WO2007110986A1

WO2007110986A1 - Network computing system and multi-processor computer

Info

Publication number: WO2007110986A1
Application number: PCT/JP2006/319677
Authority: WO
Inventors: Chisato Numaoka; Ken Kutaragi; Masayuki Chatani
Original assignee: Sony Computer Entertainment Inc.
Priority date: 2006-03-28
Filing date: 2006-10-02
Publication date: 2007-10-04
Also published as: JP2007264886A; JP4963855B2

Abstract

The need for improving the resource management and task scheduling efficiency in decentralized processing on a network is unavoidably increasing. In a multi-processor computer, one sub-processor (102) included in a multi-processor (20) receives and manages information of resources, which can be provided via a network from computers thereon, and, when receiving a resource use request, causes a resource manager for processing the provisions of managed resources to be loaded on a local memory (122) for operation. The other sub-processors (102) each receive and manage a resource use request from a computer on the network, and cause a resource scheduler for assigning a resource provided by an intervention of the resource manager to the resource use request to be loaded on the local memory (122) for operation.

Description

Technical field

[0001] The present invention relates to a multiprocessor computer and a network computing system. The present invention particularly relates to a technique of distributed processing by a plurality of computers on a network.

Background art

As a form of distributed computing, a technique called grid computing has been frequently used in recent years. Grid computing is a method of constructing a virtual high-performance multiprocessing system by combining multiple computers with limited performance on a single unit via a network. Grid computing users can achieve efficient computing using resources such as processors and memory provided by this virtual multiprocessing system.

[0003] In grid computing, basically, heterogeneous PCs are frequently combined. For this reason, in general, an interface is provided for assigning tasks (also called “jobs” or “processes”) to each computer, and the resources of each computer are assigned via this interface. It will be.

Disclosure of the invention

Problems to be solved by the invention

[0004] Here, when the frequency of task assignment to resources is relatively low, there are few problems with task scheduling. In this case, once a task is assigned to multiple computers, the processor can be almost occupied until each task is completed.

[0005] However, when the time required to execute one task is relatively short with one processor having a relatively high performance of each processor, the efficiency of management of resources distributed in the network and task scheduling is improved. The impact on the overall performance of is large. In fact, recent one-chip multiprocessors have made a leap forward compared to conventional general-purpose processors. As performance is improved, improving resource management and high efficiency of task scheduling are becoming issues that cannot be avoided. In addition, if a system framework that can spread distributed processing by multiple computers to more computer users is not established, it will not be possible to increase the efficiency of task scheduling.

[0006] The present inventor has made the present invention on the basis of the above problems, and an object thereof is to construct a mechanism for efficiently using hardware resources on the network. .

Means for solving the problem

[0007] In order to solve the above problems, a network computing system according to an aspect of the present invention includes a plurality of computers connected to each other via a network, and each of the plurality of computers includes a local memory. Hardware resource information that can be provided via a network from a computer on the network by any processor in the multiprocessor including a plurality of processors that operate independently. When a hardware resource use request is received and received, a computer that operates a resource manager that processes the provision of hardware resources to be managed while being loaded into local memory, and each has local memory Includes multiple processors that operate independently A resource scheduler that receives hardware resource usage requests from computers on the network and allocates hardware resources provided by mediation of resource managers to the node resource usage requests by any processor in the multiprocessor. The computer that operates in the state loaded in the memory, the hardware resource information is transmitted to the resource manager when the hardware resource can be provided to the other computer, and the hardware resource of the other computer is also transmitted. When using it, a client computer that runs a resource management client that sends a request to use the hardware resource to the resource scheduler, and a computer on the network When the computer provides and uses the hardware resources with other computers, information on the amount of resources provided or used and the information necessary to identify the provider and users are displayed. Talia ring that is received from the resource management client and processes the payment of the usage fee A computer.

Here, the “multiprocessor” may be a multiprocessor made into one chip.

The “processor” may operate independently without interrupt processing, and the “plurality of processors” may execute processing asynchronously with each other. “Hardware resource” is the power that various hardware can conceive, such as V of multiple processors! /, And for computers with multiple multiprocessors, It may refer to any multiprocessor. Or it may be a part of a storage medium such as a memory or hard disk, a communication device such as wireless communication, or an external peripheral device.

[0009] According to this aspect, when an idle resource occurs in the multiprocessor computer, the resource information indicating that the idle resource can be provided is transmitted to the resource manager, and conversely, the idle resource of another computer is used. When it should be used, a resource usage request is sent to the resource scheduler. As a result, in the multiprocessor computer serving as the client, the provision and use of resources can be switched as appropriate according to the resource utilization status. In other computers on the network, the resource manager and the resource scheduler are distributed to different processors and operate independently, so that resources can be provided and used quickly and efficiently. Since the payment of the usage fee is processed for the provision and use of these idle resources and the profit is secured for both the resource provider and the user, the provision and use of the resource can be promoted.

