A cloud resource secheduling method based on quantum and artificial intelligence analysis and a secheduling devce implementing the same

Abstract

A scheduling method according to one embodiment of the present invention, in a scheduling method implemented by a computing device, may include: receiving request information, which is information related to a scheduling request of a cloud server; analyzing the request information using a predetermined analysis method to calculate a quantum dominance index; selecting one of a quantum analysis method and an artificial intelligence analysis method based on the quantum dominance index; and calculating response information, which is a response to a scheduling request, using the selected analysis method.

Description

Technology Field

[0001] The present invention relates to a method for scheduling cloud resource management using quantum and artificial intelligence analysis and an apparatus for implementing the same. Background Technology

[0003] The cloud market is growing rapidly and is being widely utilized in various forms across many industries. In this context, the efficient distribution and allocation of cloud resources is a critical factor in providing efficient cloud services. To this end, scheduling resources in a cloud computing environment is essential.

[0004] Resource scheduling methods currently used in most cloud platforms are AI-based and operate primarily based on heuristic or meta-heuristic algorithms. For example, Kubernetes' default scheduler uses priority-based filtering and scoring mechanisms, while Apache Mesos utilizes the Dominant Resource Fulfillment algorithm.

[0005] Existing methods relying solely on artificial intelligence have the following limitations. First, it is difficult to find optimal solutions for NP-Hard problems such as bin packing, job scheduling, and multi-constrained resource allocation. Furthermore, as the scale of the problem increases, computational complexity grows exponentially; consequently, in environments requiring real-time scheduling, there is a problem where one is forced to rely only on solutions that are realistically acceptable rather than the optimal solution. Prior art literature

[0007] Korean Registered Patent Publication No. 10-2822125 (Registered on June 13, 2025) The problem to be solved

[0008] The present invention aims to solve the aforementioned problems by utilizing not only artificial intelligence analysis but also quantum analysis methods to provide a cloud resource scheduling method and a scheduling device for implementing the same. means of solving the problem

[0010] A scheduling method according to one embodiment of the present invention, in a scheduling method implemented by a computing device, may include: receiving request information, which is information related to a scheduling request of a cloud server; analyzing the request information using a predetermined analysis method to calculate a quantum dominance index; selecting one of a quantum analysis method and an artificial intelligence analysis method based on the quantum dominance index; and calculating response information, which is a response to a scheduling request, using the selected analysis method.

[0011] In addition, the step of calculating the quantum superiority index may be a step of calculating a plurality of evaluation indicators that reflect the desired resources and constraints of the cloud server based on the request information, and calculating the quantum superiority index using the evaluation indicators.

[0012] In addition, the above request information may include a first request information, which is information about the desired computing resources for each task, which is an independent resource unit of the cloud, and a second request information, which is information about the constraints on the scheduling of each task.

[0013] In addition, the above-determined analysis method may be a method of analyzing by utilizing the above-determined fairy information to represent a vertex representing a task and an edge-shaped graph expressing the relationship of tasks based on constraints, and then utilizing a first evaluation metric that considers the degree of connectivity between neighboring vertices and a second evaluation metric that considers the degree of density among all vertices.

[0014] In addition, the first evaluation indicator may be an indicator calculated as the ratio of the number of actual connections between neighboring vertices to the maximum number of possible connections between neighboring vertices.

[0015] In addition, the second evaluation indicator may be an indicator calculated based on the weights of the second-1 evaluation indicator, which considers the degree of connectivity between all vertices, and the second-2 evaluation indicator, which considers the degree of grouping of all vertices.

[0016] In addition, the above-mentioned predetermined analysis method may be a method of analysis that further utilizes a third evaluation metric representing the number of vertices and a fourth evaluation metric considering the time dependency constraint and the time window overlap ratio.

[0017] In addition, the step of calculating the quantum superiority index may be a step of calculating the quantum superiority index by inputting the first to fourth evaluation indicators into a pre-trained artificial intelligence model.

[0018] Additionally, the step of selecting one of the above quantum analysis method and the above artificial intelligence analysis method may be a step of selecting the above quantum analysis method when the above quantum superiority index is greater than or equal to a first reference value, and selecting the above artificial intelligence analysis method when the above quantum superiority index is less than a second reference value.

[0019] Additionally, if the quantum superiority index is less than the first reference value and greater than or equal to the second reference value, the method may further include the step of selecting a hybrid analysis method that utilizes both the quantum analysis method and the artificial intelligence analysis method, without selecting one of the quantum analysis method and the artificial intelligence analysis method.

