JPH09319653A - Information processor, information processing system and control method for the same - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide efficient distributed parallel processing while keeping the dependency of respective computers and effectively utilizing a hardware distributed shared memory mechanism by providing a distributed task/sled mechanism utilizing the hardware distributed shared memory mechanism. SOLUTION: Computers 201, 301 and 401 are connected so as to mutually share a physical memory through the hardware distributed shared memory mechanism. When generating distributed tasks 21, example, a distributed task managing server part 253 issues requests to kernel parts 251 and 351 and generates tasks 11 and 12. When generating and executing threads 31 and 32 through the computer 201, for example, a physical memory area corresponding to a virtual memory area to be used is secured from any physical memory of the computers 201, 301 and 401. Then, the secured physical memory area is made correspondent to that virtual memory area and this physical memory area is accessed by using the hardware distributed shared memory mechanism.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an information processing apparatus, an information processing system, and a control method thereof, in which a plurality of information processing apparatuses are connected by a hardware distributed memory mechanism and respective memories can be directly used. It is about.

[0002]

2. Description of the Related Art Conventionally, with the increase in performance of computers such as workstations and the increase in speed of networks, a plurality of workstations connected by a network are combined in a denser form and clustered (workstations). Cluster), and a method to be used for highly parallel processing has been proposed. Filed patent (File No.137922,1
37945, 137949), it is possible to create a task (distributed task) that spans multiple computers connected by a network and targets the virtual memory operating system (OS). We proposed a method (distributed task / thread model) to execute multiple threads in parallel by utilizing the.

A task is usually a unit of allocation of program execution resources such as program text and data.
A thread is generated and executed inside the task, is a unit of program execution, and is a unit of processor resource allocation. In other words, the normal task / thread model is a model that makes it possible to effectively use processor resources in a multiprocessor computer. Also,
The distributed task / thread model is an extension of this model, in which a task that has conventionally been closed inside a single computer can now exist across a plurality of computers. Therefore, this distributed task / thread model is a model in which a thread operating inside a task can operate by effectively utilizing a plurality of processors on a plurality of computers.

A method of realizing the distributed task / thread model will be briefly described as follows.

First, a distributed shared memory mechanism (software distributed shared memory) is realized by software. This is because the software that manages the distributed shared memory manages the memory in page units and holds a copy of the page on each computer like a cache system. Then, when writing to those copied pages occurs, the distributed shared memory mechanism is realized by performing the consistency holding operation as in the cache system. At the time of this consistency holding process, the packet communication method used in the conventional network communication is used,
For example, the copy page is updated by transferring the contents of the page between computers.

Using this software distributed shared memory mechanism, the memory space (program text or data space) of a plurality of tasks generated on a plurality of computers is shared, and these task groups are combined into one large task. A distributed task / thread model is realized by managing so that it can be handled as a distributed task, creating multiple threads inside the task, and enabling parallel execution.

On the other hand, in recent years, with the advent of a communication medium having reliability such as an optical cable and having a bandwidth close to the internal bus of a computer to some extent, it is becoming possible to directly show the physical memory to other computers. A system has been proposed in which the memory (or part thereof) arranged in a computer can be directly accessed from any computer as a single memory space. In other words, multiple computers connected by a high-speed and highly reliable communication medium are connected to the NUMA (Non Uniform Memory Ac
cess) type multiprocessor system. By distinguishing the mechanism that realizes distributed shared memory among multiple computers by such hardware from the distributed shared memory mechanism by software described above,
It is called a hardware distributed shared memory mechanism.

[0008]

However, on a plurality of computers having the hardware distributed shared memory mechanism as described above, an OS for a conventional workstation or the like is installed.
If you try to implement the above as it is, (1) treat as one computer with multiple computers as a whole, and only one OS exists on some computer or divided among multiple computers. Manages the whole, or
(2) Each OS has a conventional OS on each computer, and basically each OS manages only each computer and communicates via a communication line when necessary.
Will be one of.

In the case of the method (1), since the entire system is treated like one parallel computer, parallel processing can be performed by effectively utilizing the processors on each computer, while any computer can be operated during system operation. Since it is difficult to freely turn on / off or add / delete to / from the system, the independence of individual computers is lost, and it is not suitable for distributed processing.
On the other hand, in the case of the method (2), although individual computers have independence and are suitable for distributed processing, the hard distributed shared memory mechanism can be used without changing the conventional OS. Therefore, there is a problem that efficient parallel processing cannot be realized.

Further, the distributed task / thread mechanism by the soft distributed shared memory utilizing the above-mentioned packet communication,
It may be possible to implement it on multiple computers that have a hard distributed shared memory mechanism as it is, but the memory of other computers will be accessed via the copy page, and it is possible to directly access the memory of other computers. It is not possible to take advantage of the hardware distributed shared memory mechanism.

The present invention has been made in view of the above problems, and realizes a distributed task / thread mechanism using a hardware distributed shared memory mechanism, and maintains the independence of each computer and the hardware. An object of the present invention is to provide an information processing device, an information processing system, and a control method thereof that realize an efficient distributed parallel processing by utilizing a distributed shared memory mechanism.

