KR20250069168A - Apparatus for distributed scheduling of analytics workflow and method thereof - Google Patents

Abstract

The present invention relates to a distributed scheduling device for an analysis workflow and a method thereof, comprising: a communication module, a memory, and a processor connected to the communication module and the memory, wherein the processor analyzes an analysis workflow received through the communication module to select at least one application task included in the analysis workflow, and distributes each application task to a different cloud in consideration of security information and movement cost of target data required for each selected application task, and causes the application task to be distributedly executed in the cloud to which each application task is distributed.

Description

{APPARATUS FOR DISTRIBUTED SCHEDULING OF ANALYTICS WORKFLOW AND METHOD THEREOF}

The present invention relates to a distributed scheduling device for an analysis workflow and a method thereof, and more specifically, to a distributed scheduling device for an analysis workflow and a method thereof that enable each application task of an analysis workflow to be processed in a distributed, parallel manner in a plurality of cloud environments.

Data generated by public or private entities are vast and managed by different organizations. When data management organizations provide data, data users download data online from the data management organization or visit the data management organization offline to access it. In this process, high network communication costs for moving large amounts of data or data security issues may arise.

The analysis workflow, which performs a series of tasks such as processing, handling, learning, analyzing, visualizing, and storing from a large amount of data, consists of application tasks that access the data and perform tasks. Each application task must access data from different data management organizations. However, if the analysis workflow is performed in a single location, network communication between the analysis workflow execution location and the data management organization is inevitable.

The background technology of the present invention is disclosed in Korean Patent Publication No. 10-2015-0017052.

The purpose of the present invention is to provide a distributed scheduling device and method for an analysis workflow that enables each application task of an analysis workflow to be processed in a distributed and parallel manner in multiple cloud environments.

A distributed scheduling device for an analysis workflow according to some embodiments of the present invention comprises a communication module, a memory, and a processor connected to the communication module and the memory, wherein the processor analyzes an analysis workflow received through the communication module to select at least one application task included in the analysis workflow, and distributes each application task to a different cloud by considering security information and movement cost of target data required for each of the selected application tasks, and causes the application tasks to be distributedly executed in the cloud to which each application task is distributed.

In some embodiments of the present invention, the processor is characterized by including a classification analyzer that analyzes at least one of a plurality of application tasks included in the analysis workflow, input data information of each application task, and required computing resources of each application task, a target data query unit that queries target data information including security information of target data required for each application task, a cloud resource query unit that queries available clouds based on the required computing resources of each application task, a scheduler that selects an optimal cloud for each application task by considering the security information and movement cost of the target data among the searched clouds, and an executor that requests execution of each application task on the selected optimal cloud.

In some embodiments of the present invention, the classification analyzer is characterized in that it determines whether the analysis workflow is a valid analysis workflow through classification analysis of the analysis workflow, and if it is a valid analysis workflow, it checks application tasks to be executed in the analysis workflow, and if it is an invalid analysis workflow, it terminates the analysis workflow and waits for the next analysis workflow.

In some embodiments of the present invention, the target data query device is characterized in that it queries a data management organization of the target data based on input data information of the application task to confirm whether the target data is an accessible target, and queries target data information including security information and access information of the target data.

In some embodiments of the present invention, the scheduler is characterized in that it extracts at least one cloud that can be accessed according to the security information of the target data from among the searched clouds, and selects a cloud having the lowest movement cost of the target data from among the extracted at least one cloud as an optimal cloud.

In some embodiments of the present invention, the movement cost of the target data is characterized in that it is a value calculated based on at least one of the size of the target data and the network distance.

In some embodiments of the present invention, the executor is characterized in that it monitors the status of each application task when each application task is executed normally.

A method for distributed scheduling of an analysis workflow according to some embodiments of the present invention is characterized by including a step in which a processor analyzes an analysis workflow received from a client and selects at least one application task included in the analysis workflow, a step in which the processor places each application task in a different cloud by considering security information and a movement cost of target data required for each of the selected application tasks, and a step in which the processor controls the application tasks to be distributedly executed in the cloud in which each application task is placed.