Another aspect of the present invention is a multiprocessor computer. This multiprocessor computer is provided by mediation of a resource manager that receives a hardware resource use request from a computer on a network and a multiprocessor including a plurality of processors each having a local memory and operating independently. Allocate hardware resources in response to requests to use hardware resources. Send and receive data to and from other multiprocessor computers running resource schedulers, and provide and use hardware resources of computers on the network. And a network communication unit for transmitting and receiving data to and from a clearing computer that processes the payment of the usage fee. Multiprocessor, other multiprocessor computer When using other hardware resources, a request to use hardware resources is sent to the resource scheduler that runs on any multiprocessor computer on the network, and when other computer hardware resources are used It includes a processor that runs a resource management client that sends information about the content of the request and the amount of resources used to the Clearing Now computer.

[0011] Also in this aspect, in a scene in which an idle resource of another computer is to be used in a multiprocessor computer, the provision of the resource can be received by transmitting a use request. At this time, the contents of the usage request and the usage amount or the source amount are transmitted to the clearing-nose computer, so that the clearing-nose computer is used to settle the usage fee. Since the payment of the usage fee is processed for the provision and use of such idle resources and the profit is secured for both the resource provider and the user, the provision and use of the resource can be promoted.

[0012] It should be noted that any combination of the above-described constituent elements, and the constituent elements and expressions of the present invention are mutually replaced between a method, an apparatus, a system, a computer program, a recording medium storing the computer program, a data structure, and the like. These are also effective as an embodiment of the present invention. The invention's effect

[0013] According to the present invention, it is possible to construct a system that promotes the provision and use of hardware resources on a network.

Brief Description of Drawings

FIG. 1 is a diagram illustrating an example of a basic configuration of a network computer using a one-chip multiprocessor.

FIG. 2 is a diagram showing an example of a network computing system in the present embodiment.

FIG. 3 is a diagram showing an example in which a resource manager and a resource scheduler operate on a server including two multiprocessors.

FIG. 4 is a diagram showing an example in which a resource manager and a resource scheduler operate in a server including two multiprocessors connected by a high-speed network or a high-speed bus. FIG. 5 is a diagram showing an example in which a resource manager and a resource scheduler operate on sub-processors in each of two servers distributed on a network.

[FIG. 6] A diagram showing a processing procedure in three processes of a resource scheduler, a resource manager, and a resource management client.

FIG. 7 is a diagram for explaining a mechanism by which a resource management client selects a resource manager according to a resource allocation record through an operation monitoring server.

FIG. 8 is a diagram for explaining a mechanism in which a resource management client selects a resource manager according to the scheduling efficiency performance through the operation monitoring server.

FIG. 9 is a diagram showing processing procedures in three processes of a resource scheduler, a resource manager, and a resource management client in the second embodiment.

FIG. 10 is a diagram showing a configuration of a one-chip multiprocessor in a first modified example.

FIG. 11 is a diagram showing a configuration of a one-chip multiprocessor in a second modified example. Explanation of symbols

[0015] 20 multiprocessors, 41 servers, 50 network computing systems, 51 networks, 54 servers, 60 servers, 62 servers, 64 servers, 66 servers, 102 subprocessors, 122 low power memories, 202 resource managers, 206 Resource scheduler.

BEST MODE FOR CARRYING OUT THE INVENTION

[0016] (First embodiment)

The network computing system in the present embodiment is configured by connecting a plurality of computers incorporating a plurality of processors to a network. Real-time resource management in this embodiment is realized by cooperation of each computer. First, the basic configuration of each computer incorporating a plurality of processors is described, and then the entire system is described.

FIG. 1 shows an example of a basic configuration of a network computer using a one-chip multiprocessor. Network computer 10 includes multiprocessor 20, RF processor 29, AD converter 28, bridge 19, network interface 18, non-volatile memory 21, DVD player 22, USB23, graphics card 25, main memory 17, antenna 31, Conto mouth Equipped with a controller 24 and a monitor 27.

[0018] The multiprocessor 20 includes one main processor 100, a plurality of sub-processors 102, an I / O interface 104, and a memory controller 106, which are connected to the ring bus 108 so that each operates in parallel. Composed. Multiprocessor 20 in the present embodiment includes six processors as sub-processors 102. Each sub processor 102 is uniquely recognized on the ring bus 108 by an ID such as `` SPU0 '' or `` SPU1 '', for example, which processor sends a request to which processor via the ring bus 108! / To be able to identify!

The memory controller 106 is connected to the main memory 17 and mediates access to the main memory 17 by each sub-processor 102. In the present embodiment, a power variation for explaining an example in which the main memory 17 is provided outside the multiprocessor 20 may be a configuration in which the main memory 17 is included in the multiprocessor 20. ,. \ / The O interface 104 is connected to a bridge 19 for connection with various IZO devices, a graph status card 25, an RF processing unit 29 for processing wireless communication, and the like, and the main processor 100 and sub processor 102 control external devices. And mediate communication with external devices.

