KR101055548B1 - Semantic Computing-based Dynamic Job Scheduling System for Distributed Processing - Google Patents

Abstract

A dynamic job scheduling system based on semantic computing for distributed processing is provided. A semantic computing-based dynamic job scheduling system includes a data input / analyzer that receives data, a data object corresponding to the data, and a requirement for processing the data; A job processing relay unit controlling the overall operation so that the data is processed; A resource manager configured to collect and maintain attribute information of the resource through communication with at least one resource and to select available resources corresponding to the data characteristics and requirements; And a scheduling management unit which selects a resource for processing the data from the selected available resources through communication with the resource, distributes the data to the selected resource, and provides a processing result for the distributed data.

Distributed Processing System, Dynamic Job Scheduling, Semantic Computing, Resource Management

Description

Semantic Computing-based Dynamic Job Scheduling System for Distributed Processing {SEMANTIC COMPUTING-BASED DYNAMIC JOB SCHEDULING SYSTEM FOR DISTRIBUTED PROCESSING}

Embodiments of the present invention relate to a dynamic job scheduling system based on semantic computing for distributing and distributing jobs to heterogeneous resources geographically dispersed.

In a grid computing or distributed computing environment, a method of distributing work to geographically dispersed resources is described as "a dynamic resource allocation method based on a user's QoS and application characteristics in a grid. 0621787) ". The above-described conventional technology derives an objective function by converting a nonlinear model into a linear model to process a task in a short time in consideration of the performance of resources in a grid computing environment. However, it is difficult to schedule efficiently by reflecting various dynamic states of resources by displaying only CPU performance, workload, and network state, which represent resource calculation capability as factors included in the objective function. In addition, there is a problem in compatibility and scalability in defining and adding a new kind of resource and updating or applying an optimal job scheduling method according to user requirements and characteristics of data to be processed.

Another method of distributing jobs to existing geographically dispersed resources is "a job scheduling apparatus and method of a distributed processing system" (Korean Patent Laid-Open Publication No. 1998-079594). The prior art described above provides the ability to schedule a task at any period, the ability to skip a task, and a method of scheduling tasks occurring on resources in a network remotely. However, in the geographically distributed computing environment, the above-described conventional techniques may cause an imbalance in resource usage or congestion of the network by not considering the performance and dynamic change of resources, congestion of the network, and the like.

The present invention provides a dynamic task scheduling system based on semantic computing that enables easier resource management and easy system expansion in a heterogeneous computing environment.

The present invention provides a dynamic task scheduling system based on semantic computing that can process a task using a resource most suitable according to a user's budget and data characteristics.

The present invention provides a dynamic task scheduling system based on semantic computing that provides an efficient scheduling technique for distributing and allocating tasks to distributed computing resources.

A semantic computing-based dynamic job scheduling system according to an embodiment of the present invention includes a data input / analyzer that receives data for a task, data characteristics corresponding to the data, and requirements for processing the data; A job processing relay unit which controls the overall operation so that the data is processed; A resource manager configured to collect and maintain attribute information of the resource through communication with at least one resource and to select available resources corresponding to the data characteristics and requirements; And a scheduling manager configured to select a resource to process the data among the selected available resources through communication with the resource, distribute the data to the selected resource, and provide a processing result for the distributed data. .

In an embodiment of the present invention, the data input / analysis unit converts the input data characteristic into an integer form and then transfers the data analyzer to the work processing relay unit, and converts the input requirement into an integer form and then performs the operation. It includes a requirements analyzer that forwards to the processing relay.

In one embodiment of the present invention, the job processing relay unit is an available resource requester for requesting a resource list for the available resources from the resource manager, and a resource list of the available resources provided by the resource manager in response to a request of the resource list. And a resource selector for selecting a resource corresponding to the data characteristic and the requirement among the available resources of the received resource list by using an inference rule defined for the data characteristic and the requirement. .

In an embodiment of the present invention, the resource manager may include a resource information requester for transmitting an attribute information request message to the resource, a resource information receiver for receiving the attribute information provided by the resource with respect to the request message, and the resource. A resource ontology manager for expressing the received attribute information using an ontology standard language for management, a resource information repository for storing the attribute information expressed in the ontology standard language in association with the resource, and control of the job processing relay unit And a resource information transmitter that selects available resources corresponding to the data characteristics and requirements and provides a resource list composed of the available resources to the work processing relay.