In some embodiments of the present invention, the step of selecting at least one application task is characterized in that the step of the processor includes the step of checking at least one of a plurality of application tasks included in the analysis workflow, input data information of each application task, and required computing resources of each application task, the step of the processor verifying each application task to determine whether the analysis workflow is a valid analysis workflow, and if it is a valid analysis workflow, the step of the processor checking application tasks to be executed in the analysis workflow.

In some embodiments of the present invention, the step of deploying each application task to a different cloud is characterized in that it includes the step of the processor retrieving target data information including security information of target data required for each application task, the step of the processor retrieving available clouds based on required computing resources of each application task, and the step of the processor selecting an optimal cloud for each application task from among the retrieved clouds, considering the security information and movement cost of the target data.

Some embodiments of the present invention are characterized in that, in the step of searching for the target data information, the processor searches a data management organization of the target data based on input data information of the application task to confirm whether the target data is an accessible target, and searches for target data information including security information and access information of the target data.

Some embodiments of the present invention are characterized in that, in the step of selecting the optimal cloud for each application task, the processor extracts at least one cloud that can be accessed according to the security information of the target data from among the searched clouds, and selects the cloud with the lowest movement cost from among the extracted at least one cloud as the optimal cloud.

In some embodiments of the present invention, the movement cost of the target data is characterized in that it is a value calculated based on at least one of the size of the target data and the network distance.

Some embodiments of the present invention are characterized in that, in the step of controlling the distributed execution, the processor monitors the status of each application task when each application task is executed normally.

A method for distributed scheduling of an analysis workflow according to some embodiments of the present invention is characterized by including: a step in which a processor analyzes an analysis workflow received from a client and selects at least one application task included in the analysis workflow; a step in which the processor searches for information on target data required for each of the selected application tasks; a step in which the processor searches for available clouds based on required computing resources of each application task; a step in which the processor places each application task on a different cloud among the searched clouds in consideration of security information and a movement cost of the target data; and a step in which the processor controls the application task to be distributedly executed in the cloud in which each application task is placed.

In some embodiments of the present invention, the step of selecting at least one application task is characterized in that the step of the processor includes the step of checking at least one of a plurality of application tasks included in the analysis workflow, input data information of each application task, and required computing resources of each application task, the step of the processor verifying each application task to determine whether the analysis workflow is a valid analysis workflow, and if it is a valid analysis workflow, the step of the processor checking application tasks to be executed in the analysis workflow.

Some embodiments of the present invention are characterized in that, in the step of searching for information on the target data, the processor searches a data management organization of the target data based on input data information of the application task to confirm whether the target data is an accessible target, and searches for target data information including security information and access information of the target data.

Some embodiments of the present invention are characterized in that, in the step of deploying each application task to a different cloud, the processor extracts at least one cloud that can be accessed according to the security information of the target data from among the searched clouds, and selects a cloud with the lowest movement cost from among the extracted at least one cloud as the optimal cloud.

In some embodiments of the present invention, the movement cost is characterized in that it is a value calculated based on at least one of the size of the target data and the network distance.

Some embodiments of the present invention are characterized in that, in the step of controlling the distributed execution, the processor monitors the status of each application task when each application task is executed normally.

The distributed scheduling device and method of an analysis workflow according to the present invention have the effect of enabling each application task of an analysis workflow to be processed in a distributed, parallel manner by distributing and executing each application task to different clouds while considering the security information and movement cost of target data of each application task included in the analysis workflow.

The distributed scheduling device and method of an analysis workflow according to the present invention have the effect of enabling efficient execution of an analysis workflow in various cloud environments by jointly utilizing data from different data management organizations.