A network interface 18, a non-volatile memory 21, a DVD driver 22, and a USB 23 are connected to the bridge 19. The network interface 18 mediates communication with a network such as a home LAN or the Internet. The USB 23 mediates connection with a peripheral device such as a keyboard via the controller 24. The graphics card 25 includes a frame memory 26 and outputs an image signal to the monitor 27. The RF processing unit 29 includes a down converter 30 and is connected to the IZO interface 104 via the AD conversion 28. The RF processing unit 29 establishes wireless communication with the outside via the antenna 31.

The main processor 100 includes a processor core 110, an L1 cache 112, an L2 cache 114, and a sub memory controller 116. The sub memory controller 116 includes a DMAC (Direct Memory Access Controller) 118. In the main processor 100, an operating system (hereinafter referred to as “OS”) operates, and a plurality of sub processors 102 operate based on the basic processing of the OS. [0022] Each of the plurality of sub-processors 102 includes a processor core 120, a local memory 122, and a sub-memory controller 126. The sub memory controller 126 includes a DMAC 128. Since each of the sub-processors 102 has its own local memory 122, once the target program is loaded into the local memory 122, the target memory can be accessed without accessing the main memory 17 unless necessary. Can be continued. In addition, the sub processor 102 is designed not to interrupt processing due to interrupts, so it is suitable for executing various processes described later at high speed and distributing and using resources in real time over the network. ing.

[0023] The sub processor 102 may execute a program that forms part of the OS, such as a device driver or a part of a system program, or a dedicated OS that controls the entire sub processor 102 operates. Also good. The main processor 100 and the sub processor 102 are configured by an instruction set architecture (ISA) having different instruction sets. The main processor 100 operates a program that manages only the use of resources included in the computer, whereas the sub processor 102 manages the use and provision of resources included in other computers on the network. The program works.

FIG. 2 shows an example of a network computing system in the present embodiment. The network computing system 50 includes a server 41 and a sano 54 each including a multiprocessor 20, a high performance computer 52 including a plurality of multiprocessors 20 connected by a switching network 42, a personal computer 53 including a multiprocessor 20, a multiprocessor Home appliances 55 including 20 and personal computers 56 including processors other than the multiprocessor 20 are included in the form of being connected to the network 51. The network 51 may be the Internet or a LAN such as a home network. Note that the configuration of the network combining system 50 in this figure is merely an example, and the hardware configuration and the number of units connected to the network 51 are arbitrary.

[0025] The multiprocessor 20 of the server 41 includes a plurality of sub-processors 102 and one main processor 100. Each component included in the power processor is basically the same as each component included in the network computer 10 in FIG. The same. Server 41 has 102 Then, the process for executing the real-time resource management of this embodiment operates. For example, a resource manager program runs on the first sub-processor 102, a resource scheduler program runs on the second sub-processor 102, and a resource management client program runs on the third sub-processor 102. As a modification, the resource manager program operates on one of the plurality of sub-processors 102 included in the server 41, and the resource scheduler program operates on one of the sub-processors 102 included in the server 54. May be. The resource management client may run on any computer included in the network computing system 50.

[0026] Clearing Knows 77 records the use and provision status of resources by each computer, and processes the charge for the use. The operation monitoring server 98 manages the performance and efficiency of allocation for resource usage requests. The operations of the resource manager, resource scheduler, and resource management client will be described in detail below, and then the operations of the tallying house 77 and the operation monitoring server 98 will be described in detail.

[0027] (Resource Manager)

The resource manager in the present embodiment determines whether or not computing resources such as the multiprocessor 20 and the subprocessor 102 included in the multiprocessor 20 in any of the computers connected to the network 51 can be provided to other computers. It is a process operating on one sub-processor 102 for management. When any of the computing resources becomes available to other computers, the resource manager receives the computer power resource information having the available resources via the network 51.

[0028] The resource manager operates in cooperation with a resource scheduler described later. The resource manager according to the present embodiment receives a resource use request from the resource scheduler, executes a process for allocating the resources specified in the request, and returns a response to the resource scheduler. Here, when allocating resources distributed to multiple computers in response to resource usage requests, the resource manager divides the usage requests according to the specs of each computer and returns them to the resource scheduler. [0029] The resource manager has a resource management table that classifies resources that can be provided by resource type. For example, the resource manager subtracts the value of the resource counter, which is the number of resources that can be allocated for each resource type, every time resource allocation is performed. This avoids allocating multiple resources on a specific computer at the same time or allocating resource resources on the same computer over and over, and places a heavy burden on a specific computer. Can be suppressed.