[0012]

An information processing apparatus according to the present invention for achieving the above object has the following arrangement. That is, an information processing apparatus in which a plurality of external information processing apparatuses and physical memories are connected to each other by a hardware distributed shared memory mechanism so that they can be shared by each other, and the information processing apparatuses and the plurality of external information processing apparatuses are distributed and arranged. And a plurality of external information processing devices for generating a distributed task including a plurality of tasks and a physical memory area corresponding to a virtual memory area used for executing the distributed task. Securing means for securing from the physical memory of, the managing means for managing the correspondence between the physical memory area secured by the securing means and the virtual memory area, and the virtual memory area generated in the execution of each task of the distributed task. Depending on the use, the physical memory area associated by the management means is accessed using the hardware distributed shared memory mechanism. And an access means for.

An information processing system of the present invention for achieving the above object has the following configuration. That is, an information processing system in which a plurality of information processing devices are connected to each other so that their physical memories can be shared by a hardware distributed shared memory mechanism, and is configured by a plurality of tasks distributed in the plurality of information processing devices. And a physical memory area corresponding to a virtual memory area used for execution of the distributed task in each of the plurality of information processing apparatuses. A securing unit that secures the memory, a management unit that manages the correspondence between the physical memory area secured by the securing unit and the virtual memory area in each of the plurality of information processing devices, and a storage unit of the plurality of information processing devices. In each case, according to the use of the virtual memory area generated in the execution of each task of the distributed task, the management unit associates the virtual memory area with each other. And an access means for accessing the physical memory area by using the hardware distributed shared memory mechanism.

The control method in the information processing system of the present invention for achieving the above object is an information processing system in which a plurality of information processing devices are connected by a hardware distributed shared memory mechanism so that physical memories can be shared with each other. And a generating step of generating a distributed task including a plurality of tasks distributed in the information processing apparatus and the plurality of external information processing apparatuses, and used for executing the distributed task. A securing step of securing a physical memory area corresponding to the virtual memory area from the physical memories of the information processing apparatus and the plurality of external information processing apparatuses;
A management step of managing the physical memory area secured in the securing step and the virtual memory area in association with each other, and a management step corresponding to the use of the virtual memory area generated in the execution of each task of the distributed task An access step of accessing the attached physical memory area using the hardware distributed shared memory mechanism.

[0015]

BEST MODE FOR CARRYING OUT THE INVENTION Preferred embodiments of the present invention will be described below with reference to the accompanying drawings.

[First Embodiment] FIG. 1 is a block diagram for explaining the arrangement of a computer system according to the first embodiment. In the figure, 101 is a high-speed communication line such as an optical network, and 201, 301, and 401 are a plurality of computers connected by the communication line 101. 202, 203, 30
2, 303, 304, 402, 403 are computer 20
One or more MMUs provided in 1, 301, 401
(Memory management unit) A built-in processor (hereinafter, simply referred to as a processor or a CPU). 210, 31
Reference numeral 0 denotes a display and a keyboard as a user interface provided in the computers 201 and 301.
220 and 420 are secondary storage devices such as hard disk (HD) devices provided in the computers 201 and 401, respectively.

Display and keyboard 210, 31
The 0 and the secondary storage devices 220 and 420 do not necessarily have to be installed in the computer. For example, a computer 401 that is not equipped with a display and a keyboard is used as a server machine or a processor pool.

Numerals 250, 350 and 450 are computers 20
O that can operate independently on 1, 301 and 401
S. Reference numeral 102 denotes an inter-CPU interrupt mechanism for performing asynchronous communication between OSs operating on each computer.
260, 360, 460 are operating systems running on each computer
It is an information exchange area for performing communication between parties.

Reference numerals 230, 330 and 430 denote physical memories 240, 340 and 4 which form a single memory space as a whole.
Reference numeral 40 is a local memory for each computer. The local memory is mainly required for the OS and the like, but since a part of the physical memory 230, 330, 430 can be substituted,
The presence or absence of it does not hinder the application of the present invention. Each processor is configured to be able to access the physical memory 230, 330, 430. However, in general, the configuration is such that the access to the physical memory belonging to another computer is slower than the access to the physical memory inside the same computer as the processor performing the memory access because the communication line 101 intervenes. is there.

In this embodiment, the OS 250,
350 and 450 are local memories 24 on each computer.
0, 340, 440, even if the physical memory 23
It is possible to operate on 0, 330 and 430 as described above.

FIG. 2 is a conceptual diagram showing the relationship between the OS and the distributed tasks according to the first embodiment. As shown in FIG. 1, each computer 201, 301, 401 has a communication line 101.
Shall be connected to each other. 250,35
Reference numerals 0 and 450 are OSs that employ a virtual memory system that operates on each computer.

Reference numerals 251, 351 and 451 denote kernel units, which control tasks and threads, virtual memory processing, memory and processor resource management, I / O resource management, and a hard distributed shared memory mechanism inside the OS. By using the physical memory on its own computer as a shared memory, the process of registering / deleting it, the process of referring to the shared memory on another registered computer, and the like are performed.