In one embodiment of the present invention, the scheduling management unit according to the received dynamic attribute information using a resource state information receiver for receiving the dynamic attribute information of the resource and an inference rule defined for the attribute information of the resource. An inference engine unit for determining a rank of the available resources, a job divider for distributing and allocating the data to a resource having the highest rank among the available resources, and receiving a processing result of the data from the resource to which the data is allocated And a processed data collector for collecting the received processing results.

According to an embodiment of the present invention, resources selected according to a user's requirements and data characteristics are defined as rules based on knowledge, wherein the defined rules are newly added, modified, or deleted by a user or a system expert. There is an easy advantage, and by selecting the most suitable resources according to the user's budget and data characteristics, it can be processed within the deadline to increase the user satisfaction (QoS).

According to an embodiment of the present invention, an error that may occur during job processing may be evaluated by evaluating data to be processed by a user as a static property of a resource and defining and processing various scheduling techniques according to various dynamic changes as knowledge-based rules. Rapid response to loss of work can be provided to provide efficient job handling. In addition, by providing low work loss, high throughput, and short response time, high efficiency and stability in processing user tasks can be provided.

According to an embodiment of the present invention, a new resource is added by describing and managing resources using OWL (Web Ontology Language), an ontology standard language established by the World Wide Web Consortium (W3C), in a computing environment composed of heterogeneous resources. And it is easier to manage resources such as deletion and update, and can improve the scalability and compatibility of the system.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. However, the present invention is not limited or limited by the embodiments. Like reference numerals in the drawings denote like elements.

1 is a block diagram illustrating a configuration of a dynamic job scheduling system 100 based on semantic computing according to an embodiment of the present invention.

The semantic computing-based dynamic job scheduling system 100 according to an embodiment of the present invention distributes and allocates a job to a resource unit 30 geographically dispersed according to the user's 10 requirements and data characteristics.

Referring to FIG. 1, the semantic computing-based dynamic job scheduling system 100 according to an embodiment of the present invention includes a data input / analyzer 110, a job processing relay 120, a resource manager 130, and scheduling It includes a management unit 140.

The data input / analysis unit 110 receives the characteristics and requirements of the data together with the data to be processed from the user 10 and requests the work processing relay unit 120 to select a resource unit 30 to process the work. It plays a role.

The job processing relay unit 120 selects the resource unit 30 to be processed and performs a supervisory role for performing the entire work. The job processing relay unit 120 requests the information of the resource unit 30 to the resource manager 130 according to the request of the data input / analysis unit 110 and receives the resource information result message from the resource manager 130. After selecting 30), the scheduling management unit 140 requests a job to be processed.

The resource manager 130 collects and maintains heterogeneous resources. The resource manager 130 transmits the resource information collection request message to the resource unit 30, receives the resource information response message for the resource information collection request message from the resource unit 30, and receives the resource unit 30 through the received resource information response message. Save, update, and maintain information in The resource manager 130 transmits a resource information result message indicating the information of the resource unit 30 that is maintained when the resource information request is received from the job processing relay unit 120 to the job processing relay unit 120.

The scheduling manager 140 distributes and allocates a job to the resource unit 30 when a job processing request is received from the job processing relay 120. The scheduling manager 140 distributes and allocates a job to the resource unit 30 through inference, receives a result of the job from the resource unit 30 that processes the job, collects the result, and collects the result to the user 10. to provide.

Referring to Figure 2 will be described in detail the internal configuration and operation of the data input / analysis unit 110.

As shown in FIG. 2, the data input / analyzer 110 receives a user data receiver 210 that receives data to be processed from the user 10, and a data analyzer 230 that extracts characteristics of the input data. And a requirement analyzer 250 for extracting resource usage budget and deadline information from the input data.

The user data receiver 210 receives data to be processed by the user 10. Here, data is a generic term for work to be processed using computing resources such as images, text, calculation, and the like. Prior to transmitting the data to be processed, the user 10 inputs and transmits data characteristics such as the type and size of data, a budget, a deadline, and the like. Data from natural sciences and engineering fields, such as weather forecasts and genetic analysis, are so large that they cannot be transmitted at one time. In this case, the user data receiver 210 receives data from the user 10 and stores the data in the work queue until the data is allocated to the resource unit 30. Transmission to the scheduling manager 140.