[0030] If the resource manager cannot allocate a resource usage request, it returns a response indicating that allocation is impossible to the resource scheduler. In a variation, the resource manager maintains a manager management table that manages at least some resource managers that operate on sub-processors on the same or different servers. The resource manager generates resource amount information indicating the amount of resources managed by the resource management table, and periodically exchanges resource amount information with other resource managers managed by the manager management table. The resource manager refers to the manager management table, selects the resource manager that manages more available resources, and returns the identification information of the selected resource manager to the corresponding resource scheduler.

[0031] (Resource Scheduler)

The resource scheduler in the present embodiment is a process that is executed while occupying one sub-processor. When the resource scheduler receives a resource use request from any of the computers connected to the network 51, it sends the use request to the resource manager asynchronously. The resource scheduler waits for a resource to be allocated to the transmitted usage request, and after the resource is allocated by the resource manager, transfers the usage request to the computer that holds the allocated resource. If the computer that owns the resource is divided into multiple, the resource manager divides the usage request into multiple, and the resource scheduler transfers these divided usage requests to the corresponding computers. The scheduling queue is preferably held in the local memory 122 in the sub processor 102 in which the resource scheduler operates! /, But may be held in the main memory 17. [0032] As a scheduling algorithm of the resource scheduler, any scheduling method known to those skilled in the art can be employed. For example, FIFO-type scheduling that processes use requests in the order of arrival may be employed. In FIFO type scheduling, when a resource manager cannot allocate a resource in response to a resource usage request, information indicating that the usage request can be identified is sent to the computer that issued the usage request. To do. As another scheduling method, if the resource manager receives the identification information of another resource manager from the corresponding resource manager, the resource manager asynchronously sends a use request to the resource manager and returns a response to the use request. If there is, the usage request may be sent to the computer that owns the resource according to the allocation. As another scheduling method, a method of processing a usage request based on priority may be adopted. In this case, usage requests that include more resource constraints are processed preferentially. Conversely, usage requests with fewer resource constraints may be processed preferentially. Alternatively, priority may be added to the usage request, and the processing order may be determined according to the priority. In this case, you may want to have a scheduling mailing queue according to the priority.

[0033] (Resource management client)

The resource management client in the present embodiment is a process that operates as a client on each computer included in the network computing system in the present embodiment. Resource management clients have at least three roles, for example:

[0034] (1) When a resource that can be provided externally occurs in the multiprocessor 20 that is managed by the resource management client, the resource management client shifts provided resource information indicating the content of the resource that can be provided. To the resource manager. The same applies when the resource management client manages multiple multiprocessors. For example, in the high performance computer 52 of FIG. 2, even if the resource management client is operated only by the multiprocessor connected to the network interface NIC, the resources of all the multiprocessors connected to the switching network 42 are managed. Yo! / (2) When a resource shortage occurs on the computer on which the resource management client operates, the resource management client transmits a use request to one of the resource schedulers on the network 51.

(3) When the resource management client receives a usage request from another computer, the resource management client allocates a resource included in the multiprocessor 20 managed by itself to the computer that has transmitted the usage request.

[0037] In a computer having a processor other than the multiprocessor 20 such as the personal computer 56 in FIG. 2, the simplified resource management client in which the roles (1) and (3) are omitted is operated, and the multiprocessor 20 By issuing a use request on the network 51 to which a plurality of computers with a built-in computer are connected, the computer with the built-in multi-processor 20 performs the processing.

[0038] FIGS. 3 to 5 show examples in which three processes of a resource manager, a resource scheduler, and a resource management client are arranged on one or more multiprocessor servers.

FIG. 3 shows an example in which the resource manager and the resource scheduler operate in the server 60 including one multiprocessor 20. In the example of this figure, among the two sub-processors 102 included in the multiprocessor 20, the resource manager 202 operates on the sub-processor 102 to which the ID of “SPU1” is assigned, and the ID of “SPU2” is assigned. A resource scheduler 206 operates in the sub processor 102.

FIG. 4 shows an example in which a resource manager and a resource scheduler operate in a server 62 including two multiprocessors 20 connected by a high-speed network or a high-speed bus. In the example of this figure, two sub-processors 102 are selected so as to span two multiprocessors 20, and a resource manager 202 and a resource scheduler 206 operate in each. In other words, the resource manager 202 operates in the sub processor 102 to which the ID of “SPU1” included in one multiprocessor 20 is assigned, and the subprocessor to which the ID of “SPU2” included in the other multiprocessor 20 is assigned. In 102, the resource scheduler 206 operates. The two multiprocessors 20 are connected by a high-speed bus or a high-speed network. FIG. 5 shows an example in which the resource manager and the resource scheduler operate on the sub-processor 102 in each of the two servers 64 and 66 distributed on the network 51. In the example of this figure, two sub-processors 102 are selected so as to span two servers 64 and 66, and a resource manager 202 and a resource scheduler 206 operate in each. That is, the resource manager 202 operates in the sub processor 102 to which the ID “SPU2” included in the multiprocessor 20 of one Sano 64 is assigned, and the “SPU3” included in the multiprocessor 20 of the other server 66. The resource scheduler 206 operates in the sub processor 102 to which the ID is assigned.