Denoted at 252, 352 and 452 are distributed shared memory management server units, which are memories of a plurality of tasks existing on a plurality of computers by utilizing the above-mentioned shared memory registration / deletion / reference mechanism inside the OS. Realize sharing of space (program text, data, etc.).

Reference numerals 253, 353 and 453 denote distributed task management server units, which manage a plurality of tasks sharing a memory space by the distributed shared memory management server units 252, 352 and 452 in the OS as one distributed task. Then
Among them, multiple threads can be created and executed in parallel, and multiple processor resources on multiple computers can be effectively used.

As described above, basically, the OS is individually operated for each computer, and by adopting a configuration in which the OSs cooperate with each other, the independence of each computer is maintained and the distributed processing is also performed. To adapt.

Since 11, 12, 13, and 14 form a distributed task, the distributed shared memory management server units 252 and 35 are provided.
2, 452 are common tasks that share a memory space and are managed collectively by the distributed task management server units 253, 353, 453. In 21 and 22, a plurality of tasks sharing a memory space by the distributed shared memory management server units 252, 352, 452 are collected by the distributed task management server units 253, 353, 453, and managed as one large task. It is a task. 31,3
Reference numerals 2, 33, 34, 35, 36, and 37 are threads that are managed by the distributed task management server unit and are generated inside each task and executed in parallel.

In this embodiment, an example in which the kernel part, the distributed shared memory management server part, and the distributed task management server part are included in the OS is shown, but a microkernel structure such as MachOS (Carnegie Mellon University) is adopted. In the case of an operating system, the kernel part corresponds to a microkernel, and the distributed shared memory management server part and the distributed task management server part are program modules independent of the microkernel, which are tasks created and executed on the microkernel. It may be realized in a form. As described above, in the present embodiment, there is no particular limitation on the implementation form of these modules.

Next, how to share the memory space of a plurality of tasks on a plurality of computers by using the hard distributed shared memory mechanism will be described.

In this embodiment, the case of the demand paging system will be described. In this method, a physical memory area (hereinafter, referred to as a physical page because it is handled in a unit called a normal page) is not allocated at the time of task generation. For this reason, generation of a thread, which is a program execution unit, and execution of the thread are started, and a page fault occurs when a virtual address where a physical page does not exist is accessed. In this method, first at this timing, a physical page is reserved, necessary contents (program text or data, or all zeros) are written in the reserved physical page, mapped to the virtual address space of the task, and the thread continues. I will let you.

FIG. 3 is a flow chart showing the flow of distributed task generation and thread generation / execution processing, which is mainly performed by the distributed task management server section. FIG. 4 is a flowchart showing the flow of processing for sharing the task memory space by the page fault processing after the page fault occurs. In addition, with each flow chart,
The case of the distributed task 21 shown in FIG. 2 is shown as an example.

In step S101, the distributed task management server unit that has received a request from the user starts the generation of the distributed task 21. In this example, the distributed task management server unit 25
3 will be described as having received a request from the user. Subsequently, in step S102, the distributed task management server unit 253 requests the kernel units 351 and 251 to
The tasks 11 and 12 that constitute the distributed task 21 are generated.

Here, the task means a common execution environment (program text, data, etc.) for allowing a plurality of threads to execute therein. Therefore, the task generation means that the OS prepares the environment according to the designation from the user, creates a management block for managing the task in the OS, and starts the management.

In general, the kernel unit can generate a task only within its own computer. When the kernel unit having this task generation function is configured to directly receive the task generation request from the distributed task management server unit on another computer, for example, the distributed task management server unit 25
3 receives a task generation request from the user, task 1
For the task 1, the task is directly requested to the kernel unit 251 and for the task 12, the task is requested to the kernel unit 351 to generate the tasks 11 and 12.

On the other hand, when the kernel unit is not configured to directly receive the task generation request from the distributed task management server on another computer, for example, when the distributed task management server unit 253 receives a request from the user. At
The task 11 directly requests the kernel unit 251 to generate a task, and the task 12 requests the kernel unit 351 to generate a task via the distributed task management server unit 353 to generate the tasks 11 and 12. . As a method of passing through the distributed task management server unit 353,
Normally, the network communication means between computers provided by the kernel section may be used to communicate between the distributed task management server sections to make a request. If that is not possible, the information exchange area may be used. A request may be made by communicating between the distributed task management server units by asynchronous communication using the 260, 360 and the CPU interrupt mechanism 102 and the like. In this way, the distributed task management server units on each computer proceed with processing while cooperating with each other.

Next, in step S103, the distributed task management server unit requests the kernel unit to allocate the page fault processing to the generated tasks 11 and 12 so that the distributed shared memory management server unit will perform them. That is, the distributed shared memory management server unit 252 for the task 11 and the distributed shared memory management server unit 352 for the task 12 are set to perform the page fault process. As a result, tasks 11, 12
When a page fault occurs, control is transferred to the distributed shared memory management server units 252 and 352.

