JP2005044274A - Grid computing system, and computing resource collection method in grid computing system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To continue service without exerting an influence on the entire system even in a fault in individual computing resources by finding in real time the optimum computing resource and by enhancing fault resistance with a need for a directory eliminated. <P>SOLUTION: The computing resources 101-113 form a grid community on a semantic information network and registers self state information as semantic information on the filters of the semantic information network to appropriately update its semantic information. A portal site 120 transmits stimulus information containing requirement information on the state of the computing resources based on a request from a computer 130 at a request origin to a plurality of computing resources 101-113. When the computing resources 101-113 receiving the stimulus information fulfill by themselves requirements based on the requirement information contained in the stimulus information, firing information is returned to the portal site 120, The portal site 120 asks the computing resource returning the firing information about processing. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a grid computing system that dynamically combines a plurality of computing resources dispersed on a network to form a virtual machine, and a computing resource collection method in the grid computing system.

  In grid computing, in order to determine which computing resource (computer resource such as a personal computer or server) is requested to perform work (calculation processing, etc.), information on the state of the computing resource (eg, CPU usage rate) And request work according to the information. Conventionally, the state of a plurality of computing resources is centrally managed by a directory prepared in a specific management server.

  For example, in the computing resource collection function (MDS: Monitoring and Discoverry Service information service) in Globus Tool Kit (registered trademark), which is recognized as the de facto standard as middleware constituting the grid computing system, as shown in FIG. Uses a directory inquiry method.

  In FIG. 13, each of the computing resources 1001 to 1003 periodically reports information on the state of its own computing resources (for example, CPU usage rate) to a directory 1010 prepared in the central management server. The user 1020 who intends to use the calculation resources 1001 to 1003 refers to the directory 1010 and searches for available calculation resources (discovery). When an available computing resource is found, each computing resource is accessed (lookup) (for example, see Non-Patent Document 1).

  When the user 1020 collects information on the computing resources 1001 to 1003 by the conventional directory inquiry method described above, there are the following problems.

First, when the directory is abnormal, the entire system managed by the directory is damaged. Secondly, the status information of the computing resource described in the directory is not always the latest. Third, when the directories are hierarchized, there is a problem that the overall management of the directories becomes complicated.
Globus Toolkit 2.2 MDS Technology Brief Draft 4-January 30 2003, [online], January 30, 2003, [Search July 16, 2003], Internet <URL: http://www.globus.org/mds /mdstechnologybrief_draft4.pdf>

  As described above, in the grid computing system using the conventional directory, it is necessary to update the state information of the computing resource as appropriate. However, if the real-time property of the update is to be improved, communication for transmitting the information is performed. There is a problem that traffic increases. In addition, since the status information of each computing resource is managed by a single directory, if the device in this directory part fails, the entire system will not function.

  The present invention has been made to solve such a problem, and an object of the present invention is to be able to find an optimal computing resource in real time, to improve the fault tolerance by eliminating the need for a directory, An object of the present invention is to provide a grid computing system and a computing resource collection method in the grid computing system that can continue services without affecting the entire system even when a computing resource fails.

  In the grid computing system of the present invention, a plurality of computing resources form a grid community on a semantic information network, and each of the computing resources registers its state information as semantic information in a filter of the semantic information network. Means for updating the semantic information; means for requesting the computing resource to process the computing resource; requesting processing to the portal site computer; and receiving the request from the portal site computer Based on a request from a requesting computer that requests processing to a resource, means for transmitting stimulus information including requirement information regarding the state of the calculation resource to the plurality of calculation resources, and the calculation resource that has received the stimulus information Is based on the requirement information included in the stimulus information, Means for returning ignition information to the computer at the portal site; means for requesting processing for the computing resource to which the computer at the portal site has returned the ignition information; and the computing resource that has received the request for processing. Comprises means for performing processing.

