / output devices 5a-5m are determined and the system startup for business processing is started.
4 is a flowchart showing the processing procedure of n. Hereinafter, the operation in the present embodiment will be described with reference to FIGS. Here, the PC servers 1a to 1n, the server management devices 2a to 2n, the external input / output devices 5a to 5m, and the OS 6
PC server 1i, server management device 2
i, the external input / output device 5i and the OS 6i, the operation will be described. When the entire distributed system is powered on, other P
The C servers 1a to 1n and the server management devices 2a to 2n also simultaneously start up the system by the same procedure, and execute the following processing.

First, when the power of the entire distributed system is turned on, the server management apparatus 2i sets the number of degenerated processors to 0 in the failure information of the PC server 1i, and sets the operation status to the initialized state. If there is other degeneration information such as a memory or a cache, they are also set to the values at the time of non-degeneration (step 100).

The server management device 2i starts initialization and self-diagnosis of the PC server 1i (step 110). P
After turning on the power, the processors 8a to 8n of the C server 1i
Until the start of the initialization process is instructed from the server management apparatus 2i, the operation is stopped by, for example, maintaining the reset state of the hardware. When the server management device 2i instructs the start of the initialization process, the processors 8a to 8n
Initialization processing is started by the initialization and self-diagnosis program in the above. Thereby, the memory 9, the external input / output device connection device 11, the control devices 13, 15, 17, 19, 2
1, keyboard 14, mouse 16, display 18,
The FDD 20 and the HDD 22 are initialized and diagnosed, and become operable. The external input / output device connection device 11
Keeps the connection destination offline until the connection destination instruction is issued from the server management device 2i. Since the external input / output device connection device 11 is defined as a load device for the OSs 6a to 6m, the processors 8a to 8n remain in a state of waiting for the system startup by the OSs 6a to 6m after the completion of the initialization process.

Here, when it is confirmed that the PC server 1i has failed in the initialization process and is inoperable, the server management device 2i sends a failure report indicating that the PC server 1i has been stopped to the communication path. 3 to all the other server management devices 2a to 2n (steps 120 and 130). Thereafter, the server management device 2i keeps the PC server 1i in a stopped state and waits for recovery by maintenance personnel (step 14).
0).

On the other hand, upon confirming the completion of the initialization processing of the processors 8a to 8n, the server management device 2i confirms the degeneration status of the PC server 1i (steps 120 and 15).
0). The PC server 1i includes, for example, the processors 8a to 8n
As a result of the self-diagnosis, if a failure is found in the processor 8c, the processor 8c is disconnected to complete the initialization processing. When recognizing the disconnection of the processor 8c, the server management device 2i records the fact (the number of failed processors) in its own failure information (step 160).

The server management device 2i calculates the effective performance score of the PC server 1i based on the updated configuration information and failure information relating to the PC server 1i, and records it in the configuration information of the PC server 1i (step 170). According to the example shown in FIG. 4, since the number of degenerated processors is 2 and the weight is 1 for the PC server 1b, the calculation is performed with the effective performance score as (basic performance score-2). Since the number of degenerated processors is 0 for other PC servers, the basic performance score is directly used as the effective performance score.

The server management device 2i calculates the reliability score of the PC server 1i based on the updated failure information on the PC server 1i, and records it in the failure information of the PC server 1i (step 180). According to the example shown in FIG.
C server 1a, PC server 1b, PC server 1d and P
Since the operating status of the C server 1n is in the operating state, the number of failures is recorded as it is as the reliability score. However, since the operating status of the PC server 1c is in the stopped state, the reliability score is represented by the number of failures. Recorded as +100.

The server management device 2i thus sets P
When the configuration information and the failure information of the C server 1i are automatically generated,
The contents of the nonvolatile memory 25 are updated, and an initialization completion report including the configuration information and the failure information is transmitted to all the other server management apparatuses 2a to 2n via the communication path 3 (step 190). On the other hand, the other server management devices 2a to 2a
2n, receives an initialization completion notification or a failure notification. At this stage, the server management device 2i acquires the configuration information and the failure information regarding all the PC servers 1a to 1n from the other server management devices 2a to 2n via the communication path 3 and sequentially records them in the nonvolatile memory 25. .

The server management device 2i is connected to all PC servers 1a.
To 1n configuration information, fault information, and all external input / output devices 5a
The connection destinations of all the PC servers 1a to 1n and all the external input / output devices 5a to 5m are determined based on the configuration information of the first to fifth meters (step 200). As a result, it is assumed that the connection destination of the PC server 1i is determined to be the external input / output device 5i. The server management device 2i instructs the external input / output device connection device 11 of the PC server 1i to the external input / output device 5i to be connected to the PC server 1i based on the result of the determination, and the PC server 1i and the external input / output device 5i. Are connected (step 210). In addition,
Each of the server management devices 2a to 2n performs the same processing individually, and the PC server 1a to 1n on which each is mounted.
As a result of determining the connection destinations, if the number of PC servers is equal to or less than the number of external input / output devices, the connection destinations of all the PC servers 1a to 1n are basically determined.

