JP2007272275A - Computer system, parallel initialization method, and boot program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To shorten the time for start-up of the whole system in a computer system comprising a plurality of nodes where individual OSs are operated. <P>SOLUTION: The computer system 1 comprises a plurality of nodes 100 connected to each other. The plurality of nodes 100<SB>1</SB>-100<SB>n</SB>have a first node 100<SB>2</SB>and a second node 100<SB>1</SB>. The first node 100<SB>2</SB>comprises a first initialization processing part 22 for executing the predetermined initialization processing, and an execution status communication part 23 for sending the internal status (status information 40) of a memory changed by the predetermined initialization processing by the first initialization processing part 22 to the second node 100<SB>1</SB>. The second node 100<SB>1</SB>has an execution status reflection part 14 for changing its own internal status based on the status information 40. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a computer apparatus including a plurality of nodes and an initialization method thereof, and more particularly to a cluster system in which a plurality of computer apparatuses are connected, or a multi-core processor system in which a plurality of processor cores are connected, and a parallel initialization method used for them. And a boot program.

  In recent years, cluster systems have been widely used. The cluster system is a system in which computers equipped with at least a processor, a memory, and a network device are connected by a network. The cluster system does not have a shared memory, and a separate OS (Operating System) operates on each computer. In such a cluster system, each computer operates independently. However, the OS, the library, and the application cooperate with each other, and the computers are linked to each other so that a single computer can be used as a whole for users and other computers. It is also possible to operate as if there is. That is, a plurality of computers can be managed so as to handle one computer. For this reason, even if one of the computers in the cluster system stops, the entire system does not stop, and repair and replacement can be performed while continuing the processing.

  On the other hand, there is an SMP (Symmetric Memory Processor) type computer system as a computer composed of a plurality of processors. The SMP type computer system shares memory and devices, and is different from the cluster system in that all processors are controlled by one OS. Furthermore, recently, a multi-core processor system in which a plurality of processor cores are mounted in one chip has appeared. Multi-core processor systems include a cluster type and an SMP type. An independent OS is operated in each processor core of the cluster type multi-core processor. For this reason, the cluster type multi-core processor system is treated as a plurality of microprocessors from the OS. On the other hand, in the SMP type multi-core processor, one OS is operated in the whole processor. As representative examples of cluster type multi-core processors, there are MP211 from NEC Electronics, MPcore from ARM, and CoreDuo from Intel as representatives of SMP type multi-core processors.

  Thus, in a computer system having a plurality of computers, processors, and processor cores (hereinafter collectively referred to as nodes), it is possible to perform calculations at high speed by performing parallel processing. This parallel processing has been widely used for speeding up application programs. In recent years, in order to shorten the startup time of a computer system, an approach for parallel processing of OS startup processing has appeared.

  An example of a conventional parallel bootable computer is described in Japanese Patent Application Laid-Open No. 2004-213530 (see Patent Document 1). In Patent Document 1, a first initialization processing unit that performs an OS initialization process in a state in which a thread cannot be used, and a second initialization processing unit that performs an OS initialization process in a state in which a thread is usable And a computer (information processing apparatus) having third initialization processing means for performing OS initialization processing while a user thread is available. The first initialization processing means sequentially initializes the device driver (I / O device). The second initialization processing means executes initialization of a plurality of device drivers in parallel by using threads. The third initialization processing means executes initialization of the device driver in parallel using normal parallel processing after the user level process becomes available. In the SMP type computer, the device driver initialization can be executed in parallel on a plurality of processors by the second and third initialization processing means, so that the startup time can be shortened.