  The computing resource collection method in the grid computing system of the present invention includes a procedure in which a plurality of computing resources form a grid community on a semantic information network, and each computing resource means its own state information as a filter of the semantic information network. A procedure for registering as information and updating the semantic information, a procedure for requesting the computing resource to process the computer requesting processing to the portal site computer, and the portal site receiving the request The computer transmits a stimulus information including requirement information regarding the state of the calculation resource to the plurality of calculation resources based on a request from the requesting computer that requests the calculation resource to process the stimulus information. The received calculation resource is automatically determined based on the requirement information included in the stimulus information. If the information satisfies the requirements, a procedure for returning the firing information to the computer of the portal site, a procedure for the computer of the portal site to request processing for the computing resource that has returned the firing information, and the processing And a procedure for processing the computing resource that has received the request.

In the grid computing system of the present invention, a plurality of computing resources form a grid community on the semantic information network, and each computing resource registers its state information in a filter. Based on the processing request from the requesting computer, the portal site sends stimulus information including requirement information about the state of the computing resource to the computing resource, and requests the computing resource that has returned the firing information to process. .
Thereby, it is possible to find an optimal computing resource in real time, and since the directory is unnecessary, the fault tolerance is high. Furthermore, the service can be continued without affecting the entire system even when an individual computing resource fails.

  Hereinafter, a grid computing system according to a preferred embodiment of the present invention will be described with reference to the drawings. In the embodiment, an example of a grid computing system that exchanges status information of computing resources using a semantic information network will be described. In this semantic information network, transmission of information is controlled based on the meaning of the information. As a specific example, a semantic information network called SIONet (Semantic Information Oriented Network) is used. SIONet is the name of the P2P system developed by NTT (registered trademark), and SIONet itself will be described later in the section “Supplemental explanation about SIONet”.

  In this way, a plurality of computing resources exchange their own state information using the semantic information network as a communication means, and based on this, the computing resources having a desired state are searched and obtained as a result of the search (several). A virtual computer (grid computing) is realized by distributing and coordinating work (processing) to computing resources. For this reason, the following functions are expanded or added in order to retrieve information on the status of computing resources through the semantic information network.

First, the computational resource search method is expanded from simple file name mapping to a computational resource search method “based on semantic information”. By using this method, it is possible to search not only the CPU load status but also the status of other calculation resources such as the network load and the types of applications that can be executed in the collection parameter search range of the calculation resource search.
Second, using a community concept in a P2P (Peer to Peer) system, a group of computing resources is statically constructed as a community, and is divided based on policies such as available applications and security levels. By using a P2P system that can control information search across communities, unnecessary traffic can be reduced and efficient search can be performed as compared with conventional search by broadcast.

Third, by managing the participation and withdrawal of computing resources using the community participation authentication function in the P2P system, it is possible to widely recruit computing resources from general users' PCs and the like.
Fourthly, by constructing a portal site based on HTTP (HyperText Transfer Protocol) and using it as a reception window for grid calculation processing from the user, it is possible to easily implement grid computing for general users.

Hereinafter, an example in which a system is constructed using SIONet having a function of easily realizing community management and construction among P2P technologies will be described with reference to the drawings.
FIG. 1 is a diagram for explaining a grid computing system according to an embodiment of the present invention, and FIG. 2 is a diagram for explaining functions of each component of the grid computing system shown in FIG. It is. As shown in FIG. 1, the grid computing system includes computing resources 101 to 113, a portal site 120, and a request source computer (user terminal) 130.

  SIONet software and grid middleware software are installed in the computational resources 101 to 113 which are SIONet entities, and these software are started. In SIONet, a terminal (calculation resource) that starts SIONet software is referred to as a SIONet entity (also simply referred to as “entity”).

The computing resource activated as an entity constitutes a community for grid calculation using the SIONet community creation and participation function (here, in SIONet, a general community is referred to as an event place). The event place is semantic information that is a function of SIONet by dynamically or statically “semantic information” of the state of the computational resource and network characteristics (described later) by grid middleware installed in each computational resource. Register regularly with filters (also called simply “filters”).
The contents registered in the semantic information filter of each computing resource include, for the computing resource status, IP address / system administrator name / OS version / CPU load / applicable application / architecture type / memory capacity, etc. . As for network characteristics, there are bandwidth / latency / protocol and the like. In particular, the CPU load state, network (NW) latency, and the like are estimated as the future use state from the past use state using an agent and registered as semantic information.