The external input / output device connection device 11 is connected to the external input / output device 5i selected by the instruction of the server management device 2i.
Via the external input / output switching mechanism 4 to bring the connection destination online. Thereby, the PC server 1
After the initialization of i is completed, the processors 8a to 8n in the online waiting state of the OS 6i loading device load the OS 6i mounted on the connected external input / output device 5i into the memory 9 and start system startup. (Step 2
20).

FIG. 7 is a flowchart showing a detailed procedure of the initialization completion notification process (step 190) in FIG. By this procedure, the server management device 2i
The configuration information and the failure information of the server 1i are reported to all the other server management apparatuses 2a to 2n, and the configuration information and the failure information of all the other server management apparatuses 2a to 2n are acquired.
Here, the number of all the PC servers 1a to 1n is obtained based on the number of the initialization completion reports and the failure reports received at the predetermined time, and it is possible to flexibly cope with a change in the system configuration before the power is turned on. . Other server management devices 2a to
2n also performs the same processing at the same time, so that all server management apparatuses 2a to 2n share the configuration information and the failure information of all PC servers 1a to 1n.

The server management device 2i sends an initialization completion report including the configuration information and the failure information of the PC server 1i to all the other server management devices 2a to 2n (step 1).
91). The configuration information and the failure information transmitted here are only the minimum information necessary for determining the external input / output devices 5a to 5m, such as the server group, the effective performance score, and the reliability score. The server management device 2i includes all the PC servers 1a
An initial value n = 1 is set in order to obtain the number of １1n (step 192). Note that n = 1 is set in consideration of the own PC server 1i. Next, the server management device 2
i sets the initialization processing completion notification or the failure notification waiting time from the other PC servers 1a to 1n (step 1).
93). This waiting time is determined for all PC servers 1a to 1n.
Is a value sufficient to complete the initialization process. Then, the server management device 2i repeatedly performs the following processing until the waiting time ends.

First, when the received information is the initialization completion report, the received configuration information and fault information are stored in the nonvolatile memory 25.
(Step 196), and 1 is added to the server number n (step 197). On the other hand, if the received information is a failure report, the received failure information is recorded in the non-volatile memory 25 (step 199), and 1 is added to the server number n (step 197). As a result of this processing, the number of PC servers 1a to 1n can be obtained while recording each information in the nonvolatile memory 25.

FIG. 8 is a flowchart showing a detailed procedure of determining the external input / output device connection destination (step 200) in FIG. PC server 1a to 1 in this embodiment
The external input / output devices 5a to 5m to which 1n is connected are all server management devices 2a to 2n, and all the PC servers 1a to
1n configuration information and fault information and all external input / output devices 5
By sharing the configuration information a to 5m, each is determined by autonomous judgment. Hereinafter, the procedure for determining the connection destination will be described as an operation in the PC server 1i with reference to FIG. 4, FIG. 5, and FIG. In other PC servers 1a to 1n,
The connection destination is determined by the same procedure.

At the start of the procedure in FIG. 8, the configuration information and fault information of all the PC servers 1a to 1n and the configuration information of all the external input / output devices 5a to 5m are updated to the state at the time of completion of the initialization processing. Thus, the entire distributed system is in a state where business processing can be started by determining a connection destination based on such information.

The server management device 2i, based on the effective performance score and the reliability score in FIG.
The connection order of n is determined (step 201). First, the server connection order is determined to be higher in the order of higher effective performance score, and for those having the same effective performance score, the server having the lower reliability score is ranked higher. As a result, the PC servers 1 which are still in the same order are shown in FIG.
The server with the smaller server number from a to 1n is ranked higher. FIG.
In the example shown in (1), the PC server 1n having the highest effective performance score for the server group IM is set to the server connection order 1, and the PC servers 1a to 1c having the same effective performance score are ordered in ascending order of reliability score, that is, PC
The server 1a has the server connection order 2 and the PC server 1b has the server connection order 3. The PC server 1c determines that the connection is impossible because the reliability score exceeds 100, and sets the server connection rank to 0.

Subsequently, the server management device 2i determines the device connection order of the external input / output devices 5a to 5m from the configuration information of the external input / output devices 5a to 5m in FIG. 5 (step 202). The device connection order is ranked in the order of the performance indices shown in FIG. 5, and those having the same order are ranked by the reliability index. The performance index and the reliability index both indicate the order set for each server group, and a smaller value indicates a higher order, indicating that a PC server with higher performance and higher reliability is required. In addition,
In the present embodiment, the device connection order is indirectly expressed by the performance index and the reliability index, and the server connection order is calculated again by the server management device 2i every time the system is started up. It may be uniquely determined by the system designer.