  According to the parallel initialization method described in Patent Document 1, in order to boot one OS with a plurality of threads, an SMP type computer system or an SMP type multi-core processor system capable of operating a plurality of threads simultaneously on a plurality of nodes is used. Speed-up effect. However, in the cluster system in which the OS is operated separately on each computer, the number of processors controlled by one OS is smaller than the number of threads, and thus the effect of dividing into a plurality of threads cannot be sufficiently obtained. That is, with the boot speed-up method described in Patent Document 1, the expected effect cannot be obtained in a cluster system or a multi-core processor system having a plurality of nodes each operating an individual OS. That is, the conventional parallel boot method has a high speed effect in the SMP type computer system and the SMP type multicore processor system, but the speedup effect is not high even when applied to the cluster system and the cluster type multicore processor system. .

  On the other hand, conventional techniques related to parallel initialization processing are described in Japanese Patent Application Laid-Open No. 2002-189706 (see Patent Document 2) and Japanese Patent Application Laid-Open No. 6-259393 (see Patent Document 3). Japanese Patent Application Laid-Open No. 2004-228561 describes a technique for reducing the startup time of the entire system by suppressing an increase in the amount of communication between processors of individual processing units in a system having a plurality of processors. According to the technique of Patent Literature 2, when the system is started, the initial setting information held for each individual processing unit is used to perform the initial setting processing of each processing unit in parallel, and communication between the processing units is unnecessary. . Patent Document 3 describes a boot control method in a distributed memory parallel computer having a host computer and a processor element connected via a network. According to the technique of Patent Document 3, the second and subsequent nodes are started by starting the second and subsequent nodes using the data on the memory after starting the first node (processor element). Can start up faster.

In a computer system having a plurality of nodes, the number of device drivers that are initialized when the system is started and the amount of processes that are started vary from node to node. In particular, a main node (hereinafter referred to as a master node) has many devices to be initialized and processes to be activated, and the activation time is longer than that of other nodes. This increases the startup time of the entire system. In the techniques described in Patent Documents 2 and 3, since the startup process is performed in parallel for each node, a difference occurs in the startup time for each node. That is, the startup time of the entire system increases depending on the node with the longest startup time.
JP 2004-213530 A JP 2002-189706 A Japanese Patent Laid-Open No. 06-259393

  An object of the present invention is to provide a parallel initialization method capable of shortening the startup time of the entire system in a computer system including a plurality of nodes each operating an individual OS.

  [Means for Solving the Problems] will be described below using the numbers and symbols used in [Best Mode for Carrying Out the Invention] in parentheses. This number / symbol is added to clarify the correspondence between the description of [Claims] and the description of the best mode for carrying out the invention. It should not be used for interpreting the technical scope of the invention described in [Scope].

The computer system (1, 1 ′) according to the present invention includes a plurality of nodes (100 _{1 to} 100 _n , 100 ₁ ′ to 100 _n ′) connected to each other. The plurality of nodes (100 _{1 to} 100 _n , 100 ₁ ′ to 100 _n ′) include a first node (100 ₂ , 100 ₂ ′) and a second node (100 ₁ , 100 ₁ ′). The first node (100 ₂ , 100 ₂ ′) includes a first initialization processing unit (22) that executes predetermined initialization processing (initialization of a predetermined device driver and / or activation of a predetermined process). And an execution state communication unit (23) for transmitting the state information (40) changed by the initialization process in the first initialization processing unit (22) to the second nodes (100 ₁ , 100 ₁ ′). . The second node (100 ₁ , 100 ₁ ′) includes an execution state reflecting unit (14) that changes its own internal state based on the state information (40). As described above, the second node (100 ₁ , 100 ₁ ′) according to the present invention can reflect the result of the initialization process performed by the other nodes (100 ₂ , 100 ₂ ′) in the system. . That is, the initialization process in the second node can be distributed to other nodes and reflected in itself. In addition, the initialization process that has been conventionally executed in the second nodes (100 ₁ , 100 ₁ ′) can be distributed to other nodes (100 _{2 to} 100 _n , 100 ₂ ′ to 100 _n ′).