FIG. 1 is a block diagram illustrating a distributed scheduling service providing system of an analysis workflow according to one embodiment of the present invention.
FIG. 2 is a block diagram showing a distributed scheduling device of an analysis workflow according to an embodiment of the present invention.
Figure 3 is a block diagram for explaining the function of the processor illustrated in Figure 2.
FIG. 4 is a flowchart illustrating a distributed scheduling method of an analysis workflow according to an embodiment of the present invention.

Hereinafter, an embodiment of a distributed scheduling device and method of an analysis workflow according to the present invention will be described with reference to the attached drawings. In this process, the thickness of lines and the size of components illustrated in the drawings may be exaggerated for clarity and convenience of explanation. In addition, the terms described below are terms defined in consideration of functions in the present invention, and may vary depending on the intention or custom of the user or operator. Therefore, the definitions of these terms should be made based on the contents throughout this specification.

Recently, with the development of cloud technology, individuals or companies can utilize the technology of cloud service providers rather than their own computing servers. Or, individuals or companies can build their own private clouds. Through the cloud environment, virtual computing resources can be created and allocated to users.

Data management organizations also build cloud environments to perform data-related tasks while providing data. This reduces network communication costs and security issues because data does not move outside the data management organization. However, if data is provided only from a single data management organization, data from other data management organizations cannot be used for tasks.

Therefore, a distributed scheduling technology is needed that enables the management agency of the target data required for each application task of the analysis workflow to distribute and execute application tasks in a trusted cloud environment.

Accordingly, the present invention proposes a distributed scheduling device and method for an analysis workflow for distributing and parallel processing of data managed at geographically different locations into a plurality of application tasks.

The present invention relates to a distributed scheduling device and method for an analysis workflow that enables analysis workflows supporting analysis of data managed at geographically different locations to be executed in a distributed manner by utilizing computing resources of multiple cloud environments.

Specifically, the present invention relates to a scheduling device and method for efficiently distributing each application task of an analysis workflow to a cloud environment in consideration of security information and movement costs of target data, thereby enabling distributed parallel processing.

FIG. 1 is a block diagram illustrating a distributed scheduling service providing system of an analysis workflow according to one embodiment of the present invention.

Referring to FIG. 1, a distributed scheduling service providing system of an analysis workflow according to one embodiment of the present invention includes a client (100), a distributed scheduling device of an analysis workflow (hereinafter referred to as a 'distributed scheduling device', 200), a plurality of data management organizations (300), and a cloud service providing organization (400).

A client (100) can transmit an analysis workflow service request to a distributed scheduling device (200). Here, the analysis workflow service request can include application tasks included in the analysis workflow.

A client (100) may include a wired or wireless communication terminal, device, system, etc. used by a user who wishes to receive a distributed scheduling service of an analysis workflow.

The distributed scheduling device (200) analyzes the analysis workflow service request received from the client (100) to select at least one application task included in the analysis workflow, and distributes each application task to different clouds by considering the security information and movement cost of target data required for each selected application task, and enables each application task to be distributedly executed in the cloud to which it is deployed. At this time, the distributed scheduling device (200) can inquire about target data required for each application task of the analysis workflow from the data management organization (300) and inquire about target data information such as security information and access information. In addition, the distributed scheduling device (200) can inquire about cloud resources of the cloud service provider (400).

A detailed description of this distributed scheduling device (200) is given with reference to FIG. 2.

Multiple data management organizations (300) are organizations that manage data required for application tasks of the analysis workflow, and each data management organization (300) can manage different data in different locations.

A cloud service provider (400) may refer to an institution that provides cloud resources, such as a cloud provider, a data management institution (300), or a private cloud.

The cloud service provider (400) is connected to the data management agency (300), and some of the data management agencies (300) can provide cloud services.

Meanwhile, although only one cloud service provider (400) is illustrated in FIG. 1, there may be multiple cloud service providers (400). FIG. 2 is a block diagram showing a distributed scheduling device of an analysis workflow according to an embodiment of the present invention, and FIG. 3 is a block diagram explaining the function of the processor illustrated in FIG. 2.