The data analyzer 230 receives and analyzes characteristics of data to be processed by the user. Here, the characteristics of the data may include information on whether the user's data is a large amount of data requiring a large storage space or calculated data requiring fast calculation. The data analyzer 230 converts the analyzed characteristic of the data into an integer form and transmits the characteristic of the converted data to the work processing relay 120.

The requirement analyzer 250 receives and analyzes a user's requirement. Here, the user's requirements may include a budget that the user can spend using resources and a deadline for when the data to be processed should end. The requirement analyzer 250 converts the user's requirements into an integer form and transmits the converted requirements to the job processing relay unit 120.

An internal configuration and operation of the job processing relay 120 will be described in detail with reference to FIG. 3.

As shown in FIG. 3, the job processing relay unit 120 receives a resource selection request from the data input / analysis unit 110 to request available resources not currently used by the resource management unit 130. Knowledge-based inference of the resources suitable for the characteristics of the data and the requirements of the user based on the available resource receiver 330, the available resource receiver 330 receiving the information result message of the available resources from the resource manager 130, and the available resource list. It may include a resource selector 350 to select through, and a task manager 370 to perform a supervisory role for performing the overall task.

The available resource requester 310 requests a list of currently available resources from the resource manager 130. Here, the available resources mean idle computing resources in which the utilization rate of CPU and RAM is not used for a predetermined ratio among the resource units 30 geographically dispersed.

The available resource receiver 330 receives a list of currently available resources from the resource manager 130 through the network and transmits the received available resource list to the resource selector 350. Available resource receiver 330 is a list of resources available from the resource management unit 130 until the job is completed in case the job is not available, such as failure of the job processing or sudden power shutdown of the resource and disconnection of the network line Receive and keep it.

The resource selector 350 receives a list of available resources, data characteristics, and user requirements. In this case, the resource selector 350 selects a resource to be matched by defining a method of selecting a resource within a range not exceeding a budget that can be spent when using the resource among user requirements using various rules. For example, if the size of the data is large and there is enough time to complete the deadline, the auxiliary storage will choose a cheaper resource if it is large and not fast. In an embodiment of the present invention, a rule may be defined in the form of selecting a resource as shown in Table 1, and a rule may be additionally defined or modified and updated by a system administrator or a user.

Rule Description One DesktopComputer (? X) ^ CPUSpeed (? X,? Y) ^ [? Y> = 3000.0] ^ organization (? X,? Z) ^ [? Z = usrOrg] → Same agency high speed computer (? X) You want resources belonging to the same institution with faster processing speed. 2 Server (? X) ^ Bandwidth (? X,? Y) ^ [? Y> = 100.0] ^ StorageSize (? X,? Z) ^ [? Z> 5000.0] → Large High Speed Server (? X) In a high-speed network environment, a resource capable of processing a large amount of data is desired. 3 Server (? X) ^ CPUSpeed (? X,? Y) ^ [? Y> = 3400.0] ^ Cost (? X,? Z) ^ [? Z> 15] ^ StorageSize (? X,? K) ^ [? k <10] → Small data calculation server (? x) Even if resources are expensive, we want small resources with small data size and fast processing. 4 Notebook (? X) ^ location (? X,? Y) ^ [? Y = usrLoc] ^ Cost (? X,? Z) ^ [? Z <5] → Low cost notebook (? X) Regardless of the speed and size that needs to be handled, you want a low-cost resource in the same area as the user. 5 Cluster (? X) ^ numCPU (? X,? Y) ^ [? Y> = 5] ^ CPUSpeed (? X,? Z) ^ [z> 3000.0] ^ OSType (? X,? K) ^ [? K = WinFamily] ^ StorageSize (? X,? H) ^ [? H> = 1000.0] ^ Bandwidth (? X,? S) ^ [? S> = 100.0] → High Performance High Speed Cluster (? X) I want more than 5 available CPUs, my OS is Windows, and I want a resource that can handle a large amount of data in a high-speed network environment. …

The resource selector 350 transmits the resource list selected through the defined rule to the scheduling manager 140.