In addition, the second node (100 ₁ , 100 ₁ ′) is configured to execute a predetermined initialization process (initialization of a predetermined device driver and / or activation of a predetermined process). ). The execution state reflecting unit (14) changes the internal state by the state information (40) transferred from the first node (100 ₂ , 100 ₂ ′) and the initialization process in the second initialization processing unit (12). change. For this reason, the second node (100 ₁ , 100 ₁ ′) itself also executes part of the device driver initialization process and the process startup process, and a part of the other nodes (100 _{2 to} 100 _n , 100 ₂ '~ 100 _n '), and the result can be reflected in itself.

Each of the plurality of nodes according to the present invention _{(100 1 ~100 n, 100 1} '~100 n'), the role specifying unit (11 role of the node to determine whether the slave node is the master node, 21), the assignment designation unit (11, 21) for designating the initialization processing target (device driver to be initialized and / or the process to be activated) in the own node, and the assignment designation unit (11, 21) The initialization processing unit (12, 22) that executes initialization processing (device driver initialization and / or (process activation)) and the role specification unit (11, 21) determine that the own node is a slave node. If the status information (40) changed by the initialization processing (device driver initialization and / or process activation) in the initialization processing unit (22) is transferred to another node ( 100 ₁ , 100 ₁ ′) and an execution state communication unit (23).

When the role specifying unit (11, 21) determines that the own node is the master node, the state information (40) transferred from the other nodes (100 ₂ , 100 ₂ ′) is set to the internal state of the own node. And an execution state reflecting unit (14) for reflecting.

  The node (100, 100 ′) further includes a storage device (103) for storing assignment information (30) designating at least one of a device driver (320) to be initialized and a process (330) to be started. preferable. The sharing designation unit (11, 21) refers to the sharing information (30) to determine a device driver to be initialized and a process to be activated.

The node (100, 100 ′) stores role specification information (20) in which role information (220) for specifying a master node or a slave node is associated with an identifier of one of a plurality of nodes. It is preferable to further comprise a storage device (104). The role designation unit (11, 21) refers to the role information (220) corresponding to the identifier of the own node based on the role designation information (20) and determines the role of the own node. The execution state communication unit (23), when the own node is designated as a slave node in the role designation unit (21), refers to the role designation information (20), and the node (100 ₁ , It is preferable to further include an execution state communication unit that transfers the state information (40) to 100 ₁ ′). Each node (100, 100 ′) holds the role designation information (20), so that it can be confirmed which node is the master node.

  The state information (40) is preferably information indicating the state in the memory changed by the initialization of the device driver and the start of the process. The node designated as the master node can change the contents in its own memory based on this status information (40).

The plurality of nodes (100 _{1 to} 100 _n ) are preferably a plurality of computer devices (100) connected to each other via a network (200). Each of the plurality of computer devices (100 _{1 to} 100 _n ) preferably includes a CPU (101) that executes an individual OS (Operating System).

In a computer system according to another aspect, the plurality of nodes (100 ′) may be a plurality of processor cores (100 ₁ ′ to 100 _n ′) connected to each other via an internal bus (200 ′). preferable. Each of the plurality of processor cores (100 ₁ ′ to 100 _n ′) preferably executes a separate OS (Operating System).

  According to the computer system, the parallel initialization method, and the boot program according to the present invention, the startup time of the entire system can be shortened in a computer system including a plurality of nodes each running an individual OS.

  Embodiments of a computer system according to the present invention will be described below with reference to the accompanying drawings. In the drawings, the same or similar reference numerals indicate the same, similar, or equivalent components. The embodiment of the computer system according to the present invention will be described by taking a cluster system as an example.

(Constitution)
FIG. 1 is a configuration diagram of an embodiment of a cluster system 1 according to the present invention. Referring to FIG. 1, a cluster system 1 according to the present invention includes a plurality of computers 100 _{1 to} 100 _n connected to each other via a network 200. The network 200 is provided with a network switch (not shown), and the computers 100 _{1 to} 100 _n are connected to each other via the network switch. Hereinafter, the computers 100 _{1 to} 100 _n will be referred to as the computer 100 when collectively described.