[0042] When the examples of FIGS. 3 to 5 described above are compared, the communication efficiency increases in the order of FIGS. 3, 4, and 5. In the present embodiment, since the resource manager 202 and the resource scheduler 206 operate asynchronously and independently, any of the forms shown in FIGS. 3 to 5 can be operated without changing the basic principle.

[0043] FIG. 6 shows a processing procedure in the three processes of the resource scheduler, the resource manager, and the resource management client. In the figure, as a modification of the force in which only one clearing house 77 is shown, a plurality of clearing nodes 77 may be included in the network 51.

[0044] (Resource management client authentication)

Clearing Knows 77 only accepts information from the authenticated resource management client 200 and refuses to accept false information. The authentication of the resource management client 200 is executed when the resource management client 200 first accesses a predetermined clearing nose 77. In this authentication process, the resource management client 200 sends a certificate or resource management client ID for certifying that the software is properly created to the clearing node 77. If this certificate is successfully authenticated by Clearinghouse 77, an electronic signature is sent to the resource management client 200. When the authenticated resource management client 200 communicates with the clearinghouse 77, it sends a message with an electronic signature attached thereto.

[0045] (Registration of resource provider) The resource provider sets whether or not to provide part or all of the computer's resources to other computers through the provision setting interface 220. The provided setting interface 220 may be configured integrally with the resource management client 200, or may be configured as software separated from the resource management client 200. The offer setting interface 220 calculates an evaluation value of dividends to be paid when the provided resource is used by another computer, and displays it on the computer screen. After confirming the displayed evaluation amount, the resource provider sets the resource provision amount such as the number of sub processors, the number of multiprocessors, the memory capacity, and the provision time through the provision setting interface 220. The provision setting interface 220 of the resource management client 200a operating on the resource provider's computer network provides the provided resource information including the set resource provision contents, the computer identification information, and the provider's personal information. Transfer to Clearing Now 77 via (S2). The computer identification information is, for example, an IP address assigned to the computer in the network 51. The provider's personal information includes, for example, the provider's name, contact information such as a telephone number and email address, information necessary to identify the provider, such as an account number and credit card number, and payment of usage fees to the provider. This is necessary information. Communication between the computer and Clearing Now 77 should be kept secure. The clearing node and the mouse 77 write the information received from the provision setting interface 220 to the provider list 222 indicating the provider of the resource to be managed.

(Registration of resource users)

The user who wishes to use the resource inputs the setting for concluding the use contract with Clearing Now 77 through the use setting interface 224. The usage setting interface 224 may be configured integrally with the resource management client 200 or may be configured as software separated from the resource management client 200. Those who wish to use the resource conclude the use contract with Clearing Knows 77 through the use setting interface 224 via the resource management client 200b operating on the computer. As a usage contract, it is necessary to specify the user name, contact information such as phone number and e-mail address, account number and credit card number. The information and the information necessary for charging the user's usage fee are registered in Clearinghouse 77 (S4). When such information is registered in the clearing node 77, the clearing node 77 sends an electronic signature for uniquely identifying and authenticating the resource user to the resource user's computer.

[0047] (Procedure for using and providing resources)

First, the resource management client 200a of a computer in which an available idle resource has been generated transmits provided resource information for registering the idle resource to the selected resource manager 202 (S10).

[0048] The resource management client 200b of the computer that has detected a resource shortage such as a shortage of sub-processors issues a use request to the selected resource scheduler 206 to request the use of an idle resource in another computer (S 12) The usage request is put in the request queue 208 of the resource scheduler 206. When transmitting this usage request, the resource management client 200b reads out the electronic signature of the storage area where the electronic signature of the resource user is stored, and transmits it to the resource scheduler 206 together with the usage request. In the example of this figure, the request queue 208 is implemented in the form of a FIFO queue. Note that the resource scheduler 206 occupies the sub-processor 102 and can execute processing. The request queue 208 may be in the local memory 122 in the sub-processor 102! Yo!

[0049] The resource scheduler 206 extracts a usage request from the head of the request queue 208, and asynchronously transmits the usage request to the corresponding resource manager 202 (S14). That is, the resource scheduler 206 does not wait for the response of the resource manager 202 to this usage request, and if the allocation resource queue 210 is empty at this time, the next usage request is taken out from the request queue 208 and processed, and the allocation resource queue If 210 is not empty, go to S16. Note that the resource scheduler 206 occupies the sub-processor 102 and executes its processing. The resource allocation resource queue 210 may be in the local memory 122 in the sub-processor 102 or in the main memory 17. Good. The resource manager 202 adds the usage request received from the resource scheduler 206 to the request queue 212. [0050] The resource manager 202 retrieves the usage request from the head force of the request queue 212, and manages and allocates the idle resource power to the appropriate resource, and uses resource allocation information indicating the result of the allocation. It returns a response to the resource scheduler 206 that is the source of the request (S16). This resource allocation information is a plurality of usage request packets formed by dividing the original usage request, and each computer is assigned to the allocated resource. These plurality of usage request packets enter the allocation resource queue 210. The resource manager 202 occupies the sub processor 102 and the processing is executed. The request queue 212 may be in the local memory 122 in the sub processor 102 or may be in the main memory 17. .