Referring to FIG. 2, a distributed scheduling device (200) of an analysis workflow according to one embodiment of the present invention includes a communication module (210), a memory (220), and a processor (230).

The communication module (210) is configured to communicate with a client (100), a data management organization (300), a cloud service provider (400), etc. via a communication network, and can transmit and receive various information such as an analysis workflow service request. At this time, the communication module (210) can be implemented in various forms such as a short-range communication module, a wireless communication module, a mobile communication module, and a wired communication module.

The memory (220) is a configuration that stores data related to the operation of the distributed scheduling device (200). In particular, the memory (220) may store applications (programs or applets) that allow each application task of the analysis workflow to be efficiently distributed to a cloud environment in consideration of the security and movement cost of the target data and processed in a distributed parallel manner, and the stored information may be selectively selected by the processor (230) as needed. That is, the memory (220) stores various types of data generated during the execution of the operating system or application (program or applet) for driving the distributed scheduling device (200). At this time, the memory (220) refers to a nonvolatile storage device that continuously maintains stored information even when power is not supplied and a volatile storage device that requires power to maintain the stored information.

The processor (230) may be implemented as at least one of an Application Specific Integrated Circuit (ASIC), a Digital Signal Processor (DSP), a Programmable Logic Devices (PLD), a Field Programmable Gate Arrays (FPGAs), a Central Processing Unit (CPU), a microcontroller, and/or a microprocessor.

The processor (230) analyzes the analysis workflow received through the communication module (210) to select at least one application task included in the analysis workflow, and distributes each application task to different clouds by considering the security information and movement cost of target data required for each selected application task, and allows each application task to be distributedly executed in the cloud to which it is distributed.

This processor (230) includes a syntax analyzer (231), a target data queryer (232), a cloud resource queryer (233), a scheduler (234), and an executor (235), as illustrated in FIG. 3.

The parser (231) can analyze and verify the analysis workflow received from the client (100).

The analysis workflow can be defined in various structured formats such as JSON, YAML, etc., and includes a number of ordered application tasks that input and process data. Accordingly, the parser (231) can understand the input data information of the application tasks included in the analysis workflow, and verify the order and execution rules between the application tasks.

Specifically, the parser (231) can analyze multiple application tasks included in the analysis workflow, input data information of each application task, required computing resources of each application task, etc. Here, the input data information includes input values, data, etc., and the input values are values required for processing and include values such as numbers and characters, and the data can include access information (including server addresses, names, paths, etc.).

In addition, the syntax analyzer (231) can determine whether the analysis workflow is a valid analysis workflow by analyzing the distinction of the analysis workflow. At this time, the syntax analyzer (231) can determine whether the analysis workflow is a valid analysis workflow by checking whether the analysis workflow is defined according to a predefined format, the order and execution rules between application tasks, etc. For example, the syntax analyzer (231) can determine whether the analysis workflow is a valid analysis workflow by checking whether the relationships between nodes in the analysis workflow are normally well connected.

If the analysis workflow is invalid, the parser (231) can terminate the current analysis workflow and wait for the next analysis workflow.

If a valid analysis workflow is present, the parser (231) can select application tasks to execute in the analysis workflow.

The target data query unit (232) can query information on target data required for each application task selected by the syntax analyzer (231). That is, the target data query unit (232) can query target data information including security information and access information of the target data. Here, the security information means a security level and can include information on whether it cannot be leaked or can be leaked. If the security information cannot be leaked, the target data can only be operated (managed) by the data management organization (300).

Since an application task may require zero or more target data, the target data query device (232) can query the security information (security scope), access information (management location), etc. of the target data. Based on the input data information of the application task, the target data query device (232) can query the data management organization (300) of the target data to confirm whether it is an accessible target, and can query target data information including the security information and access information of the target data.

The cloud resource query device (233) can query available clouds based on the required computing resources of each application task. At this time, the cloud resource query device (233) can query accessible cloud resources such as a cloud service provider (400), a data management agency (300), and a private cloud.