The server management device 2i manages the P
The configuration information of the external input / output devices 5a to 5m is acquired based on the server connection order of the C server 1i (step 203).
The external input / output device 5i in the same server group as the PC server 1i and in the same device connection order is searched (step 2).
04, 205). This is repeated until the corresponding external input / output device 5i is found (step 206). If the relevant external input / output device 5i is found, it is set as the connection destination of the PC server 1i (step 207).

On the other hand, if there is no corresponding external input / output device 5i as a result of the above search, it is determined that there is no connection destination, and
The server management device 2i holds the connection waiting state of the PC server 1i to make it stand by as a substitute server (step 208). If a failure occurs in the PC servers 1a to 1n that are executing the business process after the start of the business process, the PC server 1i in the standby state performs the alternative operation.

As described above, according to the present embodiment, P
The connection order is determined for each of the C servers 1a to 1n and the external input / output devices 5a to 5m, and each PC server 1a to 1n is connected to the external input / output devices 5a to 5m in the same server group and having the same device connection order. Since each connection was made,
It is possible to connect the C servers 1a to 1n to appropriate external input / output devices 5a to 5m corresponding to the respective processing capacities to perform business processing. In particular, according to the present embodiment, the PC servers 1a to 1n represented by the basic performance score are not based on the inherent processing capability due to the specifications and the like, but are based on the actual configuration in consideration of the change in the configuration such as the degeneration of the processor. Since the server connection order is obtained using the effective performance score, each of the PC servers 1a to 1 is set in accordance with the processing capacity at the time of system startup.
n can be connected to the optimal external input / output devices 5a to 5m.

Next, the operation when the standby PC server 1i starts the alternative operation will be described with reference to FIG. 9 and FIG. FIG. 9 shows the PC servers 1a to 1n when the system is started up.
Connection status between the external input / output devices 5a to 5m, the connection status after the start of the alternative operation, the connection status after the PC server 1j is restored, and P
It is the figure which showed the connection status at the time of C server 1i restarting in tabular form. Among the configuration information and the failure information managed by the server management devices 2a to 2n, the server group, the server connection order and the operation status are shown, and the external input / output device connection status is also newly shown. The external I / O device connection state is defined by each P
This is information on the connection destinations of the C servers 1a to 1n. When connected to any one of the external input / output devices 5a to 5m, the identification information of the external input / output devices 5a to 5m indicates any of the external input / output devices. If none of 5a to 5m is connected, "None" is set to that effect. All PC server 1a
1 to 1n, each PC server 1
Since information is exchanged between a to 1n and the server connection order and the device connection order are in one-to-one correspondence, the external input / output device connection state can be directly known from the server connection order and the device connection order. However, when performing the alternative operation while keeping the states of the operating PC servers 1a to 1n as they are, in order to partially break the correspondence between the server connection order and the device connection order, the external input / output device connection state is changed to the server connection order. And the device connection order cannot be directly known. That is, when performing the alternative operation, the PC server 1i performing the alternative operation sends information such as a server group to the other PC servers 1a to 1n when the system is started up. This is because the servers 1a to 1n do not transmit information about the connection destination to the PC server 1i, and thus cannot establish a correspondence. FIG. 9A illustrates a transition state of the connection state by directly indicating the external input / output device connection state. In order to control the alternative operation, the server management devices 2a to 2a
n records the external input / output device connection status in the configuration information.

In FIG. 9B, the PC server 1j has failed and stopped, and the standby PC server 1i has connected to the external input / output device 5d to which the PC server 1j has been connected, and has started alternative operation. Indicates the situation. In FIG. 9C, P
Assuming that the reliability of the PC server 1j is improved when the C server 1j is restored, only the server connection order is changed without changing the external input / output device connection state. Also, in FIG. 9D, when the PC server 1i is restarted, the server connection order is reviewed, so that the PC server 1j is again activated.
Shows a state in which a business process is started by connecting to the external input / output device 5a.

FIG. 10 shows the failed PC server 1j.
Management device 2j when the server management device 2j that manages the server notifies the other server management devices 2a to 2n that a failure has occurred.
It is a flowchart which showed operation | movement of a-2n. Less than,
The operation in which the standby PC server 1i starts the alternative operation will be described based on this flowchart.

When the PC server 1j fails and becomes inoperable after the business process starts, the server management device 2j managing the PC server 1j disconnects the connection with the external input / output device 5a, Server management devices 2a to 2n
Is notified via the communication path 3 to the occurrence of a failure. Each of the server management devices 2a to 2n that has received the notification of the occurrence of the failure records the stop as the operating status in the failure information of the PC server 1j and changes the connection destination to none from the external input / output device 5a (step 301).

After changing the failure information and the configuration information of the PC server 1j, each of the server management devices 2a to 2n checks whether or not there is a standby PC server from the operating status of the failure information (step 302). If there is no standby PC server, the process ends except for the server management device 2j. Since only the server management device 2j cannot perform the alternative operation, the maintenance work is required by the maintenance staff, and thus an alarm is issued to stop the PC server 1j (step 303).