  The computing resource activated as an entity responds (fires) to an inquiry (stimulation) from the portal site when the condition is satisfied.

  FIG. 3 is a diagram showing the software structure of the above-described calculation resource. The calculation resource (entity) 200 functions as a SIONet entity, and the software structure of the calculation resource 200 includes various OSs 201 (Windows). (Registered trademark) and UNIX (registered trademark)), communication middleware 202 (SOAP (Simple Object Access Protocol), etc.), P2P-PF (platform) 203, SIONet API (Application Program Interface) 204, P2P It has a hierarchical structure composed of middleware 211.

  Further, the P2P middleware 211 is equipped with a discovery engine controller 212 for executing a network service based on semantic information, and includes a calculation processing resource collection function 213, a calculation processing allocation function 214, and a PF (platform) linkage function. 215 and the like are provided. The calculation resource collection function 213 has a function of registering state information as a calculation resource of the own terminal and network state information in a semantic information filter. Examples of the calculation resource state information include an IP address, system management, and the like. Name, OS / version, CPU load, application, architecture type, memory, etc. are registered, and bandwidth, latency, protocol, etc. are registered as network status information. The PF (platform) cooperation function 215 includes an authentication PF, a copyright protection PF, a charging PF, and the like.

The portal site 120 is a portal site for grid computing, and has a function of accepting a grid computing processing request from the requesting computer (user terminal) 130 and constructing a virtual machine.
The portal site 120 installs SIONet software and grid middleware software, and operates as a SIONet entity. In addition, a Web service function and a virtual machine function are constructed in the upper layer.
The virtual machine management function of the portal site 120 is performed by a VMM (Virtual Machine Manager). The Web service function is a function that accepts grid computing processing from the user (requesting computer), and provides parameters for configuring various virtual machines as menus in response to requests from the user.
The virtual machine function is a function that divides a process given in response to a calculation request from a user into a plurality of parts and manages the progress and result of the process allocated to each calculation resource. The configuration of the virtual machine is determined individually according to the content of the processing request from the user.

  In addition, the portal site 120 extracts the calculation processing request from the user as “semantic information” in order to efficiently search for the calculation resource that actually configures the virtual machine according to the request from the user, and the meaning of SIONet Search (stimulate) using the information search function. Multiple portal sites will be installed to improve reliability. A user accesses a portal site by a specified URL (Uniform Resource Locator). By the DNS function, one portal site address is selected from a plurality of portal site addresses by round robin, and the portal site having the address is accessed. A connection is made. As a result, load distribution of the portal site is realized.

  FIG. 4 is a diagram showing the software structure in the portal site described above. The calculation resource (grid calculation processing resource) 300 functions as a SIONet entity. The software structure of the portal site 300 is as follows. Based on various OS 301 (Windows (registered trademark), UNIX (registered trademark), etc.), communication middleware 302 (SOAP, etc.), P2P-PF (platform) 303, SIONet API (Application Program Interface) 304, P2P middleware 311 and P2POPen API 316 have a hierarchical structure.

  The P2P middleware 311 is equipped with a discovery engine controller 312 for executing a network service based on semantic information, and includes a calculation processing resource collection function 313, a calculation processing allocation function 314, and a PF (platform) linkage function. 315 and the like are provided, which have functions of converting user requests into semantic information and searching (stimulating), selecting discovered calculation resources, and requesting calculation processing. In addition, when a Web service function 317, a grid VM manager 318, a grid VM synthesizer 319, a compiler 320, and the like are provided, and a calculation request from a user can be received by HTTP and the processing contents can be divided, Divide and aggregate results if divided. The PF (platform) cooperation function 315 includes an authentication PF, a copyright protection PF, a charging PF, and the like.

  The request source computer (user terminal) 130 is a user terminal that requests the grid computing system to perform calculation processing. Connect to the portal site 120 via HTTP through a browser and make a grid computing calculation processing request. Since the calculation process itself is performed in the grid (in the network), it is only a requirement that the browser is installed regardless of the terminal specifications. A calculation processing request by the request source computer (user terminal) 130 is implemented by selecting parameters (for example, application, calculation processing, processing fee, etc.) using a menu provided by the portal site 120.