The task manager 370 supervises the execution of the entire task until the system is started and the processing is completed in order for the user's data to be processed. At this time, if the resource selected by the resource selector 350 cannot process a job normally due to sudden power shutdown or network disconnection while processing the user's data, the job manager 370 detects this and selects a resource reselection. Command the selector 350. In addition, when the job manager fails, the job manager 370 may instruct a job retransmission and reprocessing accordingly.

An internal configuration and operation of the resource manager 130 will be described in detail with reference to FIG. 4.

As shown in FIG. 4, the resource manager 130 includes a resource information requester 410 for requesting attribute information of a resource, a resource information receiver 430 for receiving a static attribute of a resource, and a job processing relay unit. The resource information transmitter 450 responsive to the available resource list according to the request of 120, the resource ontology manager 470 represented by OWL in order to manage the information and the relationship of the resource, and the attribute information value for the resource. Resource information storage 490 to store.

The resource information requester 410 periodically requests the resource unit 30 for a resource attribute value. Here, the attribute value of the resource is a static attribute that may include the OS type of the resource, the maximum clock speed of the CPU, the size of the auxiliary storage, and the like; It is a dynamic attribute that can change according to the current state of resources such as CPU utilization, main memory usage, and auxiliary memory available space. Table 2 shows the static attributes of resources, and Table 3 shows the dynamic attributes.

Property
Name Description Value type Minimum Value Unit ID Resource identifier Integer 0 --- OSType Type of OS String --- --- CPUArchType CPU Structure String --- --- organization Organization to which the resource belongs String --- --- CPUSpeed CPU maximum clock speed Float 700 Megahertz CPUNumber CPU count Integer One Cachesize size of cache memory Float 256 Kilo bytes RAMSize Size of main memory Float 32 Mega bytes Storagesize Size of auxiliary memory Float 10 Giga bytes Bandwidth Network bandwidth Float --- Mega BPS Cost Resource usage amount Integer --- --- … … … … …

Property
Name Description Value type Value range CPUStatus CPU usage Boolean true or false CPUUtilization CPU utilization Float 0-100 RAMUtilization Main memory utilization Float 0-100 RAMAvailableSize Main memory available space Float 0 ~ Available memory size Storutilization Auxiliary storage utilization Float 0-100 StorAvalableSize Auxiliary memory available space Float 0 ~ Available storage size NetStatus Status of the network Float RTT JobCompletion
Time Estimated time to complete the job stored in the buffer Float 0 ~ Estimated Time Rank Rank of Resource Integer 1 to 10 … … … …

In one embodiment of the present invention, the resource manager 130 may store and maintain resource information by adding, deleting, and modifying fields included in static and dynamic attributes of a resource by a system administrator. In addition, the resource manager 130 may automatically delete the corresponding resource information from the resource information storage 490 for a resource that does not respond even after a predetermined time or a specified number of times.

The resource information receiver 430 receives an information value about a static attribute of a resource from the resource unit 30. Here, the static attribute of the resource may be a response to a periodic request message by the resource information requester 410 or may be automatically transmitted by generating an event by detecting that the resource unit 30 has changed its information. . The resource information receiver 430 stores the received resource information in the resource information storage 490.

When the resource information transmitter 450 receives the available resource request message from the available resource requester 310 of the job processing relay 120, the resource information transmitter 450 searches for the currently available resources in the resource information storage 490 to search for resources that meet the conditions. After selecting and generating the resource list information, it is transmitted to the available resource receiver 330 of the job processing relay unit 120.

The resource ontology manager 470 classifies resources according to types of resources. The resource ontology manager 470 may classify resources using OWL, an ontology standard language established by the W3C.

FIG. 5 illustrates resources in the form of a class. As illustrated, a top-level class is Resource, and may be classified into a ComputingResource for processing calculation-oriented data and a DataResource for processing a large amount of data according to the type of resource. . And, when a new kind of resource is added, a new type of resource can be defined as a ResourceType # N class and included in the ontology so that the system can be easily extended and have a flexible structure.