If there is a standby PC server, each of the server management apparatuses 2a to 2n determines the PC server 1i having the highest server connection order among the standby PC servers as the substitute PC server (step 304). As a result, the server management devices 2a to 2n change the operating status of the PC server 1i to active and set the external input / output device connection status to the external input / output device 5a.

If the server management devices 2a to 2n are the server management devices 2i that manage the PC server 1i that starts the alternative operation, the server management devices 2a to 2n connect to the external input / output device 5a (steps 305 and 306). ), The system is started, and the alternative operation is started (step 307). The other server management devices 2a to 2n end the processing only by updating the failure information and the configuration information. Thus, all the server management devices 2a to 2n share the same configuration information and fault information.

Note that the failed PC server 1j
Since all the connections with the external input / output devices 5a to 5m are disconnected, the power supply is cut off, the fault location is recovered, and the diagnosis is performed independently while the business processes by the other PC servers 1a to 1n are continued. be able to. The PC server 1j
After the recovery, it can be used as an alternative PC server by keeping it in a standby state. When only one standby PC server exists, the standby PC server will perform the alternative operation unconditionally. In this case, too, the standby PC server is always set as the alternative PC server by following the above procedure. Will be determined.

The alternative operation by the PC server 1i is canceled by restarting the entire system after the PC server 1j is restored. However, even if the entire system is in operation, the alternative operation by the PC server 1i for the convenience of business processing is performed. Can be stopped, the business process is taken over again from the takeover destination PC server 1i at that time, and the PC server 1
It can be restarted with j. In this case, both the PC server 1i and the PC server 1j are temporarily put into a standby state, and the server connection order is restored again so that the PC server 1j is determined as the substitute PC server. As a result, the PC server 1j can be returned to the original business process.

Next, the operation at the time of restoration of the PC server 1j will be described.
This will be described with reference to FIGS. 6, 7, 9 and 11. FIG.
Is a flowchart showing the operation of the PC servers 1a to 1n other than the PC server 1j when the PC server 1j recovers. The other PC servers 1a to 1n include a PC server that is performing a business operation and a standby PC server.

After the recovery work by the maintenance staff, the PC server 1j is powered on and executes the initialization processing according to the procedure shown in FIG. Here, the difference from the power-on state of the entire system is that since the other PC servers 1a to 1n are already executing business processes, the initialization processing completion reports from the other server management devices 2a to 2n are not changed. This is that the external input / output devices 5a to 5m cannot be connected without permission according to the server connection order and the device connection order after the server connection order is determined.

In FIG. 6, the server management device 2j confirms the completion of the initialization process of the PC server 1j by turning on the power, reports the completion of the initialization process in step 190, and sends the information from the other server management devices 2a to 2n. Wait for the initialization completion report. In order to cope with this, the PC server 1a during the business process
When the server management devices 2a to 2n receive the initialization processing completion notification from the server management device 2j,
The server management device 2j
Can share the configuration information and the failure information with the other server management devices 2a to 2n, and allow the PC server 1j to perform the connection processing with the external input / output devices 5a to 5m.

That is, each of the server management devices 2a to 2n
When receiving the initialization completion report from the server management device 2j of the restored PC server 1j (step 311), the P
The configuration information, the failure information, and the external input / output device connection status of the C servers 1a to 1n are transmitted to the server management device 2j as an initialization processing completion notification (step 312). The failure information includes degeneration information and operation status.

Referring to FIG. 7, the P
The server management device 2j of the C server 1j records the configuration information, failure information, and external input / output device connection status of the corresponding PC server 1a to 1n as server management information (step 1).
96). Thereafter, the server management device 2j proceeds to the connection destination determination processing (step 210) of the external input / output devices 5a to 5m in FIG.

In the process of determining the connection destination of the external input / output devices 5a to 5m in step 210 in FIG. 6, the server management device 2j transmits the server connection order to the device connection order from the other server management devices 2a to 2n. By prioritizing the received external input / output device connection status, connection to the external input / output devices 5a to 5m to which the PC servers 1a to 1n that are performing business operations are connected is avoided. That is, the server management device 2j updates the external input / output device connection status currently held in the memory 9 with the received external input / output device connection status.

On the other hand, in FIG.
Each of the server management devices 2a to 2n of the servers 1a to 1n updates the server management information based on the configuration information and the failure information received from the server management device 2j (step 313), and determines the server connection order based on the update result. Review. (Step 314). However, since these PC servers 1a to 1n are performing business operations, the server management devices 2a to 2n end the processing as they are (step 315). Waiting P
As for the C server 1k, since there is no space in the external input / output devices 5a to 5m, it remains in the standby state (step 316).