[0051] If the allocated resource queue 210 is not empty, the resource scheduler 206 extracts the usage request packet from its head, and forwards the usage request packet to the resource management client 200a of the computer holding the allocated idle resource. (S 18). In the computer that provides the resource that has received the use request packet, the resource management client 200a allocates the resource corresponding to the use request and executes the requested process.

[0052] When the processing requested by the resource scheduler 206 is successfully completed, the resource management client 200b transmits to the clearing nose 77 information that can identify the use request and the destination computer of the use request. To do. Upon receiving the usage request, Talia Ring House 77 executes a process of charging the usage fee to the user specified from the electronic signature of the resource user included in the usage request according to the amount of resources allocated in the usage request. To do. On the other hand, the personal information ability account number or credit card number of the resource provider specified by the identification information of the computer to which the use request is transmitted is taken out and assigned to the account specified by these information by the use request. An instruction to pay a usage fee determined according to the amount of resources allocated is issued. Here, the amount of the usage fee paid to the provider of the resource used does not exceed the amount of the usage fee charged to the resource user! Set as / ヽ.

[0053] (Select resource manager according to actual resource allocation)

FIG. 7 shows that the resource management client 200 is reset by the operation monitoring server 98. It is a figure explaining the mechanism which selects the resource manager 202 according to a source allocation track record. The operation monitoring server 98 receives the resource allocation record 230 at regular intervals from one or more resource managers 202 operating on a sub processor of the server connected to the network 51 (S50), and manages the resource allocation record 230. . The resource allocation performance 230 at regular intervals includes, for example, the number of usage requests received at regular intervals and the amount of resources successfully allocated at regular intervals. In the example in this figure, resource allocation record 230 is sent from computer A indicating that the number of usage requests within a certain period of time is 60, and computer B power is also used with 150 requests within a certain period of time. A resource allocation record 230 indicating that the request has been made is sent, and a resource allocation record 230 indicating that the number of usage requests was 80 times within a certain time is sent from the computer C. The resource management client 200 receives the resource allocation record 230 of each computer from the operation monitoring server 98 (S52), and refers to the resource allocation record to select a resource manager with higher resource operation efficiency.

[0054] The resource provider receives a dividend for the usage fee according to the amount of the resource used, and receives a higher dividend if the idle resource provided for use by others is used more. be able to. Therefore, select one of the resource managers 202 with a higher resource allocation record so that more dividends can be provided to the resource provider, and select the resource manager 202 according to the change in resource allocation record. Change over (S54).

[0055] The resource management client 200 may automatically change the resource manager 202 according to the resource allocation result, or may change the resource manager 202 based on an explicit instruction from the resource provider. When changing based on explicit instructions from the resource provider, the resource management client 200 presents the resource provider with a list containing the IDs of available resource managers and their resource allocation performance 230, Receive explicit selection instructions from resource providers.

[0056] (Selection of resource scheduler according to scheduling efficiency performance)

FIG. 8 is a diagram illustrating a mechanism in which the resource management client 200 selects the resource scheduler 206 according to the scheduling efficiency performance through the operation monitoring server 98. The operation monitoring server 98 is a sub-server of the server connected to the network 51. The schedule efficiency record 240 is received at regular intervals from one or more resource schedulers 206 that operate in the mouth set (S60), and the schedule efficiency record 240 is managed. The schedule efficiency record 240 may be an average waiting time until a resource is allocated for a request for use in scheduling, or may be a request allocation success rate. The resource management client 200 receives the schedule efficiency record 240 of each computer from the operation monitoring server 98 (S62), and refers to the schedule efficiency record 240 in order to select the resource scheduler 206 with higher schedule efficiency.

[0057] A resource user desires that resources be allocated more continuously.

Therefore, in order to allocate resources more efficiently to resource users, select one of the resource schedulers 206 with higher V and schedule efficiency results, and change the resource scheduler 206 according to changes in the schedule efficiency results. Switch the selection (S6 4).

[0058] The resource management client 200 may automatically change the resource scheduler 206 according to the schedule efficiency record, or may change it based on an explicit instruction from the resource user. When changing based on an explicit instruction from the resource user, the resource management client 200 presents the resource user with a list including the IDs of available resource schedulers and their schedule efficiency results 240, and then performs resource allocation. Receive explicit selection instructions from the service user.