The configuration and characteristics of the grid computing system using SIONet described above will be summarized below.
In the grid computing system, middleware for a grid computing resource collection function is constructed on the P2P (SIONEt) API of an entity that is a component of the system. By using the “semantic information search function” realized by the middleware on SIONet, it is possible to construct an optimal virtual machine in response to a calculation processing request from a requesting computer requesting a calculation processing request. Computation resources can be discovered immediately and efficiently.

  The computing resource and the portal site are both SIENet entities, and the computing resource registers the state of its own terminal as semantic information in a filter statically and dynamically. The contents registered as semantic information in the middleware for the grid resource collection function are calculation resource information and network characteristic information. The contents converted into semantic information as the computing resource information include IP address, system administrator name, OS and version, CPU load, executable application, architecture type, memory capacity (total capacity and free capacity), hard disk capacity ( Total capacity and free capacity). Further, the contents converted into semantic information as network characteristics include network load / bandwidth / latency / protocol and the like. In particular, the CPU load status and network load status are not registered in real time, but the agent function based on statistical information is used to estimate the future use of computing resources from the current usage status. The state is estimated and registered in the filter as semantic information as needed.

With the above-mentioned mechanism, it has more search parameters than the previous system, and the contents indicate the latest and optimal information of the calculation resources, so the optimal calculation resources can be collected efficiently. Can do.
In addition, since the computing resource that becomes the entity of SIONet is used as a grid computing resource, it is necessary to create a community (event place) that is a static aggregate in advance and be a member of the community. . The calculation resource is statically included in the event place, and the state of the terminal is stored in the semantic information filter, so the portal site performs a search based on the semantic information and allocates the optimal calculation resource to the extra traffic. It is possible to collect effectively without increasing the directory and without the need for a directory.
Furthermore, by using an input / output management function for an event place using an authentication function, management of registration / deletion of computing resources can be performed. Accordingly, the computing resource can be used not only for specially prepared resources but also for an unspecified number of terminals used by general users.

In addition, the event place has a function of automatically connecting the two adjacent entities and continuing the connection even when the entity constituting the event place goes down for some reason. Therefore, by using this event place, even if the computational resource goes down for some reason, only that terminal (computational resource) will leave the event place and will not affect the entire system.
Furthermore, by using an event place, it is possible to construct a community (event place) according to the application and the security level, and it is possible to give diversity to the system.

The portal site accepts a processing request from a requesting computer (user terminal), converts a computing resource for configuring a virtual machine that satisfies the request as semantic information, and uses the SIONet function based on the semantic information. (The portal site needs to participate in the same event place as the calculation resource in advance in order to perform semantic information search by SIONet).
In computing resource search from the portal site, the processing status of computing resources (for example, CPU usage rate, processing acceptance status based on statistical information based on network (NW) usage rate, memory capacity, free hard disk capacity), etc. Given that parameter, search for the optimal computational resource (stimulus).

  A plurality of portal sites that accept calculation processing requests from a requesting computer (user terminal) are deployed in order to continue the service even in the event of a failure (improvement of reliability). The DNS function is used to determine which portal site the user selects. That is, the user accesses the portal site using the URL, but a plurality of registered IP addresses are responded by round robin, thereby distributing the load and improving the fault tolerance.

  The portal site realizes processing acceptance by HTTP in order to reduce the requesting computer (user terminal) that requests grid computing calculation processing, and realizes versatility accessible from a browser.

  The configuration and characteristics of the grid computing system according to the embodiment of the present invention have been described above. The grid computing system will be described below by showing a more specific example.

  FIG. 5 is a diagram for explaining a configuration example of the SIONet network, and shows an example in which the event place is divided into a plurality according to the application and the degree of security. In the example illustrated in FIG. 5, four event places A to D are included in the network.