FIG. 9B showing the time when the PC server 1j fails.
FIG. 9C shows that the PC server 1i is performing an alternative operation of the PC server 1j. In FIG. 9C showing when the PC server 1j is restored, the server connection order is reviewed. Indicates that the PC server 1j is changed from the fourth rank to the third rank, so that the system configuration needs to be reconfigured. However, at this point, since the business process is ongoing, the current external input / output device connection state is maintained regardless of the server connection order.

By restarting the PC server 1i,
The procedure for restoring the PC server 1j to business processing will be described with reference to FIGS. 6, 9, and 11. This process is based on the premise that the business process by the PC server 1i can be temporarily stopped.

After the business processing is stopped, the PC server 1i
The initialization process is executed again according to the procedure described in (1). The processing procedure is the same as the procedure at the time of restoration of the PC server 1j. The server management apparatus 2i of the PC server 1i receives the initialization processing completion notification (step 190), and the other server management apparatuses 2a to 2a.
2n transmits the configuration information, the failure information, and the external input / output device connection status (steps 311 and 312), and the server management device 2i starts the external input / output device connection destination determination process (step 200). In this case, since the state of the PC server 1i has not been changed, there is no change in the server connection order, and the result of the server connection order review in step 200 and step 314 is the same as the state when the PC server 1j in FIG. Will be the same. However, since the server connection order of the PC server 1i is lower than that of the standby PC server 1j, connection with the external input / output device 5a cannot be performed, and the PC server 1i enters a standby state. At the same time, all server management devices 2a to 2n
Changes the external input / output device connection status of the PC server 1i to none from the external input / output device 5a.

At this time, the standby server management device 2j also
Upon receiving the initialization processing completion notification from the server management device 2i, the server performs response processing according to the procedure shown in FIG. The server management device 2j determines the external input / output device 5a as a connection destination as a result of the review of the server connection order in step 314. The server management device 2j proceeds to step 316 because it is waiting in step 315, and determines from the external input / output device connection state determined in step 314 that the connection destination of the external input / output device 5a is connectable, Proceed to step 317. As a result, the PC server 1j connects to the external input / output device 5a and starts system startup (step 31).
8).

As described above, in the present embodiment, the business processing of the PC server 1j is restored by the initialization processing of the PC server 1i. However, the business processing of the PC server 1i is stopped by all the server management apparatuses 2a to 2n. The same effect can be obtained by connecting the standby PC server 1j to the external input / output device 5a to start the business process.

In the present embodiment, the alternative operation of the PC server that has become inoperable and the return processing after restoration have been described. However, although it is not impossible to operate the PC server, the processing capacity is reduced due to, for example, degeneration of the processor. Accordingly, the present invention can also be applied to a case where a notification of an alternative operation is transmitted when it is desired to perform an alternative operation on another PC server.

In the present embodiment, each server management device 2a-2n is made to select only one optimal external input / output device 5a-5m based on the configuration information and the failure information. You can make it select,
For example, the second best external input / output device 5a-5
Alternatively, only one m may be selected.

Embodiment 2 In the first embodiment, in principle, PC servers with higher processing capacity and higher reliability are connected in order, and the remaining PC servers with lower connection order are put on standby. Alternate operation is performed on a highly reliable server from among the servers. However, if a PC server with a low processing capacity is made to stand by and an alternative operation of a PC server with a high processing capacity is performed, the alternative PC server becomes a bottleneck and the processing capacity of the entire system decreases. May occur.

Therefore, in the present embodiment, a specific PC server having relatively high processing capability and reliability is set as a substitute server from the start of the system to prevent a reduction in the processing capability of the entire system. I did it.
In the present embodiment, this is realized by adding operation control information for making the PC server stand by as a substitute server when each server is started up, to the configuration information of the PC server. Hereinafter, the operation in the present embodiment will be described using the example of the configuration information and the failure information shown in FIG.

In FIG. 12A, the server management apparatuses 2a to 2a
2n shows the server group, the effective performance score, the reliability score, the server connection order, the operation status, the external input / output connection status among the configuration information and the fault information held by 2n, and additionally shows the operation control. In the first embodiment, the server connection order is determined in the same server group unit. However, here, even in the same server group, the PC servers 1c and 1e for which the operation control is designated as standby and the other PC servers 1c and 1e. The operation control of the PC servers 1a, 1b, and 1n designated as driving separately determines the connection order. External input / output device 5a-
The connection destination with the PC 5m is first compared with the server connection order and the device connection order of the PC servers 1a, 1b, 1n for which the operation control is designated as operation, and the PC servers 1n, 1a, 1b are ordered in descending order of the server connection order. To the external input / output devices 5a,
5b, 5c. As a result, the server group IM
Are connected to any of the PC servers 1n, 1a, 1b whose operation control is designated as operation, and the connection of the PC servers 1c, 1e whose operation control is designated as standby. The PC server 1c having the highest effective performance score and the highest reliability score stands by in the first server connection order. As described above, even a PC server with a high processing capability and high reliability, for which a standby instruction is set, is a substitute PC server.