The resource information store 490 stores attribute information values of resources as static attributes and dynamic attributes of the resources shown in Tables 2 and 3 in one of resource ontology forms as shown in FIG. 5.

An internal configuration and operation of the scheduling manager 140 will be described in detail with reference to FIG. 6.

As shown in FIG. 6, the scheduling manager 140 infers the resource state information receiver 610 which receives the state value of the resource from the resource unit 30, and selects the resource unit 30 based on the received state value of the resource. The engine unit 630 and the resource selected from the resource selector 350 of the job processing relay unit 120 are configured as resources for processing a job, and the inference engine unit 630 allocates and distributes the job to selected resources. A job divider 650 and a processed data collector 670 which collect the completed jobs from the resource unit 30 and make them into a file. Here, the inference engine unit 630 is a rule manager 631 that can add, delete, modify the rule by a system administrator or an expert, and searches for the fact that satisfies the conditions necessary for the execution of the defined rule and derives new facts. A rule resolver 633 and a query processor 635 to process the query among the refined resources.

The resource state information receiver 610 receives a dynamic state information value of a resource from the resource unit 30. Here, the dynamic state information value of the resource is a dynamic resource state shown in Table 3, and may include CPU usage, CPU usage, main memory usage, auxiliary storage usage, network status, and the like. The resource state information receiver 610 transmits the received resource state information to the inference engine unit 630 in the form of a data set.

The rule manager 631 of the inference engine unit 630 may further define, modify, and delete a rule to be performed by a system administrator or an expert. Here, the rule may combine the static and dynamic attributes of the resources shown in Tables 2 and 3 to create an individual with one of the resource classes shown in FIG. 5. At this point, system administrators or experts can add new rules, delete them, or change priorities to be applied according to the environment, such as the dynamic state of resources and network conditions. Rules for dynamic job scheduling according to an embodiment of the present invention are shown in Table 4.

Rule Domain rule One Resource (? X) ∧ StorAvalableSize (? X,? Y) ∧ [? Y> jobSize] → AvailableResource (? X) If the remaining disk space is larger than the job size to process, it is defined as available resource. 2 Resource (? X) ∧ CPUSpeed (? X,? Y) ∧ [? Y> 3.0] ∧ CacheSize (? X,? Z) ∧ (? Z> 4096) ∧ CPUUtilizationr (? X,? K) ∧ [? K <30] → HighSpeedLowWorkloadResource (? X) If the maximum clock speed of the CPU is 3.0GHz or more, the cache size is 4096 kilobytes or more, and the CPU utilization is less than 30%, it is defined as the low-speed high-speed processing CPU resource. 3 Resource (? X) ∧ CPUUtilization (? X,? Y) ∧ [? Y> 90] ∧ StorUtilization (? X,? Z) ∧ [? Z> 90] ∧ RAMUtilization (? X,? K) ∧ [? K > 90] ∧ NetStatus (? X,? N) ∧ [? N> threshold]
HasRank (? X, 1) If CPU usage, disk usage, and memory usage all exceed 90% and the network is congested, the rank of the resource is one. 4 Resource (? X) ∧ CPUUtilization (? X,? Y) ∧ [? Y> 40] ∧ [? Y <50] ∧ StorUtilization (? X,? Z) ∧ [? Z> 40] ∧ [? Z <50 ] ∧RAMUtilization (? X,? K) ∧ [? K> 40] ∧ [? K <50] ∧ NetStatus (? X,? N) ∧ [? N <threshold]
HasRank (? X, 5) If CPU usage, disk usage, and memory usage are 40-50%, and the network is in a normal state, the resource rank is 5. 5 Resource (? X) ∧ CPUUtilization (? X,? Y) ∧ [? Y <10] ∧ StorUtilization (? X,? Z) ∧ [? Z <10] ∧ RAMUtilization (? X,? K) ∧ [? K <10] ∧ NetStatus (? X,? N) ∧ [? N <threshold]
HasRank (? X, 10) If CPU utilization, disk utilization, and memory utilization are all less than 10% and the network is in a normal state, the rank of resources is 10. 6 Resource (? X) ∧ NetStatus (? X,? Y) ∧ [? Y <threshold] ∧ bandwidth (? X,? Z) ∧ [? Z> = 100] → HighSpeedNetworkResource (? X) If the network environment supports more than 100Mbps and has an RTT smaller than the threshold, it is defined as a high-speed network resource. … …

The rule interpreter 633 of the inference engine unit 630 searches for a fact that satisfies the conditions necessary to perform the rule shown in Table 4 and derives a new fact.