  The portal sites 401 to 403 exist across all existing event places A to D. The portal site that has received the request for the grid computing process from the requesting computer (user terminal) 410 divides the process as necessary, determines the necessary computing resources, and searches them based on the semantic information. Here, since the semantic information search does not spill over to event places having information not related to the calculation resource requested by the portal site, it is possible to delete the search traffic.

  The detailed processing image of each entity will be described with reference to FIGS. Here, in order to simplify the explanation, an example of processing in one event place is shown. The example shown in FIGS. 6 and 7 is a system when an event place is formed by one portal site 501 and six computing resources R1 to R6, and the leftmost requesting computer (user terminal) 510 is The example is based on the premise that processing is requested to this grid system.

  FIG. 6 is a diagram illustrating an example of a sequence representing a processing image. FIG. 7 is a diagram showing an example of semantic information. The portal site 501 that has received a processing request from the requesting computer 510 searches for the computing resources R1 to R6 in order to construct a virtual machine, and each of the conditions at that time. The states of the computational resources R1 to R6 are shown. These states of the calculation resources R1 to R6 are registered in the filters F1 to F6 as semantic information.

First, preprocessing in the calculation resources R1 to R6 and the portal site will be described.
First, the individual calculation resources R1 to R6 are activated as SIONet entities, and the middleware for grid resource search / collection is activated at the upper level. This middleware dynamically or statically registers the load status of computing resources, the name of an application that can be executed, and the like as semantic information in the filters S1 to F6 of SIONet. The individual computing resources R1 to R6 that have become SIONet entities statically configure one event place (grid event place). Here, it is assumed that an event place is formed with the configuration shown in FIGS.
The portal site 501 is also a SIONet entity and is a member of the same grid event place as the computing resource group. The portal site 501 here is assumed to be located at the place shown in the figure on the event place.

Hereinafter, with reference to FIG. 6, a processing image will be described in order. The calculation resources R1, R4 and R6 satisfy the condition (requirement information), and the other calculation resources R2, R3 and R5 do not satisfy the condition.
First, the requesting computer 510, which is a grid user requesting calculation processing, finds the nearest portal site by using the DNS function (URL → IP address conversion function), and executes the request for grid calculation processing using a Web browser. . At this time, the parameters for configuring the virtual machine, such as the application to be executed, the security level, and the turnaround time, are specified. The parameter is specified using a pull-down menu set in advance in the portal site 501 (step S1).

The portal site 501 that has received the calculation process divides the tasks that can be processed in parallel according to the calculation process request content. Further, calculation resources necessary for constructing a virtual machine for performing calculation processing are calculated based on parameters received from the request source computer (user terminal) 510. These functions are implemented by a VMM (Virtual Machine Manager). Furthermore, semantic information as a search key for collecting these necessary calculation resources is determined (process A: step S2).
In the portal site 501, in order to find the optimal calculation resource, the search by the SIONet function is started using the semantic information determined in the process A, and a “search message by semantic information” is transmitted to the calculation resource R 1 (stimulus (Step S3).

The calculation resource R1 transfers the “search message based on semantic information” received from the portal site 501 to the next calculation resource R2 connected at the event place (transfers the stimulus) (step S4).
Further, the calculation resource R1 stimulated by the portal site 501 is “ignited” by the SIONet function (ignition A) because the content of the filter F1 satisfies the conditions for the stimulation of the portal site. The information that the fire has occurred is returned to the portal site 501 together with the firing condition by the calculation resource R1. The ignition conditions are as follows. CPU = Pent * u * 1.5 GHz, CPU usage rate = 20%, OS = W * n200 *, application = P * v-Ra *, memory = 128 MB, HD = 20 GB, NW = 100 Mbps, NW usage rate = 30 % (Step S5).

  The calculation resource R2 transfers the received “search message based on semantic information” as it is to the next calculation resource R3 and the calculation resource R4 connected at the event place (stimulus is transferred) (step S6). Note that since the calculation resource R2 has two connection destinations in the event place, the stimulus is transferred to the two calculation resources R3 and R4. The computation resource R2 itself does not ignite and therefore does nothing because the content of the filter F2 does not satisfy the requirements for the received stimulus. The calculation resource R3 that has received the stimulus from the calculation resource R2 does not transfer the stimulus because there is no calculation resource connected to the event place, and the information of the filter F3 does not satisfy the requirement for the stimulus. There is no fire and nothing is done.