At this stage, the PC servers 1a, 1
If any one of the external input / output devices 5a to 5c is stopped due to a failure, the external input / output device 5c having the lowest device connection order remains unconnected at the time of connection destination determination. Become. As a result, the PC designated as standby
Among the servers 1c and 1e, the first-ranked PC server 1c
Is connected to the external input / output device 5c and starts business processing.

Further, if the PC server 1n is stopped due to a failure during the business process after the connection destination of the external input / output devices 5a to 5c is determined, as shown in FIG.
PC server 1 with the highest connection order in the same server group
c is selected as the alternative PC server, the external input / output device 5a connected to the PC server 1n is reconnected to the PC server 1c, and business processing is continued. Since the PC server 1c has higher scores in both the effective performance score and the reliability score than the PC server 1n, the processing performance and reliability of the entire business process do not decrease during the alternative operation. That is, by setting a standby instruction in advance in the operation control information of the PC server 1c having a higher processing capability or the like, the PC server 1c is caused to wait when the system is started up. Even if the PC server 1c executes the business process in place of the PC server 1n, the processing capacity and the like do not decrease.

When the PC server 1n is restored, as shown in FIG. 12C, the PC server 1c is in the alternative operation, and there are no external input / output devices 5a to 5m to be assigned to the PC server 1n. Therefore, the PC server 1n
Is in a standby state.

Then, the business process in the PC server 1c is temporarily stopped, and the PC server 1c is initialized and restarted, as shown in FIG.
The PC server 1n is returned to the business process, and the PC server 1c is again put in the standby state. This is PC server 1
c, the external input / output device 5a is disconnected from the PC server 1c, and the connection destination of the external input / output device 5a is determined between the PC server 1c, which has completed the initialization process, and the standby PC servers 1e, 1n. Review the operation control information between the PCs, so that the PCs whose operation is specified in the operation control information
This is performed by determining the server 1n as a connection destination of the external input / output device 5a.

Note that the server connection order of the PC servers 1c and 1e whose operation control is designated as standby is continued from the next order of the server connection order of the PC servers 1a, 1b and 1n whose operation control is designated as operation. Can obtain the same effect. FIG. 13 shows an example of the setting contents of the server connection order. In the example shown in FIG. 13, the server connection order is determined for the PC servers 1a, 1b, and 1n designated as operation in the operation control, and then the PC servers 1c and 1e designated as standby in the operation control. 2 shows the result of assigning the order from the server connection order 4. When the system is started, the external input / output devices 5a to 5c are connected in the order of the PC servers 1n, 1a, and 1b based on the server connection order, so that the PC servers 1c and 1e have no connection destination and are automatically on standby. State.

At this stage, the PC servers 1a, 1
If any one of the external input / output devices 5a to 5c is stopped due to a failure, the external input / output device 5c having the lowest device connection order remains unconnected at the time of connection destination determination. Become. As a result, the fourth-ranked PC server 1c
Is connected to the external input / output device 5c and starts business processing.

Since the server connection order is determined continuously, an alternative operation procedure and a recovery procedure from a failure when one of the PC servers 1a, 1b, and 1n is stopped due to a failure during operation. Is the same procedure as in the first embodiment. For example, when the PC server 1n is stopped due to a failure, the fourth-ranked PC server 1c is in a standby state, so that the PC server 1n becomes an alternative PC server and continues business operations.

According to the present embodiment, a specific PC server having higher processing capability and reliability is made to stand by as a substitute server, thereby preventing a reduction in the processing capability of the entire system.

Incidentally, in each of the above embodiments, the standby P
Among the C servers, a PC server having a higher server connection order is determined as an alternative PC server. As a method of determining an alternative PC server from a plurality of standby PC servers,
Various other methods are also conceivable.

For example, the standby PC server that has become inoperable and has the closest processing capability and reliability to the PC server that has transmitted the notification to that effect may be determined as the substitute PC server, and the business process may be taken over. . This allows
The processing capacity of the entire system before and after the alternative operation can be made substantially the same.

If another PC server wants to perform an alternative operation due to a decrease in processing capacity due to processor degeneration or the like, or if an inoperable PC server is performing a processor degeneration operation, the notification time is PC in
Standby P closest to server performance and reliability
The C server is determined as an alternative PC server so that business processing can be taken over. This results in the same as above, so that the processing capacity of the entire system before and after the alternative operation can be made substantially the same. Alternatively, the standby PC server that is closest to the processing capability and reliability originally possessed by the PC server may be determined as the substitute PC server, not at the time of notification.

In the second embodiment, since the operation control information is set, each of the above-described determination methods can be easily adopted. In the first embodiment, the connection is performed in the descending order of the connection order.
Further, since the operation control information is not used, a PC server having a low processing capacity or the like is eventually substituted at the time of starting the system. However, when the system continues to operate without being down for a long time and a number of alternative operations of the PC server are performed, it is possible that a PC server having a high processing capacity or the like may be a standby PC server, and thus each of the above-described determination methods is performed. Adopting is effective.