FIG. 2 is a block diagram of a computer 100 according to the present invention. Referring to FIG. 2, a computer 100 includes a processor 101, a main storage device 102, an external storage device 103, and a NIC (Network Interface Card) 104 that are connected to each other via a bus. The external storage device 103 is exemplified by a hard disk, and stores a boot program 10, role designation information 20, and sharing information 30. The processor 101 initializes a device driver by executing the boot program 10 when the system is started, and starts a process. The main storage device 102 is exemplified by a RAM, and temporarily stores the boot program 10 executed by the processor 101. The NIC 104 is a network interface connected to a network switch (not shown) on the network 200. The NIC 104 controls transmission of data to and from other computers on the network 200. Communication between the computers 100 _{1 to} 100 _n can use TCP / IP communication. That is, each of the computers 100 _{1 to} 100 _n has a unique IP address.

The boot program 10 is a part of an OS (Operating System) executed by the processor 101, and is a program for executing a startup process. The computers 100 _{1 to} 100 _n operate on independent individual OSs. In the present embodiment, a case where Linux is used as the OS will be described. Further, the role designation information 20 and the division information 30 are stored as files in the external storage unit 103. The role designation information 20 is information for designating whether the own computer is a master node (master computer) or a slave node (slave computer). Here, the master node is a computer that becomes a main part of the starting process when the system of the cluster system 1 is started. The role designation information 20 which is a computer that takes part of the activation process at the time of system activation includes information that designates a master node or a slave node. Referring to FIG. 3A, the role designation information 20 is associated with a computer ID 210 (example: IP address) that is an identifier of the computer 100 and role information 220 that designates a master node or a slave node. Yes. Specifically, for example, in the case of Linux, if there is an entry written as “master” in the / etc / hosts file, and the IP address of the entry matches the IP address of the own computer, the master node, Otherwise, it is determined as a slave node. The shared information 30 is information for designating a device driver to be initialized by the own computer and a process to be activated by the own computer. With reference to FIG. 3B, the sharing information 30 includes information on the device driver 320 to be initialized and the process to be activated. Specifically, for example, for Linux, / etc / modules. A device driver to be initialized is described in a setting file called conf, and / etc / rc. An rc file related to the process to be started is described below d. In this embodiment, the computer 100 ₁ is designated as the master node, the computer ₁₀₀ 2 to 100 _n is assumed to be designated as slave nodes. In the present embodiment is described for convenience the computer 100 ₁ and the computer 100 ₂ operations, the is also 100 ₂ a similar structure and operation for other computers ₁₀₀ 3 to 100 _n, which is designated as a slave node needless to say Yes.

With reference to FIG. 4, the functional blocks of the cluster system 1 according to the present invention will be described. Processor 101 of the computer 100 ₁ By executing the boot program 10, roles designation unit 11, the initialization processing unit 12, the execution state communication unit 13 realizes the function of the execution status reflection portion 14. Similarly, the processor 101 of the computer 100 _2, by executing the boot program 10, roles specifying unit 21, the initialization processing unit 22 realizes the function of the execution state communication section 23.

The role assignment designation units 11 and 21 determine whether the own computer becomes a master node or a slave node based on the role designation information 20 when the computer 100 is activated. Further, the device driver to be initialized and the process to be activated are determined based on the sharing information 30. In this embodiment, roles designation unit 11 designates the computer 100 ₁ is the master node, roles specifying unit 21 the computer 100 ₂ specifies that the slave node.

  The initialization processing units 12 and 22 initialize and activate device drivers and processes determined by the role assignment designation units 11 and 21. The processing contents of the initialization processing units 12 and 22 are the same as those of the conventional boot method. For example, in the case of Linux, for a device driver, a command such as insmod is executed by the processor 101, and / etc / modules. Initialization processing is performed by incorporating a kernel module included in conf into the kernel. As for the process, / etc / rc. A program set to be d or less is executed by the processor 101 to be activated.