The calculation resource R4 that receives the stimulus from the calculation resource R2 transfers the received “search message based on semantic information” to the next calculation resource R5 and the calculation resource R6 connected at the event place as they are (transfers the stimulus). (Step S7). Since the calculation resource R4 has two connection destinations in the event place, the calculation resource R4 transfers the stimulus to two places. In addition, the calculation resource R5 and the calculation resource R6 that have received the stimulus from the calculation resource R4 do not transfer the stimulus because there is no calculation resource connected to the event place ahead of the calculation resource R4. Since some information of F5 does not satisfy the requirement for stimulation, it does not ignite and does nothing.
Further, since the content of the filter F4 satisfies the condition for the received stimulus, the calculation resource R4 “fires” by the SIONet function. The ignition conditions are as follows. CPU = Pent * u * 850 MHz, CPU usage rate = 30%, OS = W * n200 *, application = P * v-Ra *, memory = 64 MB, HD = 8 GB, NW = 10 Mbps, NW usage rate = 25% Yes (ignition B). Information that it has been ignited is returned to the portal site 501 by P2P together with the igniting condition by the computing resource R4 (step S8).
Further, since the content of the filter F6 satisfies the condition for the received stimulus, the calculation resource R6 is “fired” by the SIONet function. The ignition conditions are as follows. CPU = Pent * u * 1.7 GHz, CPU usage rate = 20%, OS = W * n200 *, application = P * v-Ra *, memory = 256 MB, HD = 100 GB, NW = 10 Mbps, NW usage rate = 15 % (Ignition C). Information that it has ignited is returned to the portal site 501 by P2P together with the igniting condition by the calculation resource R6.

The portal site 501 receives and confirms the firing response from the computing resource (this time, three responses from the computing resources R1, R4, and R6 (there are steps S5, S8, and S9)), and based on the response firing conditions. Determine the optimal computing resource. Here, since it is determined that the number of calculation resources that require the VMM is “two”, the optimal calculation resources are narrowed down to the top two from the firing conditions, and are determined as the calculation resources R1 and R6. Since the response time is set as the priority content this time, the CPU specifications, usage rate, and memory capacity are the criteria for selecting the calculation resources R1 and R6 (Process B: Step S10).
The portal site 501 allocates the calculation process to the calculation resource R1 and the calculation resource R6 based on the determination result of the process B (step S11). This allocation is performed directly by P2P in “between portal site = computation resource R1” and “between portal site = computation resource R6”.

Then, the calculation resource R1 executes the calculation process assigned to the calculation resource R1 in the process B (process C), and the process result is returned (returned) to the portal site (step S12). This response is performed by P2P directly between the portal site = the computing resource R1.
In addition, the calculation process assigned to the calculation resource R6 in the process B is executed by the calculation resource R6 (process D), and the process result is returned (returned) to the portal site 501 (step S13). This response is performed by P2P directly between the portal site = the computing resource R6.
Thereafter, in the portal site 501, the entire processing results are integrated based on the processing results of the processing resource R1 and the processing resource R6 (processing E: step S14). Further, the portal site reports the processing result of processing D to the requesting computer 510 (step S15).

  The calculation resource R4 ignites and responds, but since there is no calculation processing request from the portal site 501, a transition is made to the initial state due to timeout (process F: step S16).

Although the grid computing system according to the embodiment of the present invention has been described above, grid computing is performed while configuring the semantic information network and collecting optimal calculation resources based on the semantic information, while taking advantage of the directoryless advantage. It becomes possible to diversify the search conditions in the collection of computing resources of the system.
In addition, it is possible to update the status information (not only the CPU load, but also various information) of computing resources on demand, so that optimal computing resources can be found in real time and the performance of the grid computing system can be improved. it can.

  In addition, the directory-less system reduces dynamic information exchange and lookup procedures between computing resources and directories, leading to a reduction in extra traffic. In addition, by using a transfer function based on semantic information and an event place, broadcast search can be improved and traffic volume can be further reduced.