Next, in step S104, the distributed task management server unit 253 requests the kernel unit to generate and execute the threads 31, 32 and 33. Similar to the task generation, the threads 31 and 32 are requested to the kernel unit 251 and the thread 33 to the kernel unit 351 to generate and execute the threads.

A thread is a unit of execution for actually executing a program inside a task, and normally has a stack area and processor context information as its own environment. Therefore, the generation of a thread means that the OS prepares these environments according to the specification from the user, creates a management block for managing the thread inside the OS, and starts the management.

With the above, the flow of the generation processing of the distributed task 21 ends. When creating this distributed task, the user can directly specify how many tasks that make up the distributed task will be created on which computer, and how many threads will be created in which task. The server unit may automatically make the determination. In the present embodiment, these are not particularly limited.

Next, in the generation and execution of threads in the distributed task, a page fault occurs because a virtual address where a physical page does not exist yet is accessed, and control is transferred to the distributed shared memory management server unit.
The flow of processing until the shared memory page is actually mapped to the memory space of the distributed task will be described with reference to the flowchart of FIG.

In step S201, when a thread running inside the distributed task generates a page fault, the page fault process by the distributed shared memory management server unit is started. For example, when a page fault occurs in the threads 31 and 32, the page fault process by the distributed shared memory management server unit 252 is started, and when a page fault occurs in the thread 33, the page fault process by the distributed shared memory management server unit 352 is started. . Here, a case where a page fault occurs in the thread 31 is shown as an example. That is, the page fault processing by the distributed shared memory management server unit 252 will be described as an example.

When the page fault process is started, in step S202, the distributed shared memory management server unit 252.
Checks whether the physical page corresponding to the faulted page has already been reserved on some computer and mapped to one of the tasks that make up the distributed task.

Here, each distributed shared memory management server unit has, for each distributed task, a physical page already reserved as a physical page corresponding to the virtual memory area to be used and the physical page when referring to it. It has a distributed task management table that manages keys for specifying pages. Therefore, the confirmation in step S202 can be performed by referring to the distributed task management table or the like. This distributed task management table exists for each distributed task, exists inside any of the plurality of distributed shared memory management server units that cooperate, and can be referenced from any distributed shared memory management server unit. .

Here, the key is used to identify from which computer (a) which computer a certain physical page belongs to, and (b) where (address) on the memory it is. It is possible to specify. For example, when the computer A reserves a physical page on another computer (computer B) with a key, the computer (computer B) specified by the key receives a notification of the actual physical address, and thereby obtains an access right. It will be. That is, when access is requested together with this key, the computer B notifies the physical address indicated by the key to the computer A, and sets the physical page computer A so that the physical page computer A also manages the shared access. Also, the computer A that has received the notification of the physical address can access the physical page in the virtual storage space managed by the own OS by setting the address in the virtual storage management unit inside the own OS. Become.

Therefore, when the management table to be referred to does not exist in the distributed shared memory management server unit in the own computer, the distributed shared memory management server unit in which it exists is
Usually, it is sufficient to send necessary information using network communication means between computers provided from the kernel unit.
Alternatively, the necessary information may be sent by communication using the information exchange areas 260, 360 and the CPU interrupt mechanism 102.

Next, in step S203, if it is found that the corresponding physical page does not exist as a result of checking the presence or absence of the physical page, the process proceeds to step S204 in order to secure the physical page from any computer. move on. Also,
The physical page already exists on any computer,
If it is determined that the task is mapped to any task, the process proceeds to step S221.

In step S204, it is determined from which computer to secure the physical page. This determination may be designated in advance by the user when the distributed task is generated, or may be automatically performed by the distributed shared memory management server unit or the distributed task management server unit. In this embodiment, this determination method is not particularly limited.

When it is determined from which computer to secure the physical page, in step S205, the distributed shared memory management server unit 252 requests the kernel unit to secure the physical page as the shared memory, and When referring to a physical page, a key that uniquely identifies the physical page is obtained. At this time, if it is determined that the physical page is to be reserved from the computer 201 which is its own computer, the kernel unit 251 is requested to directly reserve the physical page. Further, when it is decided to secure a physical page from a computer other than the computer 201, for example, the computer 301, the computer 3
A request for securing a physical page is made to the kernel unit 351 on the computer 301 via the distributed shared memory management server unit 352 on 01.

As a method of passing through the distributed shared memory management server unit 352, if a request is made by communicating between the distributed shared memory management server units using a network communication means between computers which is usually provided from the kernel unit. Good. As another method, the request may be made by communicating between the distributed shared memory management server units by communication using the information exchange areas 260, 360 and the inter-CPU interrupt mechanism 102.

Upon receiving the request to secure the physical page, the kernel unit secures a free physical page inside the computer and manages the shared physical page inside the kernel as a shared page that can be referenced by other computers. Register in the table and determine the key value for specifying the page when referencing. The key is a value capable of uniquely specifying a target physical page in the system, and is also capable of uniquely specifying a computer in which the physical page exists. Then, the secured physical page and key are returned to the distributed shared memory management server unit 252.