The query processor 635 of the inference engine unit 630 selects resources to distribute the current work by executing a query based on the new facts derived by the rule interpreter 633 to maximize the efficiency of the work processing. The query processor 635 transmits index information about the selected resource to the job distributor 650.

The job distributor 650 receives the resources selected from the resource selector 350 of the job processing relay 120 and the resources selected by the query processor 635. Here, the resources selected from the resource selector 350 are transmitted at the moment of receiving the user's job processing request message and are configured as distributed resources for job processing. If the resource list re-selected by the command of the job manager 370 of the job processing relay 120 may be reconfigured to the received resource as necessary. At this time, the job is distributed and allocated to the resource unit 30 finally selected by the query processor 635 through the inference engine unit 630.

The processed data collector 670 collects the processed data from the resource unit 30, creates a file, and returns the processed data to the user.

An internal configuration and operation of the resource unit 30 will be described in detail with reference to FIG. 7.

As shown in FIG. 7, the resource unit 30 includes an application processor 710 for receiving and processing a scheduled task, a resource state measurer 730 for measuring a static / dynamic state of a resource, and a measured resource. Resource status transmitter 750 for transmitting the status information value.

 The application processor 710 serves to receive and process a user's work distributed and allocated from the job distributor 650 of the scheduling manager 140, and transmits the processed job to the processed data collector 670.

When the resource state measurer 730 receives the resource information request message requested by the resource information requester 410 of the resource manager 130, the resource state measurer 730 measures static / dynamic attributes of the resource as shown in Tables 2 and 3. do.

The resource state transmitter 750 transmits the measured static / dynamic attribute information value of the resource to the resource information receiver 430 of the resource manager 130 and the resource state information receiver 610 of the scheduling manager 140.

Semantic computing-based dynamic job scheduling system according to an embodiment of the present invention by the above configuration is to ontology information and relationships of geographically dispersed resources to process the user's work in a computing environment consisting of heterogeneous resources By establishing and distributing and allocating data through scheduling defined by various rules, it is possible to select resources suitable for user requirements and data characteristics.

8 is a diagram illustrating a process of a dynamic job scheduling method based on semantic computing according to an embodiment of the present invention.

Referring to FIG. 8, a process of selecting a resource to be scheduled according to an embodiment of the present invention will be described in detail as follows. Here, the semantic computing-based dynamic job scheduling method according to an embodiment of the present invention may be executed by the semantic computing-based dynamic job scheduling system shown in FIG. 1, in particular, the scheduling manager 140 shown in FIG. 6.

The semantic computing-based dynamic job scheduling system according to an embodiment of the present invention receives a data size which is a characteristic of data analyzed from the data analyzer 230 of the data input / analyzer 110 (S801). The resource ontology OWL file is loaded (S802) and a resource object is generated based on the static and dynamic attributes of the data in a form that can be inferred (S803).

The semantic computing-based dynamic job scheduling system according to an embodiment of the present invention applies a rule defined in Table 4 to select a resource, where the size of the input data is available to the auxiliary storage device among the dynamic attributes of the resource. It is determined whether it is smaller than the space (S804), and resources that satisfy the condition of the step (S804) are added to the available resource list (S805).

The semantic computing-based dynamic job scheduling system according to an embodiment of the present invention calculates a task completion estimated time, CPU utilization, main memory utilization, and auxiliary storage utilization in order based on dynamic attributes of resources (S806 ~). S809). After determining the grade of the resource corresponding to the list of available resources by applying the inference rule defined in Table 4 using the dynamic attributes of the resources calculated in the above-mentioned process (S806 ~ S809) (S810), based on the grade of the resource In order to sort the available resources in ascending order (S811).

The semantic computing-based dynamic job scheduling system according to an embodiment of the present invention prioritizes resources having a small estimated time to complete a task when there are resources of the same grade in the list of available resources arranged in step S811 (S812). If a resource having the same estimated time to complete work among resources of the same class exists, priority is given to a resource having a good network state (S813).