Since an application task may require computing resources such as CPU and memory when executed, the cloud resource query unit (233) may query clouds that can use the required computing resources. Here, the required computing resources may include CPU processing power, memory capacity, etc.

The scheduler (234) can place each application task in a different cloud by considering the security information and movement cost of target data required for each application task among the clouds searched by the cloud resource searcher (233). Here, the movement cost is the cost of moving the target data to the cloud, and can be calculated based on at least one of the size of the target data and the network distance, and the network distance can be calculated using the hop count, etc.

The scheduler (234) can extract at least one cloud that can be accessed according to the security information of the target data from among the searched clouds, and select the cloud with the lowest cost of moving the target data from among the extracted at least one cloud as the optimal cloud.

For example, if the security level is high (i.e., if the target data cannot be leaked (moved)), the scheduler (234) can select the optimal cloud among the clouds of the data management organization (300). If the target data cannot be moved and the cloud of the data management organization (300) cannot be used, the application task cannot be executed and may remain in an error state. In addition, if the security level is low (i.e., if the target data can be leaked (moved)), the scheduler (234) can select the cloud with the lowest movement cost among the searched clouds.

For example, a case will be described where the first target data of 100 GB in size is located in data management organization 1, the second target data of 200 GB in size is located in data management organization 2, both the first target data and the second target data are movable, and the first target data and the second target data must be processed in one place. In this case, the scheduler (234) determines that the movement cost of the first target data with a small data size is less than the movement cost of the second target data, and thus can move the first target data.

The executor (235) can request execution of each application task on the optimal cloud selected by the scheduler (234) and monitor the status of the application task.

That is, when the executor (235) is ready to execute the application tasks, it can distribute and execute each application task to the selected optimal cloud. In addition, when the application tasks are executed normally, the executor (235) can monitor the status of the application tasks, check whether the corresponding application task has been terminated, and find and execute the next executable application tasks. When all the application tasks have been terminated, the executor (235) can terminate the analysis workflow.

The distributed scheduling device (200) configured as described above efficiently distributes each application task of the analysis workflow to a cloud environment while considering the security information and movement cost of the target data, so that it can be processed in a distributed parallel manner.

FIG. 4 is a flowchart illustrating a distributed scheduling method of an analysis workflow according to an embodiment of the present invention.

Referring to FIG. 4, when an analysis workflow is received from a client (100) (S402), the distributed scheduling device (200) analyzes the analysis workflow (S404) and determines whether it is a valid analysis workflow (S406). That is, the distributed scheduling device (200) can parse the analysis workflow to check a plurality of application tasks included in the analysis workflow, input data information of each application task, required computing resources of each application task, etc. Then, the distributed scheduling device (200) can verify the order and execution rules between the application tasks to determine whether the analysis workflow is a valid analysis workflow. At this time, the distributed scheduling device (200) can check whether it is a valid analysis workflow that the scheduler (234) can understand, and if it is not a valid analysis workflow, it can terminate the corresponding analysis workflow and wait for the next analysis workflow.

When step S406 is performed, the distributed scheduling device (200) checks whether there are executable application tasks in the analysis workflow (S408) and selects an executable application task (S410).

When step S410 is performed, the distributed scheduling device (200) searches for information on target data required for the selected application task (S412). At this time, the distributed scheduling device (200) can search for target data required for the selected application task from the data management organization (300) to confirm whether the target data is accessible. In addition, the distributed scheduling device (200) can search for target data information including security information (security scope), access information (management location), etc. of each target data.

When step S412 is performed, the distributed scheduling device (200) searches for available clouds based on the required computing resources of each application task (S414). Since the application task may require computing resources such as CPU and memory when executed, the distributed scheduling device (200) can search for clouds that can use the required computing resources.

When step S414 is performed, the distributed scheduling device (200) selects an optimal cloud for each application task from among the searched clouds, considering the security information and movement cost of target data required for each application task (S416). That is, the distributed scheduling device (200) can extract at least one cloud that can be accessed according to the security information of the target data from among the searched clouds, and select a cloud with the lowest movement cost of the target data from among the extracted at least one cloud as the optimal cloud.