Next, in step S206, the distributed shared memory management server unit 252 registers the returned physical page information and key in the distributed task management table inside the distributed shared memory management server unit 252. Next, in step S207,
The distributed shared memory management server unit 252 fills the secured physical page with necessary contents. That is, the program text or data or zero value is embedded in the physical page according to the corresponding region of the virtual space. If the content to be filled is program text or data, the corresponding content is read from a file system or the like, and the value is written to a physical page.

Next, in step S208, the distributed shared memory management server unit 252 requests the kernel unit 251 to
The physical page with the filled contents is mapped to the corresponding virtual address space of the task 11 that caused the page fault. Finally, in step S209, the distributed shared memory management server unit 252 requests the kernel unit 251 to restart the execution of the thread 31 that caused the page fault.

Next, a case will be described in which it is determined in step S203 that a physical page already exists as a result of checking the presence / absence of a physical page. When the physical page already exists, it means that the physical page has already been mapped to the corresponding page of the virtual address space of the task other than the task 11 that constitutes the distributed task 21, in this example the task 12. Means That is, in this case, by performing the process of mapping the physical page to the same page in the virtual address space of task 11, task 11 and task 1
It becomes possible to share the same memory in 2. The flow of the processing is shown below.

When it is determined in step S203 that there is a physical page corresponding to the physical page in which the page fault has occurred, the process proceeds to step S221. Step S221
At, the distributed shared memory management server unit 252 refers to the distributed task management table existing inside the distributed shared memory management server unit, and acquires the key of the corresponding physical page.

In step S222, the obtained key is used as an argument to request the kernel unit 251 to refer to the physical page already existing on another computer (for example, the computer 301). This is normally handled directly by requesting and securing the physical memory on the local computer to the kernel section on the local computer (Read / Write is possible).
On the other hand, it is because the physical memory on other computers cannot be directly handled, and in order to enable this,
This is a procedure for requesting the kernel section on the own computer to refer to the physical memory and handle it directly.

For example, the kernel unit 25 that receives the reference request
1 specifies the computer 301 which is the computer where the physical page exists based on the key, communicates with the kernel unit 351 on the computer 301, obtains the access right of the physical page and physical address information, etc. Be able to handle the physical page directly. The communication between the kernel parts is normally performed by the network communication function of the kernel part, or the information exchange areas 260 and 360 and the inter-CPU interrupt mechanism 10.
This is performed by using a communication mechanism that utilizes the 2 or the like. Then, the distributed shared memory management server unit 252 is set so that the physical page can be referenced (Read / Write enabled).

Next, proceeding to step S208, the distributed shared memory management server unit 252 requests the referenceable physical page to the kernel unit 251, maps it to the virtual address corresponding to the task 11, and finally, Step S20
At 9, the execution of the thread 31 is resumed.

As described above, in the first embodiment, the physical page is secured by the demand paging method when the distributed task is generated and executed. That is, when the process of generating a distributed task is started, first, a plurality of tasks that form the distributed task are generated on a plurality of computers, and when a page fault occurs inside the plurality of tasks, a page fault process described later is performed. To generate and execute a plurality of threads inside the plurality of tasks. Then, when any of the above threads generates a page fault, the page fault process is started.

In the page fault processing, it is confirmed whether the corresponding physical page is already mapped to any of the plurality of tasks constituting the distributed task, and if it is not mapped to any task yet. , One of the plurality of computers is determined, a physical page on that computer is secured as a shared memory, a value that becomes a key that uniquely identifies the physical page at the time of reference is determined, and Register in the distributed task management table. Then, the content (program text, data, etc.) corresponding to the secured physical page is set, the physical page with the set content is mapped to the virtual address space corresponding to the task in which the page fault occurs, and the page fault Resume the generation and execution of the thread that generated.

Further, in the page fault processing started when any of the above threads generates a page page fault, the corresponding physical page is already mapped to any of the plurality of tasks constituting the distributed task. If so, the key of the mapped physical page is acquired by referring to the distributed task management table, and direct access to the corresponding physical page based on the acquired key (hardware distributed sharing). Access using a memory mechanism). Then, the acquired physical page is mapped to the virtual address space corresponding to the task in which the page fault has occurred, and the generation and execution of the thread in which the page fault has occurred are restarted.

By performing the above processing, a plurality of tasks operating on different computers can share a memory space, and by utilizing this, a plurality of tasks can be collected and one task that spans a plurality of computers can be collected. It can be handled as a distributed task.

Therefore, according to the present invention, a distributed task / thread mechanism using the hard distributed shared memory mechanism can be realized, and while maintaining the independence of individual computers, the hard distributed shared memory can be utilized to achieve efficient distribution. It is possible to realize parallel processing.

[Second Embodiment] In the above-described first embodiment, the case of the demand paging system has been described. In the second embodiment, a case will be described in which a physical memory is mapped in all memory spaces when a task is generated. Note that such a method will be referred to as a non-demand paging method. In this case, since the page fault does not occur, the distributed shared memory management server 252 which has performed the page fault process in the first embodiment.
352 and 452 are unnecessary. Therefore, the relationship between the OS and the distributed task according to the second embodiment is a state in which the distributed shared memory management server is omitted from FIG. In the following, the flow of processing according to the second embodiment will be described using the distributed task 21 as an example.