  In addition, it becomes possible to divide event places according to applications for processing computing resources and the degree of security, so that the system can be diversified and service variations can be increased. In addition, even in the event of an individual failure of a computing resource, the service can be continued without affecting the entire system. Furthermore, in addition to dedicated computing resources, terminals used by general users can be used as computing resources, contributing to the globalization of grid computing systems.

  As mentioned above, although the form of the preferred embodiment of this invention was demonstrated, the grid computing system of this invention is not limited only to the above-mentioned illustration example, In the range which does not deviate from the summary of this invention, it is various. Of course, changes can be made.

  The computing resource and requesting computer in this application may be any name and form as long as they have hardware and network connection functions such as a personal computer, a PDA (Personal Digital Assistant), and a mobile phone. The network includes both wired and wireless connection methods.

[Supplementary explanation about SIONet]
Here, “SIONet” used in the embodiment of the present invention will be supplementarily described.
As shown in FIG. 8, SIONet is a metanetwork that distributes events based on semantic information (metadata), and dynamically searches for a specific entity from a large number of unspecified entities that are super-distributed on the network.・ Can be discovered. In other words, SIONet is a network that delivers an event based on semantic information (sending to whom) instead of the destination address (sending to whom) used in the conventional network.

  In SIONet, as a definition of SIONet entity, all computers such as personal computers, workstations, portable information terminals, and mobile phones are collectively referred to as hosts. Furthermore, a host on which SIONet software is installed is referred to as “SIONE entity” or simply “entity”. An entity is a virtualization of a host on which SIONet software is installed as a unit of autonomous distributed cooperation in SIONet.

Entities are classified into three types: SE (service entity), NE (network entity), and SNE (service network entity).
The SE (service entity) is an entity that behaves as a service, and is a virtualized P2P application as an internal entity within the entity. The NE (network entity) is an entity that behaves as a network component, is an internal entity for constructing and operating SIONets, and a plurality of NEs (network entities) are distributed and coordinated to self-organize the network. .
An SNE (service network entity) is an entity that behaves from the viewpoint of both a service and a network component, and is used when constructing a pure P2P network, and self-organizes the network by autonomously distributing them.

  As shown in FIG. 8, the event is composed of semantic information and data. As the semantic information, for example, “resident in the suburbs of Tokyo”, “person with a classical music hobby”, “Meiji Dori” There are “people walking now”, “travel content holders” and so on. The event structure will be described later.

  FIG. 9 is a diagram for explaining the SIONet mechanism. The SIONet includes a plurality of event places, and the event places include a plurality of entities. Further, SIONet is an autonomous distributed cooperative network (autonomous distributed computer) in which the network is self-organized when all entities including the components of SIONet perform autonomous distributed cooperation. Network entities that are SIONet network components include "semantic information switch (SI-SW)", "semantic information router (SI-R)", and "semantic information gateway (SI-GW)". Elements include “session”, “shared link (SL)”, and the like, and by self-organizing as necessary, a secure and scalable P2P network can be constructed by a bottom-up approach.

  The semantic information switch (SI-SW) is a network entity (NE) that provides a switching mechanism for switching events based on semantic information, and the semantic information router (SI-R) is based on semantic information. , A network entity (NE) that performs event routing (event transfer) between SI and SW, and a semantic information gateway (SI-GW) is a network entity (NE) that realizes event sharing between event places. .

  The “session” is a connection between the semantic information switch (SI-SW) and the service entity (SE), and the service entity (SE) can transmit and receive events only through the session. There are three types of sessions: an event transmission session, a reception session, and a transmission / reception session. The “shared link” is a concept for performing bidirectional event sharing between a plurality of different entities. For example, an entity group (event place) is formed when a certain entity requests establishment of a shared link (SL) to another entity.