When step S416 is performed, the distributed scheduling device (200) distributes and executes the application task to the selected optimal cloud (S418). That is, when the distributed scheduling device (200) completes preparation for execution of the application task, it can distribute and execute the application task to the selected optimal cloud.

When step S418 is performed, the distributed scheduling device (200) monitors the status of the application task and determines whether the application task has been terminated (S420).

As a result of the judgment at step S420, when the application task is terminated, the distributed scheduling device (200) performs step S408.

The distributed scheduling device and method of an analysis workflow according to the present invention have the effect of enabling each application task of an analysis workflow to be processed in a distributed, parallel manner by distributing and executing each application task to different clouds while considering the security information and movement cost of target data of each application task included in the analysis workflow.

In addition, the distributed scheduling device and method of the analysis workflow according to the present invention have the effect of enabling efficient execution of the analysis workflow in various cloud environments by jointly utilizing data from different data management organizations.

Although the present invention has been described with reference to the embodiments shown in the drawings, these are merely exemplary, and those skilled in the art will understand that various modifications and equivalent other embodiments are possible therefrom. Accordingly, the technical protection scope of the present invention should be defined by the following patent claims.

100 : Client
200: Distributed scheduling device for analysis workflows
210 : Communication module
220 : Memory
230 : Processor
231 : Parser
232: Target Data Lookup
233: Cloud Resource Query
234 : Scheduler
235 : Executor
300 : Data Management Agency
400: Cloud Service Provider

Claims (20)