FIG. 5 is a flow chart showing the flow of distributed task generation and thread generation / execution processing, which is mainly performed by the distributed task management server unit.

In step S301, the distributed task management server unit starts the generation of the distributed task in accordance with the request from the user. Here, the distributed task management server unit 253
Then, the generation of the distributed task 21 is started. Subsequently, in step S302, the distributed task management server unit requests the kernel unit to generate the tasks 11 and 12 configuring the distributed task 21. This task 11, 12
The generation method of is similar to that of the first embodiment.

Next, in step S303, the distributed task management server unit determines which computer physical memory is used to configure the distributed task. This determination may be designated in advance by the user when the distributed task is generated, or may be automatically performed by the distributed task management server unit. In this embodiment, this determination method is not particularly limited. Further, the total memory space of the distributed task may be configured by the physical memory on one computer, or the memory space may be divided into appropriately divided units or page units, and the physical memory on different computers may be divided. You may assign it.

Next, in step S304, the distributed task management server unit makes a request to the kernel unit according to the above determination to secure as many physical memories as necessary to configure a task as a shared memory, and When the physical memory on another computer is used, a key for specifying the physical memory is acquired. For example, the distributed task management server unit 253 directly requests the kernel unit 251 to secure the physical memory when securing the physical memory on the computer 201, which is its own computer. Further, when securing a physical memory of a computer other than the computer 201, for example, the computer 301, a request is made to the kernel unit 351 via the distributed task management server unit 353 on the computer 301 to secure the physical memory. And earn the key.

Next, in step S305, if
When the physical memory of the computer other than the computer 201, for example, the computer 301 is secured in step S304, the distributed task management server unit 253 requests the kernel unit 251 with a key as an argument and refers to the physical memory. Be able to handle directly (Read / Write possible). This processing is also similar to that of the first embodiment.

Next, the distributed task management server unit 253
In step S306, the corresponding contents are filled in the physical memory that can be directly handled as in the case of the first embodiment.

Next, in step S307, the distributed task management server unit requests the kernel unit to map the physical memory in which the contents are filled into a plurality of tasks constituting the distributed task. For the task 11, the kernel unit 251 is directly requested, and for the task 12, the kernel unit 351 is requested via the distributed task management server unit 353 to perform mapping.

Next, in step S308, the distributed task management server unit requests the kernel unit to generate and execute the threads 31, 32, 33. This thread 3
The generation and execution of 1, 32 and 33 are the same as in the first embodiment.

With the above, the generation / execution processing of the distributed task 21 in the second embodiment is completed. When creating and executing this distributed task, users can specify directly on which computer the tasks that make up the distributed task will be created, and how many threads will be created in which task. The task management server unit may automatically make the determination. In the present embodiment, these are not particularly limited.

The processing procedure according to the second embodiment described above is summarized as follows. That is, when the process of creating a distributed task is started, first, a plurality of tasks that make up a distributed task are created on a plurality of computers, and how the physical memory that makes up this distributed task is placed on a plurality of computers. Decide what to do. The physical memory whose arrangement has been determined is secured as a physical memory on each computer, and a key value for uniquely identifying the physical memory at the time of reference is determined. Here, if there is a physical memory that cannot be directly handled without performing the reference procedure among the secured physical memory, refer to the corresponding physical memory based on the determined key. Allow direct access.
Then, the content (program text, data, etc.) corresponding to the reserved physical memory or the reserved physical memory.
To map the physical memory with the above contents set to the virtual address spaces of the plurality of tasks, and generate and execute a plurality of threads inside the plurality of tasks.

As described above, according to each of the above embodiments, the memory space of a plurality of tasks generated on a plurality of computers to form a distributed task is shared by using the hard distributed shared memory mechanism. It becomes possible. Then, by utilizing this, these multiple tasks can be managed as one distributed task, multiple threads can be created inside the distributed task, and parallel execution can be performed by utilizing multiple processors on multiple computers. It will be possible. Therefore, it is possible to realize a distributed task / thread mechanism that uses the hard distributed shared memory mechanism, and it is possible to realize efficient distributed parallel processing while maintaining the independence of individual computers and utilizing the hard distributed shared memory mechanism. Play.

The present invention may be applied to a system including a plurality of types of devices (for example, host computer, interface device, reader, printer, etc.).

Further, an object of the present invention is to supply a storage medium storing program codes of software for realizing the functions of the above-described embodiments to a system or an apparatus, and to provide a computer (or CPU) of the system or apparatus.
And MPU) read and execute the program code stored in the storage medium.

In this case, the program code itself read from the storage medium implements the functions of the above-described embodiment, and the storage medium storing the program code constitutes the present invention.

As a storage medium for supplying the program code, for example, a floppy disk, hard disk, optical disk, magneto-optical disk, CD-ROM, CD
-R, a magnetic tape, a nonvolatile memory card, a ROM, or the like can be used.