FIG. 10 is a diagram illustrating an example of the structure of an SIONet event. The SIONet event includes a control information part (CIPart (Control Information Part)), a semantic information part (SIPart (Semantic Information Part)), and a data part ( DATATAP).
The control information part (CIPart) is a control field for effective control of SIONet and is not open to the service entity (SE). The semantic information part (SIPart) stores semantic information (vocabulary and its value) of transmission data and a lexical concept (event type) of semantic information. The data part (DATATAPart) stores transmission data (various data and programs such as text data and binary data).
In addition, a stimulus is transmitting an event. A response is that the filter matches the event. Firing means starting an entity that has registered a matched filter and notifying an event.

  Moreover, FIG. 11 is a figure for demonstrating a semantic information switch and a filter, and a semantic information switch (SI-SW) provides the switching mechanism which switches an event based on semantic information. The service entity (SE) transmits and receives events via a session (connection between SI-SW and SE).

  The filter is a content that the service entity (SE) registers the event acquisition condition in the semantic information switch (SI-SW) through the session. Set the event type (event vocabulary concept) of the event you want to acquire and the matching condition with the semantic information.

FIG. 12 is a diagram for explaining the event place and the semantic information gateway, and the event place is an aggregate of semantic information switches (SI-SW).
The event place is a concept of glue pink, which is a concept used as a security guarantee unit, a unit of load distribution, grouping of groups, etc., and achieves the following objectives.
First, the event transfer range is limited. That is, it limits the spread of chain reactions based on stimulation and ignition.
Second, the target range of event routing is limited. It is a set of semantic information switches (SI-SW) connected by a shared link, and a link topology or event routing method can be selected as an event place.
Third, it is a space that guarantees the uniqueness of the semantic information system, and suppresses the explosive increase in the event types handled.

  The semantic information gateway (SI-GW) is a network entity (NE) for realizing event sharing between event places. As a result, it is possible to link event places having different uses and purposes.

  The present invention can be applied to a grid computing system, a computing resource collection method in the grid computing system, and the like.

The figure for demonstrating the grid computing system in embodiment of this invention. The figure for demonstrating the function of the component of a grid computing system. The figure which shows the software structure in a calculation resource (entity). The figure which shows the software structure in a portal site. The figure for demonstrating the structural example of a network. The figure which shows the example of the sequence showing a process image. The figure which shows the example of semantic information. The figure which shows the outline | summary of SIONet. The figure for demonstrating the mechanism of SIONet. The figure which shows the example of the structure of an event. The figure for demonstrating a semantic information switch and a filter. The figure for demonstrating an event place and a semantic information gateway. The figure which shows the general resource management mechanism of the conventional grid computing system.

Explanation of symbols

101-113 computing resource 120 portal site 130 requesting computer 200 computing resource 300 portal site 401-403 portal site 410 requesting computer 501 portal site 510 requesting computer

Claims (2)

Means for a plurality of computing resources to form a grid community on a semantic information network;
Each computing resource registers its own state information as semantic information in a filter of the semantic information network, and updates the semantic information;
A requesting computer that requests processing to the computing resource, a means for requesting processing to the computer of the portal site;
The computer of the portal site that has received the request transmits stimulus information including requirement information regarding the state of the computing resource to the plurality of computing resources based on a request from the requesting computer that requests the computing resource to process. Means,
Means for returning the firing information to the computer of the portal site when the computing resource that has received the stimulus information satisfies the requirements based on the requirement information included in the stimulus information;
Means for requesting processing of the computing resource from which the computer of the portal site has returned the ignition information;
A grid computing system comprising: means for performing processing on the computing resource that has received the processing request. A procedure in which multiple computing resources form a grid community on a semantic information network;
Each computing resource registers its own state information as semantic information in a filter of the semantic information network and updates the semantic information;
A requesting computer that requests processing to the computing resource makes a request for processing to the computer of the portal site;
The computer of the portal site that has received the request transmits stimulus information including requirement information regarding the state of the computing resource to the plurality of computing resources based on a request from the requesting computer that requests the computing resource to process. Procedure and
When the calculation resource that has received the stimulus information satisfies the requirements based on the requirement information included in the stimulus information, a procedure for returning ignition information to the portal site computer;
A procedure in which the computer of the portal site requests processing for the computing resource that has returned the ignition information;
A calculation resource collecting method in a grid computing system, comprising: a procedure for processing the calculation resource that has received the processing request.

Images