Execution state communication unit 13 of the computer 100 _1, which is designated as the master node receives the process state after state and starting the initialized device driver as the state information 40 by the computer 100 _2. Also, the execution state communication unit 23 of the computer 100 ₂ designated as a slave node sends the process state of the state and after activation of the initialized device driver initialization processing unit 22 as the status information 40 to the computer 100 ₁ . In this case, execution status communication unit 23 transfers the state information 40 specifies the destined for the computer 100 ₁ corresponding to the master node by referring to the role specification information 20. At this time, the execution state communication unit 23 refers to the role designation information 20 and designates the computer ID 210 corresponding to the master node as the destination, and transfers the state information 40. The state of the initialized device driver includes an execution code of the device driver, memory data being used in the main storage device 102, device driver information managed in the OS, and the like. In addition, the process state after activation includes data on the main storage device 102 used by the process and information on the process structure in the kernel.

Execution status reflection section 14 the state of the device drivers and processes same state as the source computer 100 ₂ based on the state information 40 transferred from the computer 100 _2. For example, a method of directly manipulating the data structure inside the OS is used. As this method, process migration or checkpointing can be used.

(Operation)
With reference to FIG. 5, the operation of the system activation process of the cluster system 1 according to the present invention will be described. The computers 100 ₁ and 100 ₂ belonging to the cluster system 1 start to start almost simultaneously. When the activation is started, the role assignment designation units 11 and 21 first determine whether the own computer is a master node or a slave node. Further, the device driver to be initialized and the process to be activated are determined (step S11). Specifically, the role assignment designation units 11 and 21 determine whether the local computer is a master node or a slave node from the / etc / hosts file and the IP address of the local computer. That is, if there is an entry written as “master” in the / etc / hosts file and the IP address of the entry matches the IP address of the own computer, it is determined to be a master node. Otherwise, it is determined as a slave node. To do. In addition, / etc / modules. The device driver to be initialized is determined from conf, and / etc / rc. The process to be started is determined by the file (start script) below d. Here, it is assumed that computer 100 ₁ master Bruno computer, the computer 100 ₂ is designated as a slave node.

  Next, the initialization processing units 12 and 22 execute initialization and activation for the device driver and process determined in step S11 (step S12). Specifically, regarding the device driver and process determined in step S11, the device driver is initialized by modprobe, and / etc / rc. d Start the process by executing the daemon startup script below.

If it is determined in step S11 that the node is a slave node (No in step S13), the state in the memory (state information 40) changed by the initialized device driver and the process after startup is set as the master node. Transmit (step S16). On the other hand, if it is determined in step 11 that the node is the master node (Yes in step S13), the state information 40 transmitted from the slave node is received (step S14), and the received state information 40 is stored in the internal computer. It is reflected in the state (step S15). In particular, the computer 100 ₂ so designated as slave nodes, the execution state communication section 23 transmits the state information 40 after initialization and boot the computer 100 _1. In this case, execution status communication unit 23 by using the function of process migration, and transfers the acquired state information 40 to the computer 100 _1. Although process migration is not a standard Linux function, there are several implementations at the research level that can be used. The execution state communication unit 13 of the computer 100 ₁ receives the state information 40 and sends it to the execution state reflection unit 14. The execution state reflecting unit 14 rewrites the contents of its own main storage device 102 based on the state information 40.

As described above, in the cluster system 1 according to the present invention, the activation process of the master node that is the activation subject in system activation can be distributed to other slave nodes, and the result can be reflected on the master node. As a result, the number of device drivers that are initialized at the time of system startup and the number of processes of the master node having a large number of processes to be started up are reduced, and the startup time of the entire system is reduced. Incidentally, it may be all of the initialization process and the activation process to be performed by the computer 100 ₁ is the master node is dispersed in the computer 100 ₂ to 100 _n are slave nodes. Needless to say, the slave nodes may be all or part of 100 _{2 to} 100 _n .