The semantic computing-based dynamic job scheduling system according to an embodiment of the present invention finally selects the highest-ranked resource from the available resource list and sends the final selected resource information index to the job distributor 650 of the scheduling manager 140. It transmits (S814).

The semantic computing-based dynamic job scheduling system according to an embodiment of the present invention may further define, modify, update, and prioritize rules as shown in Table 4 by a system administrator or an expert in addition to the scheduling method according to the above process. By changing and applying, various rules can be applied according to the change of dynamic resource state.

As described above, the present invention has been described by specific embodiments such as specific components and the like. For those skilled in the art to which the present invention pertains, various modifications and variations are possible.

Therefore, the spirit of the present invention should not be limited to the described embodiments, and all of the equivalents or equivalents of the claims as well as the claims to be described later will belong to the scope of the present invention. .

1 is a diagram illustrating a configuration of a dynamic job scheduling system based on semantic computing according to an embodiment of the present invention.

FIG. 2 is a diagram illustrating an example of the data input / analysis unit 110 shown in FIG. 1.

FIG. 3 is a diagram showing an example of the job processing relay 120 shown in FIG. 1.

4 is a diagram illustrating an example of the resource manager 130 illustrated in FIG. 1.

FIG. 5 is a diagram illustrating an example of a resource ontology model for generating an entity required for inference.

FIG. 6 is a diagram illustrating an example of the scheduling manager 140 illustrated in FIG. 1.

FIG. 7 is a diagram illustrating an example of the resource unit 30 shown in FIG. 1.

 8 is a diagram for explaining an example of a dynamic scheduling process using attribute information of a resource based on a defined rule.

<Explanation of symbols for the main parts of the drawings>

110: data entry / analysis section

120: work processing relay

130: resource management unit

140: scheduling management unit

Claims (5)

A data input / analysis unit configured to receive data to be worked on, data characteristics corresponding to the data, and requirements for processing the data; A job processing relay unit which controls the overall operation so that the data is processed; A resource manager configured to collect and maintain attribute information of the resource through communication with at least one resource and to select available resources corresponding to the data characteristics and requirements; And, A scheduling manager that selects a resource to process the data from the selected available resources through communication with the resource, distributes the data to the selected resource, and provides a processing result for the distributed data. Including, The data characteristic includes at least one of a type of data and a size of data; The requirement includes at least one of a budget amount to be used to process the data, a deadline to process the data, The data input / analysis unit, A data analyzer for converting the input data characteristic into an integer form and transferring the data characteristic to the work processing relay; Requirement analyzer converting the input requirements into integer form and forwarding them to the work processing relay Semantic computing-based dynamic job scheduling system comprising a. delete The method of claim 1, The work processing relay unit, An available resource requester for requesting a resource list of the available resources to the resource manager; An available resource receiver for receiving a resource list of the available resources provided by the resource manager according to the request of the resource list; And a resource selector for selecting a resource corresponding to the data characteristic and the requirement among the available resources of the received resource list by using the inference rule defined for the data characteristic and the requirement. system. The method of claim 1, The resource management unit, A resource information requester for transmitting an attribute information request message to the resource; A resource information receiver for receiving the attribute information provided from the resource with respect to the request message; A resource ontology manager for expressing the received attribute information using an ontology standard language to manage the resource; A resource information repository for storing attribute information expressed in the ontology standard language in association with the resource; A resource information transmitter which selects available resources corresponding to the data characteristics and requirements under the control of the job processing relay unit and provides a resource list composed of the available resources to the job processing relay unit. Including, semantic computing-based dynamic job scheduling system. The method of claim 1, The scheduling manager, A resource state information receiver for receiving dynamic attribute information of the resource; An inference engine unit for determining the rank of the available resources according to the received dynamic attribute information using an inference rule defined for the attribute information of the resource; A job distributor for distributing and allocating the data to the highest ranking resource among the available resources; And a processed data collector for receiving the processing result of the data from the resource to which the data is allocated and collecting the received processing result. The dynamic attribute information, And at least one of CPU usage, CPU usage, main memory usage, auxiliary storage usage, and network status of the resource.

Images