[0111]

According to the present invention, the configuration information of each processing device and the external input / output device is shared by each processing device, and the external input / output device to be connected to each processing device is determined based on the information. Therefore, it is possible to connect an appropriate external input / output device to each processing device according to the processing capacity and reliability of each processing device, and to allocate an appropriate business process according to the processing capacity of each processing device. it can. Further, when determining the connection destination, failure information such as processor degeneration of each processing device is also taken into consideration, so that it is possible to appropriately cope with a reduction in processing capability and reliability due to the degeneration of the configuration. That is, since the system configuration is determined based on the latest information on the processing capability and the reliability of the processing device and the OS, the processing capability is partially stopped or the function is degraded due to the occurrence of a failure, and the processing capability or the function is upgraded by upgrading the processing device. It is possible to construct an economical and highly reliable distributed system in which the balance between the processing capacity and the reliability of the entire system is always kept in an optimum state in response to the improvement of the reliability and the addition of the processing device.

Further, according to the present invention, it is possible to select the optimum processing device from the remaining processing devices when a failure occurs in the processing device without using a special OS or its extended function, and to perform business processing. Therefore, it is possible to easily construct a high-performance and low-cost system using open products.

Further, since each processing device can determine the system configuration by autonomous judgment in cooperation with the processing device management devices distributed in each processing device, it is possible to flexibly cope with an increase in the scale of the distributed system, and it is also possible to partially handle the system configuration. It is possible to construct a highly reliable distributed system capable of performing business processing with an optimal system configuration even when a processing device or a processing device management device fails.

Even when a processing unit that is performing a business process becomes inoperable, a processing unit that performs an optimal alternative operation is selected from among the standby processing units in consideration of the processing capacity and reliability, and Since the processing can be taken over, it is possible to prevent a decrease in the processing capacity and the like of the entire system as much as possible even after the alternative operation.

Further, after the recovery of the inoperable processing device, the business process being executed again can be performed, so that it is possible to return to a system environment having a suitable processing capacity at the time of system startup. .

Further, since external input / output devices satisfying the restriction conditions can be set as connection candidates by providing connection restriction conditions, even if heterogeneous processing devices and external input / output devices are mixed, it is appropriate. Connection can be made.

Further, since a specific processing device can be set for standby at the time of starting the system, detailed control of the system configuration can be performed. In particular, if a specific processing unit having relatively high processing capacity and reliability is kept on standby as an alternative processing unit from the start of the system, a reduction in the processing capacity of the entire system after the alternative operation is prevented. be able to.

[Brief description of the drawings]

FIG. 1 is a block diagram showing Embodiment 1 of a distributed system according to the present invention.

FIG. 2 is a block diagram showing a PC server according to the first embodiment.

FIG. 3 is a block diagram showing an external input / output device according to the first embodiment.

FIG. 4 is a diagram showing an example of configuration information and failure information of a PC server managed by a server management apparatus according to the first embodiment.

FIG. 5 is a diagram illustrating an example of configuration information of an external input / output device managed by a server management device according to the first embodiment;

FIG. 6 is a flowchart showing a procedure for starting up the system according to the first embodiment.

FIG. 7 is a flowchart showing a procedure for notifying completion of initialization processing and obtaining server management information managed by another server management apparatus in the first embodiment.

FIG. 8 is a flowchart showing a procedure for determining a connection destination between the PC server and the external input / output device according to the first embodiment.

FIG. 9 is a diagram showing a change in the state of the PC server and the connection state of the external input / output device according to the first embodiment.

FIG. 10 is a flowchart showing processing when a PC server failure occurs in the first embodiment.

FIG. 11 is a flowchart showing a process when the PC server is restored in the first embodiment.

FIG. 12 is a diagram illustrating a change in a state of a PC server and a connection state of an external input / output device according to the second embodiment.

FIG. 13 is a diagram illustrating a change in a state of a PC server and a connection state of an external input / output device according to the second embodiment.

FIG. 14 is a block diagram showing a conventional distributed system.

[Explanation of symbols]

1a-1n PC server, 2a-2n server management device, 3 communication path, 4 external input / output switching mechanism, 5a-
5m External input / output device, 6a-6m Operating system (OS), 7, 7a, 7b Local Ar
ea Network (LAN), 8a to 8n processors, 9 memories, 10 bus bridges, 11 external input / output device connection devices, 12 ROM, 13, 15, 17,
19, 21 control device, 14 keyboard, 16 mouse, 18 display, 20 floppy disk device (FDD), 22 HDD, 23 input / output bus, 24
Processor bus, 25 non-volatile memory, 26a-2
6n disk controller 27a-27n, 28a-2
8n disk device, 29a-29n line controller,
30 Server connection device.