When the computer executes the readout program code, not only the functions of the above-described embodiment are realized, but also the OS (Operating System) running on the computer based on the instruction of the program code. ) May perform some or all of the actual processing, and the processing may realize the functions of the above-described embodiments.

Further, after the program code read from the storage medium is written into a memory provided in a function expansion board inserted into the computer or a function expansion unit connected to the computer, based on the instruction of the program code, It goes without saying that the CPU included in the function expansion board or the function expansion unit performs part or all of the actual processing, and the processing realizes the functions of the above-described embodiments.

When the present invention is applied to the above-mentioned storage medium, the storage medium stores the program code corresponding to the above-mentioned flow chart. Briefly, in the memory map example of FIG. Each module shown will be stored in the storage medium.

That is, at least the program codes of the "generation processing module", "securing processing module", "management processing module" and "access processing module" may be stored in the storage medium.

Here, the generation processing module generates a distributed task composed of a plurality of tasks which are distributed and arranged in a plurality of external information processing apparatuses connected by the information processing apparatus of its own and the hardware distributed shared memory mechanism. It is a program module for realizing the generation processing. In addition, the reservation processing module executes a reservation process that reserves a physical memory area corresponding to a virtual memory area used for executing the distributed task from the physical memory of the information processing apparatus and the plurality of external information processing apparatuses. It is a program module for doing. The management processing module is a program module that realizes management processing for associating the physical memory area secured by the securing processing with the virtual memory area. Further, the access processing module directly uses the hardware distributed shared memory mechanism to directly allocate the physical memory area associated by the management processing according to the use of the virtual memory area generated in the execution of each task of the distributed task. This is a program module for realizing an access process for physically accessing.

[0083]

As described above, according to the present invention, distributed tasks / tasks utilizing the hardware distributed shared memory mechanism can be used.
It is possible to realize a thread mechanism and maintain the independence of each computer, and to realize efficient distributed parallel processing by utilizing the hard distributed shared memory mechanism.

Further, according to the present invention, when the use of the virtual memory area to which the physical memory area is not associated occurs in the execution of the distributed task, the physical memory area is associated with the virtual memory area. As soon as a distributed task is created, it is possible to start thread creation and execution.

Further, according to the present invention, when it is determined that the physical memory area corresponding to the virtual memory area to be used is secured in any other external information processing apparatus, the virtual memory area is determined. Can be associated with the physical memory area. Therefore, the common virtual memory area in the distributed task is associated with the only physical memory area in the system, and the tasks can share the physical memory area.

Further, according to the present invention, the physical memory area corresponding to all virtual memory areas used by the distributed task is secured prior to the execution of the generated distributed task.
After the distributed task is executed, page faults and the like do not occur, and high-speed processing can be realized.

[0087]

[Brief description of drawings]

FIG. 1 is a block diagram illustrating a configuration of a computer system according to a first embodiment.

FIG. 2 is a conceptual diagram showing a relationship between an OS and a distributed task according to the first embodiment.

FIG. 3 is mainly performed by a distributed task management server unit,
7 is a flowchart showing the flow of distributed task generation and thread generation / execution processing.

FIG. 4 is a flowchart showing a flow of processing for sharing a memory space of tasks by page fault processing after a page fault occurs.

FIG. 5 is a flowchart showing the flow of distributed task generation and thread generation / execution processing, which is mainly performed by the distributed task management server unit.

FIG. 6 is a diagram showing structural features of a control program according to the present invention.

[Explanation of symbols]

101 communication line 102 inter-CPU interrupt mechanism 201, 301, 401 computer 202, 203, 302, 303, 304, 402, 4
03 Processor 210,310 Display and keyboard 220,420 Secondary storage device 230,330,430 Physical memory 240,340,440 Local memory 250,350,450 OS 260,360,460 Information exchange area 251,351,451 Kernel part 252,352,452 distributed shared memory management server unit 253,353,453 distributed task management server unit 11,12,13,14 tasks 21,22 distributed task 31,32,34,35,36,37 threads

Claims (18)