Furthermore, one of the computer 100 ₁ as the master node is designated in this embodiment, may be a plurality of computers are specified. At this time, it is desirable that the device node 320 of the sharing information 30 and the process 330 are specified in association with the master node that is the transfer destination of the status information 40. Based on the sharing information 30, the computer 100 designated as the slave node can determine the initialized device driver 320 and the master node computer corresponding to the activated process 330 as the transfer destination of the state information 40.

  The embodiment of the present invention has been described in detail above, but the specific configuration is not limited to the above-described embodiment, and changes within a scope not departing from the gist of the present invention are included in the present invention. . According to the present invention, the boot process, which has conventionally been performed by only one computer, is shared by a plurality of computers, and the device driver and process states initialized separately at the last part of the boot process are stored in one computer. Can be collected. For this reason, in a computer system including a plurality of computers on which individual OSs are operating, the startup time of the entire system can be shortened. For example, the computer system according to the present invention can also be applied to a multi-core processor system 1 ′ having a processor core that is operated by a unique OS.

FIG. 6 is a block diagram of a multi-core processor system 1 ′ according to the present invention. Multi-core processor system 1 ', each of the plurality of processor cores 100 ₁ operated separate OS' comprises a to 100 _n '. Hereinafter, the processor cores 100 ₁ ′ to 100 _n ′ are collectively referred to as the computer device 100 ′. The processor core 100 ′ is connected to the main storage device 102 ′ via the internal bus 200 ′ and the memory controller 400, and is connected to the external storage device 103 ′ and the NIC 104 ′ via the internal bus 200 ′ and the I / O controller 500. ing. Note that a fixed area may be assigned to the main storage device 102 ′ for each processor core 100 ′. The external storage device 103 ′ and the NIC 104 ′ are shared between the processor cores 100 ′. The external storage device 103 ′ stores a boot program 10 ′, role designation information 20 ′, and sharing information 30 ′. That is, the processor core 100 ′ can execute the boot process using the common boot program 10 ′, role designation information 20 ′, and sharing information 30 ′.

  Reference numerals that are the same as or similar to the configuration of the cluster system 1 in the reference numerals of the configuration of the multi-core processor system 1 ′ indicate the same, similar, or equivalent components, and thus the description thereof is omitted. Further, since the operation of the multi-core processor system 1 'when starting the system is similar to the operation of the cluster system 1 described above, the description thereof is omitted.

FIG. 1 is a configuration diagram of an embodiment of a cluster system according to the present invention. FIG. 2 is a block diagram of the computer according to the embodiment of the present invention. FIG. 3A is a configuration diagram of role designation information according to the present invention. FIG. 3B is a configuration diagram of shared information according to the present invention. FIG. 4 is a functional block diagram in the embodiment of the cluster system according to the present invention. FIG. 5 is a flowchart showing the startup processing operation at the time of system startup of the cluster system according to the present invention. FIG. 6 is a configuration diagram in the embodiment of the multi-core processor system according to the present invention.

Explanation of symbols

1: cluster system 1 ′: multi-core processor system 10, 10 ′: boot program 20, 20 ′: role designation information 30, 30 ′: division information 11, 21: role division designation unit 12, 22: initialization processing unit 13, 23: Execution state communication unit 14: Execution state reflection unit 320: Device driver 330: Process 100, 100 _{1 to} 100 _n : Computer 100 ′, 100 ₁ ′ to 100 _n ′: Processor core 200: Network 200 ′: Internal bus 210 : Computer ID
220: Role information 101: Processor 102, 102 ′: Main storage device 103, 103 ′: External storage device 104, 104 ′: NIC (Network Interface Card)
300: Memory controller 400: I / O controller

Claims (20)