[0059] (Second Embodiment)

This embodiment is different from the first embodiment in that the consideration for use is settled between the resource user and the provider by exchanging predetermined points with the resource. In other words, in the present embodiment, points can be given to the resource provider, and other points can be used using the points. Resource exchange can be realized. “Points” here serve as a virtual currency when paying for the use of resources and can be converted into resource amounts. In addition, the “resource amount” here refers not only to the amount of hardware resources such as the number of sub processors, memory capacity, peripheral devices' devices such as WiFi device drivers, but also to resource usage time such as WiFi device driver occupation time. A concept is also included. FIG. 9 shows processing procedures in three processes of the resource scheduler, the resource manager, and the resource management client in the second embodiment. In the following, differences from the first embodiment will be mainly described, and description of common points will be omitted.

[0061] (Register point account)

The resource provider and the owner of the computer, who will be the user, can use it with Clearing Knows 77 by opening a point account for resource exchange through the account setting interface 226. Sign a contract. The account setting interface 226 transfers the computer identification information and the owner's personal information as information necessary for the use contract to the thalia ring house 77 via the network 51. The owner's personal information includes, for example, the name of the owner, contact information such as phone number and email address, identification information of the point account, etc., information necessary to identify the owner, and resource exchange points to the owner Information necessary for granting. Communication between the computer and Clearing Now 77 should be kept secure. The clearing node and the user 77 write the information received from the resource setting interface to the provider list 222 indicating the provider of the resource to be managed. Talia Ring House 77 sends an electronic signature to the computer that opened the point account to uniquely identify and authenticate the owner. The clearinghouse 77 identifies the owner's personal information including the point account by electronic signature. The point account may be placed in the clearinghouse 77 or may be placed on another server in the network 51. However, if it is located on a separate server from Clearing Now 77, it must be possible for Taliaring House 77 to have direct access to that server.

[0062] (Registration of resource provider)

The resource provider sets whether or not to provide part or all of the computer's resources to other computers through the provision setting interface 220. In the present embodiment, the provision setting interface 220 calculates the evaluation value of the resource exchange point that can be used later when the computer uses the resource of the other computer when the resource to be provided is used by another computer. Displayed on the computer screen. Also, the provided resource information in this embodiment is transferred from the provision setting interface 220 to the clearing nose 77 together with an electronic signature, and the clearing noun 77 Then, it is written in the provider list 222.

[0063] (Registration of resource users)

Those who wish to use the resource need to store in their points account points that can be exchanged for the resource.

[0064] The resource management client 200b operating on a computer that detects a shortage of resources such as a shortage of sub-processors confirms whether or not it has sufficient points to exchange for the shortage resources with a point account. In addition, when there are enough points, a use request is transmitted to the resource scheduler 206. When transmitting the usage request, the resource management client 200b reads out the electronic signature with a predetermined storage area capability, and transmits it to the resource scheduler 206 together with the usage request.

[0065] When the processing requested by the resource scheduler 206 has been successfully completed, the resource management client 200b sends a usage request and information that can identify the destination computer of the usage request to the clearing node 77. To do. Upon receiving the use request, Talia Ring House 77 identifies the user from the electronic signature included in the use request according to the amount of resources allocated in the use request, and calculates the points from the user's point account. Pull. On the other hand, the personal information ability of the resource provider identified by the identification information of the destination computer of the usage request is extracted, and the identification information of the point account is taken out and determined for the point account according to the amount of resources allocated by the usage request. Give instructions to add the amount of points. Here, it is preferable that the resource exchange point added to the provider of the used resource is set so as not to exceed the resource exchange point from which the resource user power is also subtracted.

[0066] The present invention has been described based on the embodiments. This embodiment is an exemplification, and it is understood by those skilled in the art that various modifications can be made to the combination of each component and each treatment process, and such modifications are also within the scope of the present invention. By the way. Examples of such modifications are given below.

In the first and second embodiments, the multiprocessor configured as an asymmetric type is exemplified as the multiprocessor 20 installed in the network computer 10. In the modification, a symmetric multiprocessor is installed in the network computer 10. A configuration may be adopted.

FIG. 10 shows a configuration of the one-chip multiprocessor in the first modification. The multiprocessor 300 shown in this figure is mounted on the network computer 10 instead of the multiprocessor 20 in FIG. The multiprocessor 300 includes a first processor 302, a second processor 304, a first L1 cache 306, a second L1 cache 308, a first L2 cache 310, a second L2 cache 312, and a memory controller 314. The first processor 302, the first L1 cache 306, the first L2 cache 310, and the memory controller 314 are connected to each other and operate as a processor unit that replaces the main processor 100 or the sub processor 102 of FIG. Similarly, the second processor 304, the second L1 cache 308, the second L2 cache 312 and the memory controller 314 are also connected to each other and operate as a processor unit in place of the main processor 100 or the sub processor 102 in FIG. However, only one memory controller 314 is mounted on the multiprocessor 300 and is shared by the first processor 302 unit and the second processor 304 unit. The first processor 302 and the second processor 304 are connected to each other!