Claims (16)

[Claims] A plurality of processing devices; a plurality of external input / output devices each storing an operating system and a business application; and external input / output switching means for connecting the processing device and the external input / output device. A distributed system that executes a predetermined business process by activating an operating system of the external input / output device via the external input / output switching unit by the processing device; A plurality of processing device management devices for performing operation management of the device; a communication path for performing communication between the plurality of processing device management devices; and a communication channel provided in the processing device and according to an instruction from the processing device management device. Device switching means for connecting / disconnecting a device to / from the external input / output device; The devices share the configuration information and the failure information of the respective processing devices by exchanging them via the communication path, and autonomously determine the connection destination of the external input / output device based on the shared configuration information and the failure information. The device switching means performs connection / disconnection between the processing device and the external input / output device based on the connection destination of the external input / output device determined by the processing device management device. Distributed system. 2. The apparatus according to claim 1, wherein each of the processing device management devices determines an optimal connection destination of the external input / output device in terms of processing capacity and reliability based on the shared configuration information and fault information. A distributed system as described. 3. The processing apparatus management apparatus automatically generates configuration information and failure information of the processing apparatus by performing initialization processing with reference to configuration information and failure information shared when the processing apparatus is started up. 2. The distributed system according to claim 1, wherein the automatically generated configuration information and fault information are reported to another processing device management device. 4. The configuration information of the external input / output device includes an index indicating processing capability and reliability of the processing device necessary for executing a business application stored in the external input / output device. The distributed system according to claim 1, wherein: 5. The distributed system according to claim 1, wherein the configuration information of the processing device includes information on a connection destination of each of the processing devices. 6. The processing device management device according to claim 1, wherein the processing device management device is configured to determine an effective performance score indicating an effective processing capability of the processing device based on the shared configuration information and the failure information, and a reliability indicating reliability of the processing device. The external input / output device to which the processing device is to be connected is determined based on the calculated effective performance score, reliability score, and configuration information of the external input / output device. 2. The distributed system according to 1. 7. Each of the processing device management devices determines a connection order of all the processing devices based on an effective performance score and a reliability score, and determines the external input / output device based on configuration information of the external input / output device. 7. The distributed system according to claim 6, wherein a connection order of the output devices is determined, and the external input / output device having the same connection order as the processing device is determined as a connection destination. 8. The processing apparatus management apparatus according to claim 1, wherein said processing apparatus management apparatus updates configuration information or fault information stored therein in response to a report from said processing apparatus management apparatus mounted on another processing apparatus. The distributed system according to claim 1, wherein 9. The processing device management apparatus according to claim 1, wherein the processing device is connected to each of the external input / output devices, and as a result, when the processing device is not connected to any of the external input / output devices, When receiving a notification of the alternative operation from another processing device that has been in standby as a device and performing business processing, the external input device connected to the processing device that has sent the notification has been replaced with the processing device that has sent the notification. 9. The distributed system according to claim 8, wherein an output device is reconnected to the processing device so as to take over a business process executed by the processing device that has sent the notification. 10. When there are a plurality of alternative processing devices, each of the processing device management devices assigns a service to the alternative processing device having the closest processing capability and reliability at the time of the notification of the processing device that sent the notification. The distributed system according to claim 9, wherein the processing is succeeded. 11. When there are a plurality of alternative processing devices, each of the processing device management devices may be configured to execute the alternative processing that is most similar to the processing capability and reliability inherent in the processing device that sent the notification. 10. The distributed system according to claim 9, wherein the apparatus is made to take over business processing. 12. The processing device management apparatus according to claim 1, wherein, when the processing device is not connected to any of the external input / output devices as a result of connecting the processing device and the external input / output devices, 9. The apparatus according to claim 8, wherein when a plurality of said alternative processing apparatuses are present, said alternative processing apparatus having the highest connection order takes over the business processing in place of said other processing apparatus during business processing. A distributed system as described. 13. The processing device management apparatus mounted on the processing device that has taken over the business process, after the management target processing device is restored, resumes the business process from the takeover destination processing device again. The distributed system according to any one of claims 9 to 12, wherein 14. The processing device management apparatus according to claim 1, further comprising: a connection condition between the processing device and the external input / output device included in the configuration information of the processing device and the external input / output device. 2. The distribution according to claim 1, wherein the external input / output device to be connected to the processing device is determined autonomously for each of the processing device group and the external input / output device group satisfying the constraint condition. system. 15. The processing device configuration information includes operation control information for causing each processing device to stand by as an alternative processing device when starting up each of the processing devices. 2. The distributed system according to claim 1, wherein when a standby instruction is set in the operation control information, the processing apparatus is made to stand by as an alternative processing apparatus when the processing apparatus is started up regardless of the level of processing capacity and reliability. . 16. The distributed system according to claim 15, wherein a standby instruction is set in advance in the operation control information of the processing device having high processing capability and high reliability.