Communication module;
memory; and
Including a processor connected to the above communication module and memory,
The above processor,
A distributed scheduling device for an analysis workflow, characterized in that it analyzes an analysis workflow received through the communication module to select at least one application task included in the analysis workflow, distributes each application task to a different cloud by considering the security information and movement cost of target data required for each of the selected application tasks, and causes the application task to be executed in a distributed manner in the cloud to which each application task is distributed.
In the first paragraph,
The above processor,
A segmentation analyzer that analyzes at least one of a plurality of application tasks included in the above analysis workflow, input data information of each application task, and required computing resources of each application task;
A target data query device that queries target data information including security information of target data required for each application task;
A cloud resource query engine that queries available clouds based on the computing resources required for each application task;
A scheduler that selects the optimal cloud for each application task by considering the security information and movement cost of the target data among the clouds searched above; and
A distributed scheduling device for an analysis workflow, characterized in that it includes an executor that requests execution of each application task on the above-mentioned selected optimal cloud.
In the first paragraph,
The above-mentioned classification analyzer is,
A distributed scheduling device for an analysis workflow, characterized in that it determines whether the analysis workflow is a valid analysis workflow through a distinction analysis of the analysis workflow, and if it is a valid analysis workflow, it checks application tasks to be executed in the analysis workflow, and if it is an invalid analysis workflow, it terminates the analysis workflow and waits for the next analysis workflow.
In the first paragraph,
The above target data query engine is,
A distributed scheduling device for an analysis workflow, characterized in that it checks whether the target data is an accessible target by inquiring about the data management agency of the target data based on the input data information of the above application task, and inquiring about target data information including security information and access information of the target data.
In the first paragraph,
The above scheduler,
A distributed scheduling device for an analysis workflow, characterized in that it extracts at least one cloud that can be accessed according to the security information of the target data from among the clouds searched for above, and selects the cloud with the lowest movement cost of the target data from among the at least one extracted cloud as the optimal cloud.
In paragraph 5,
The cost of moving the above target data is,
A distributed scheduling device for an analysis workflow, characterized in that the value is calculated based on at least one of the size of the target data and the network distance.
In the first paragraph,
The above executor,
A distributed scheduling device for an analysis workflow, characterized by monitoring the status of each application task when each application task is executed normally.
A step in which a processor analyzes an analysis workflow received from a client and selects at least one application task included in the analysis workflow;
The step of the above processor placing each application task in a different cloud by considering the security information and movement cost of target data required for each of the above selected application tasks; and
The above processor controls the application tasks to be distributed and executed in the cloud where each application task is deployed.
A method for distributed scheduling of an analysis workflow, comprising:
In Article 8,
The step of selecting at least one application task is:
A step in which the above processor checks at least one of a plurality of application tasks included in the analysis workflow, input data information of each application task, and required computing resources of each application task;
The step of the above processor verifying each application task to determine whether the analysis workflow is a valid analysis workflow; and
A method for distributed scheduling of an analysis workflow, characterized in that it comprises a step of checking application tasks to be executed by the processor in the analysis workflow if the analysis workflow is a valid analysis workflow.
In Article 8,
The steps for deploying each of the above applications to different clouds are:
A step in which the above processor searches for target data information including security information of target data required for each application task;
The above processor searches for available clouds based on the required computing resources of each application task; and
A distributed scheduling method for an analysis workflow, characterized in that the processor comprises a step of selecting an optimal cloud for each application task by considering security information and movement cost of the target data among the searched clouds.
In Article 10,
In the step of searching the above target data information,
A distributed scheduling method for an analysis workflow, characterized in that the processor checks whether the target data is an accessible target by inquiring about the data management organization of the target data based on the input data information of the application task, and inquires about target data information including security information and access information of the target data.
In Article 11,
In the step of selecting the optimal cloud for each of the above applications,
A distributed scheduling method for an analysis workflow, characterized in that the processor extracts at least one cloud that can be accessed according to the security information of the target data from among the searched clouds, and selects a cloud with the lowest movement cost from among the extracted at least one cloud as an optimal cloud.
In Article 12,
The cost of moving the above target data is,
A distributed scheduling method for an analysis workflow, characterized in that the value is calculated based on at least one of the size of the target data and the network distance.
In Article 8,
In the step of controlling the above distributed execution,
A distributed scheduling method for an analysis workflow, characterized in that the above processor monitors the status of each application task when each application task is executed normally.
A step in which a processor analyzes an analysis workflow received from a client and selects at least one application task included in the analysis workflow;
A step in which the above processor searches for information on target data required for each of the selected application tasks;
The step of the above processor searching for available clouds based on the required computing resources of each application task;
The step of the processor placing each application task in a different cloud among the searched clouds, considering the security information and movement cost of the target data; and
The above processor controls the application tasks to be distributed and executed in the cloud where each application task is deployed.
A method for distributed scheduling of an analysis workflow, comprising:
In Article 15,
The step of selecting at least one application task is:
A step in which the above processor checks at least one of a plurality of application tasks included in the analysis workflow, input data information of each application task, and required computing resources of each application task;
The step of the above processor verifying each application task to determine whether the analysis workflow is a valid analysis workflow; and
A method for distributed scheduling of an analysis workflow, characterized in that it comprises a step of checking application tasks to be executed by the processor in the analysis workflow if the analysis workflow is a valid analysis workflow.
In Article 15,
In the step of searching for information on the above target data,
A distributed scheduling method for an analysis workflow, characterized in that the processor checks whether the target data is an accessible target by inquiring about the data management organization of the target data based on the input data information of the application task, and inquires about target data information including security information and access information of the target data.
In Article 15,
In the step of deploying each of the above applications to different clouds,
A distributed scheduling method for an analysis workflow, characterized in that the processor extracts at least one cloud that can be accessed according to the security information of the target data from among the searched clouds, and selects a cloud with the lowest movement cost from among the extracted at least one cloud as an optimal cloud.
In Article 18,
The above moving costs are,
A distributed scheduling method for an analysis workflow, characterized in that the value is calculated based on at least one of the size of the target data and the network distance.
In Article 15,
In the step of controlling the above distributed execution,
A distributed scheduling method for an analysis workflow, characterized in that the above processor monitors the status of each application task when each application task is executed normally.