[0069] As shown in the figure, the first processor 302, the first L1 cache 306, the first L2 cache 310, the second processor 304, the second L1 cache 308, and the second L2 cache 312 are connected via the memory controller 314. It is constructed symmetrically. Of the first processor 302 and the second processor 304, one processor may operate in a main and sub relationship in which a task is assigned to the other processor.

FIG. 11 shows a configuration of a one-chip multiprocessor in the second modification. The multiprocessor 320 shown in this figure is also installed in the network computer 10 instead of the multiprocessor 20 in FIG. The multiprocessor 320 includes a first processor 322, a second processor 324, a first L1 cache 326, a second L1 cache 328, an L2 cache 330, and a memory controller 332. The first processor 322, the first L1 cache 326, the L2 cache 330, and the memory controller 332 are connected to each other and operate as a processor unit that replaces the main processor 100 or the sub processor 102 in FIG. Similarly, the second processor 324, the second L1 cache 328, the L2 cache 330, and the memory controller 332 are also connected to a processor that replaces the main processor 100 or the sub processor 102 in FIG. Operates as a unit. However, only one L2 cache 330 and one memory controller 332 are installed in the multiprocessor 320, and the L2 cache 330 and the memory controller 332 are shared by the first processor 322 unit and the second processor 324 unit. The first processor 322 and the second processor 324 are connected to each other.

As shown in the figure, the first processor 322, the first L1 cache 326, the second processor 324, the second L1 cache 328, the force L2 cache 330, and the memory controller 332 are configured symmetrically. Of the first processor 322 and the second processor 324, one processor may operate in a main and sub relationship in which a task is assigned to the other processor.

[0072] In FIG. 10 and FIG. 11, a power exemplifying a configuration having two processors as a symmetric multiprocessor may be a configuration having a larger number of processors, for example, four or more processors. ,.

Industrial applicability

[0073] It is possible to construct a system that promotes the provision and use of hardware resources on the network.

Claims

The scope of the claims

[1] A system including a plurality of computers connected to each other via a network, wherein the plurality of computers are:

Hardware resources that can be provided via a network from a computer on the network by any processor in the multiprocessor that includes multiple processors that each have local memory and operate independently A computer that operates a resource manager that processes provision of hardware resources to be managed when loaded into the local memory when receiving a hardware resource use request.

Each processor in a multiprocessor including a plurality of processors each having a local memory and operating independently receives a hardware resource use request from a computer on the network and provides it by mediation of the resource manager. A computer for operating a resource scheduler for allocating a hardware resource to be used for the hardware resource use request in a state loaded in the local memory;

When the hardware resource can be provided to another computer, the hardware resource information is transmitted to the resource manager, and when the hardware resource of another computer is used, a request for using the hardware resource A client computer for operating a resource management client that transmits the resource scheduler to the resource scheduler;

Information on the amount of resources provided or used and information necessary to identify the provider and user when a computer on the network provides and uses hardware resources with other computers A clearinghouse computer that receives payment from the resource management client and processes the payment of the usage fee;

A network computing system comprising:

[2] When the allocation by the resource scheduler for the hardware resource use request is successful, the resourceing house computer uses the resource management client in the client computer that uses the assigned node resource. The information on the usage request and the allocated resource amount is received from the ant, and the user is charged for payment of the usage fee according to the resource amount. The network computing system according to claim 1, wherein a payment to an owner of the hardware resource identified from the usage request is processed at least in part.

[3] The Clearing Knows computer manages accounts of usage fees set for computers that may provide hardware resources and computers that may use hardware resources. 3. The network combination according to claim 1, wherein the usage fee is subtracted from a hardware resource user account and at least a part of the usage fee is added to a hardware resource provider account. Utilizing system.

[4] The talia ring house computer realizes a substantial exchange between the virtual currency and the resource by adding or subtracting the usage fee from the account using the virtual currency. The network computing system according to claim 3.

[5] a multiprocessor including a plurality of processors each having a local memory and operating independently;

Computer power on the network Receiving a hardware resource use request and assigning a hardware resource provided by mediation of the resource manager to the hardware resource use request with another multiprocessor computer running a resource scheduler A network communication unit that sends and receives data to and from a Clearing-Neous computer that handles the settlement of usage fees when hardware resources are provided and used by computers on the network. And

When the multiprocessor uses a hardware resource of another multiprocessor computer, the multiprocessor sends a request for using a hardware resource to the resource scheduler operating on any multiprocessor computer on the network, and the other computer. When using hardware resources, the contents of the usage request and And a processor that operates a resource management client that transmits information on the amount of used resources to the clearing computer.