A plurality of nodes connected to each other;
The plurality of nodes include a first node and a second node;
The first node is:
A first initialization processing unit for executing a predetermined initialization process;
An execution state communication unit that transmits state information changed by the initialization processing in the first initialization processing unit to the second node;
The computer system comprising: an execution state reflecting unit that changes an internal state of the second node based on the state information. The computer system of claim 1,
The second node is
A second initialization processing unit for executing a predetermined initialization process;
The execution state reflecting unit changes an internal state based on state information transferred from the first node and initialization processing by the second initialization processing unit. The computer system according to claim 1 or 2,
The initialization process includes a process of initializing a device driver. The computer system according to any one of claims 1 to 3,
The initialization process includes a process for starting a process. With multiple nodes connected to each other,
Each of the plurality of nodes is
A role specifying unit that determines whether the role of the own node is a master node or a slave node;
A sharing specification unit for specifying the target of initialization processing in the own node;
An initialization processing unit that executes an initialization process based on the designation of the sharing designation unit;
A computer system comprising: an execution state communication unit that transfers state information changed by the initialization processing in the initialization processing unit to another node when the role designation unit determines that the own node is a slave node . The computer system according to claim 5, wherein
A computer system further comprising: an execution state reflecting unit that reflects the state information transferred from another node in an internal state of the own node when the role specifying unit determines that the own node is a master node. The computer system according to claim 5 or 6,
The node further comprises a storage device for storing assignment information for specifying a device driver to be initialized,
The sharing designation unit designates a device driver to be initialized with reference to the sharing information,
The initialization processing unit initializes the device to be initialized. The computer system according to any one of claims 5 to 7,
The node further includes a storage device for storing shared information for specifying a process to be started,
The sharing specification unit specifies a process to be activated with reference to the sharing information,
The initialization processing unit starts up the process to be initialized. The computer system according to any one of claims 5 to 8,
The node further includes a storage device that stores role information for designating a master node or a slave node and role designation information in which any one of the plurality of nodes is associated,
The role specifying unit refers to role information corresponding to the identifier of the own node based on the role specifying information, determines the role of the own node,
A computer system further comprising: an execution state communication unit that, when the own node is designated as a slave node in the role designation unit, refers to the role designation information and transfers the state information to a node designated as a master node. The computer system according to any one of claims 1 to 9,
The state information is information indicating a state in a memory changed by an initialization process. The computer system according to any one of claims 1 to 10,
The plurality of nodes are a plurality of computer devices connected to each other via a network,
Each of the plurality of computer devices is a cluster system including a CPU that executes an individual OS (Operating System). The computer system according to any one of claims 1 to 10,
The plurality of nodes are a plurality of processor cores connected to each other via an internal bus;
Each of the plurality of processor cores is a multi-core processor system that executes an individual OS (Operating System). A step of executing initialization processing to be processed by the own node;
Transferring state information indicating the state changed by the initialization process to another node;
The other node includes a step of changing its own state based on the state information. A parallel initialization method. The parallel initialization method according to claim 13, wherein
The step of executing initialization processing to be processed by the own node is as follows:
Including the step of initializing the device driver specified in the local node,
The state information includes information indicating a state changed by initialization of the device driver. The parallel initialization method according to claim 13 or 14,
The step of executing initialization processing to be processed by the own node is as follows:
Including the step of starting the process specified in the local node,
The state information includes information indicating a state changed by starting the process. A parallel initialization method. A step of executing initialization processing to be processed by the own node;
Transferring the state information indicating the state changed by the initialization process to another node. A parallel initialization method. A step of executing initialization processing to be processed by the own node;
Obtaining status information changed by initialization processing executed by another node;
A parallel initialization method that changes its own state based on the state information. The parallel initialization method according to claim 16 or 17,
The step of executing the initialization process includes:
Initializing a given device driver,
The state information includes information indicating a state changed by initialization of the device driver. The parallel initialization method according to any one of claims 16 to 18,
The step of executing the initialization process includes:
Including starting a predetermined process;
The state information includes information indicating a state changed by starting the process. A parallel initialization method. A boot program for causing a computer to execute the parallel initialization method according to any one of claims 16 to 19.

Images