[Claims] 1. An information processing apparatus in which a plurality of external information processing apparatuses and physical memories are connected to each other by a hardware distributed shared memory mechanism so that the physical memories can be shared by the information processing apparatus and the plurality of external information processing apparatuses. A generation unit configured to generate a distributed task including a plurality of tasks arranged in a distributed manner, and a physical memory area corresponding to a virtual memory area used for executing the distributed task, the information processing apparatus and the plurality of external units. A securing unit that secures from the physical memory of the information processing device, a managing unit that manages the correspondence between the physical memory area secured by the securing unit and the virtual memory area, and a virtual that occurs during execution of each task of the distributed task. Depending on the use of the memory area, the physical memory area associated by the management means is used by the hardware distributed shared memory mechanism. The information processing apparatus characterized by comprising an access means for accessing Te. 2. When a use of a virtual memory area that is not managed by the management unit occurs, it is determined whether or not a physical memory area corresponding to the virtual memory area is secured in the plurality of external information processing devices. When the determination means determines that the physical memory area corresponding to the virtual memory area is secured, the management means associates the physical memory area with the virtual memory area. The information processing apparatus according to claim 1, wherein the information processing apparatus is associated as a physical memory area. 3. The information processing apparatus according to claim 2, wherein the determination unit is executed when a page fault occurs during execution of the access unit. 4. A control means for executing the securing means when the determining means determines that the physical memory area corresponding to the virtual memory area is not secured by any external information processing device. The information processing apparatus according to claim 2, wherein: 5. The management means associates a plurality of physical memory areas associated with a plurality of virtual memory areas used in the distributed task with identification information for identifying each physical memory area. By referring to the table, it is determined whether or not there is a physical memory area corresponding to the virtual memory area used by the executing means, and the management means causes the determining means to determine the physical memory corresponding to the virtual memory area. When it is determined that the area is secured, the physical memory area is acquired based on the specific information of the management table, and is associated as the physical memory area corresponding to the virtual memory area. The information processing device described. 6. The management means converts an address of a virtual memory area used in the execution means into an address of a corresponding physical memory area.
6. The information processing device according to any one of 5 to 5. 7. The execution of the distributed task is to create a thread in each task of the distributed task and execute each created thread. The information processing device according to 1. 8. The securing means secures a physical memory area corresponding to all virtual memory areas used by the distributed task prior to execution of the distributed task generated by the generating means. The information processing device according to item 1. 9. An information processing system in which a plurality of information processing devices are connected by a hardware distributed shared memory mechanism so that physical memories can be shared by each other, and a plurality of tasks distributed in the plurality of information processing devices. And a physical memory area corresponding to a virtual memory area used for the execution of the distributed task in each of the plurality of information processing apparatuses. Securing means for securing from a physical memory included in the information processing device, managing means for managing the correspondence between the physical memory area secured by the securing means and the virtual memory area in each of the plurality of information processing devices; In each of the processing devices, the management unit is controlled by the management unit according to the use of the virtual memory area generated in the execution of each task of the distributed task. The information processing system characterized in that it comprises an access means for a physical memory area associated with access using the hardware distributed shared memory mechanism. 10. A control method in an information processing system in which a plurality of information processing devices are connected by a hardware distributed shared memory mechanism so that physical memories can be shared by each other, the information processing device and the plurality of external information processing devices. A generation step of generating a distributed task composed of a plurality of tasks distributed in a device, and a physical memory area corresponding to a virtual memory area used for executing the distributed task, Securement step of securing from the physical memory of the external information processing device, a management step of associating the physical memory area secured in the securing step with the virtual memory area, and a virtual process that occurs during execution of each task of the distributed task. According to the use of the memory area, the physical memory area associated by the management process is shared by the hardware. Control method characterized by comprising an access step of accessing using the memory mechanism. 11. When a use of a virtual memory area that is not managed by the management step occurs, it is determined whether a physical memory area corresponding to the virtual memory area is secured in the plurality of external information processing devices. When the determination step determines that the physical memory area corresponding to the virtual memory area is secured, the management step assigns the physical memory area to the virtual memory area. 11. The control method according to claim 10, wherein the control is performed as a physical memory area. 12. The control method according to claim 11, wherein the determination step is executed when a page fault occurs during execution of the access step. 13. If the determination step determines that the physical memory area corresponding to the virtual memory area is not secured by any external information processing device,
The control method according to claim 11, further comprising a control step of executing the securing step. 14. The management in which the determination step associates a plurality of physical memory areas associated with a plurality of virtual memory areas used in the distributed task with identification information for identifying each physical memory area. By referring to the table, it is determined whether or not there is a physical memory area corresponding to the virtual memory area used by the execution step, and the management step includes the physical memory corresponding to the virtual memory area by the determination step. 12. When it is determined that the area is secured, the physical memory area is acquired based on the specific information of the management table and is associated with the virtual memory area as a physical memory area. The described control method. 15. The control method according to claim 10, wherein the managing step converts an address of a virtual memory area used in the executing step into an address of a corresponding physical memory area. . 16. The execution of the distributed task is to create a thread in each task of the distributed task and execute each created thread.
16. The control method according to any one of 0 to 15. 17. The securing step secures a physical memory area corresponding to all virtual memory areas used by the distributed task, prior to execution of the distributed task generated by the generating step. Item 11. The control method according to Item 10. 18. A computer readable memory storing a program code for an information processing device, wherein a plurality of information processing devices are connected to each other so that a physical memory can be shared by a hardware distributed shared memory mechanism. A code of a generation process for generating a distributed task composed of a plurality of tasks distributed in a processing device and the plurality of external information processing devices, and a physical memory corresponding to a virtual memory area used for executing the distributed task. A code of a securing step of securing a memory area from the physical memory of the information processing apparatus and the plurality of external information processing apparatuses, and a management step of associating the physical memory area secured in the securing step with the virtual memory area. The management step according to the use of the code and the virtual memory area generated in the execution of each task of the distributed task. A computer-readable memory, characterized in that it comprises a code of an access step thus the physical memory area associated with access using the hardware distributed shared memory mechanism.