[0020] In addition, the above-determined analysis method is a method of analysis that utilizes an edge-shaped graph representing the relationship between tasks based on vertices signifying tasks and constraints using the above-determined request information, and the above-determined hybrid analysis method may be an analysis method that utilizes the above-determined quantum analysis method or the above-determined artificial intelligence analysis method for each group after dividing all vertices into a minimum number of groups through a partitioning algorithm, depending on the degree of coupling between vertices within the group.

[0021] In addition, the artificial intelligence analysis method may be a method for producing the response information by sequentially utilizing a simulated annealing technique and a genetic algorithm.

[0022] In addition, the step of calculating the response information may be a step of calculating the response information including cloud resource placement or allocation using a selected analysis method.

[0023] In addition, the step of calculating the above response information may be a step of calculating multiple candidate response information using a selected analysis method and a step of calculating a final response information among the candidate response information based on a reliability index calculated through a predetermined reliability analysis method.

[0024] In addition, the above-determined reliability analysis method may be a method of analysis that considers a first reliability index, which is an index for determining whether a constraint condition is satisfied; a second reliability index, which is an index for determining whether a predetermined objective function is satisfied; and a third reliability index, which is an index for determining by inferring past performance.

[0025] A scheduling device according to some embodiments of the present invention comprises: a processor; a memory for loading a computer program executed by the processor; and a storage for storing the computer program, wherein the computer program may include: an operation of receiving request information, which is information related to a scheduling request of a cloud server; an operation of calculating a quantum advantage index by analyzing the request information using a predetermined analysis method; an operation of selecting one of a quantum analysis method and an artificial intelligence analysis method based on the quantum advantage index; and an operation of calculating response information, which is a response to the scheduling request, using the selected analysis method.

[0026] A computer program according to some embodiments of the present invention may be combined with a computer device which is hardware and stored in a computer-readable storage medium to execute a scheduling method. Effects of the invention

[0028] The cloud resource scheduling method based on quantum and artificial intelligence analysis according to the present invention and the scheduling device implementing the same can maximize resource utilization.

[0029] In addition, scheduling optimization can be automated.

[0030] In addition, it can effectively handle large-scale constraints.

[0031] In addition, unnecessary quantum operations can be prevented.

[0032] In addition, it can effectively handle real-time workloads.

[0033] In addition, scalability and compatibility can be enhanced.

[0034] However, the effects of the present invention are not limited to the effects described above, and unmentioned effects will be clearly understood by those skilled in the art from this specification and the attached drawings. Brief explanation of the drawing

[0036] FIG. 1 is a relationship diagram of a cloud resource scheduling device according to an embodiment of the present invention. FIG. 2 is a flowchart of a cloud resource scheduling method according to an embodiment of the present invention. FIG. 3 is a diagram illustrating vertices and edges in a cloud resource scheduling method according to an embodiment of the present invention. FIG. 4 is a diagram illustrating a second evaluation metric in a cloud resource scheduling method according to an embodiment of the present invention. FIG. 5 is a diagram illustrating the process of calculating a quantum dominance index in a cloud resource scheduling method according to an embodiment of the present invention. FIG. 6 is a diagram illustrating the process of selecting an analysis method in a cloud resource scheduling method according to an embodiment of the present invention. FIG. 7 is a diagram illustrating a partitioning algorithm in a cloud resource scheduling method according to an embodiment of the present invention. FIG. 8 is a diagram illustrating an artificial intelligence analysis method and a quantum analysis method in a cloud resource scheduling method according to an embodiment of the present invention. FIG. 9 is a detailed flowchart of the step of calculating response information in a cloud resource scheduling method according to an embodiment of the present invention. FIG. 10 is a configuration diagram of a cloud resource scheduling device according to an embodiment of the present invention. Specific details for implementing the invention

[0037] Specific embodiments of the present invention will be described in detail below with reference to the drawings. However, the concept of the present invention is not limited to the presented embodiments, and those skilled in the art who understand the concept of the present invention may easily propose other inventions that are inferior or other embodiments included within the scope of the concept of the present invention by adding, changing, or deleting other components within the same scope of the concept, and such are also to be considered to be included within the scope of the concept of the present invention.

[0038] The terms first, second, A, B, etc., as used in this specification and claims may be used to describe various components, but said components should not be limited by said terms. These terms are used solely for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, the first component may be named the second component, and similarly, the second component may be named the first component. The term "and / or" includes a combination of a plurality of related described items or any of a plurality of related described items.

[0039] The terms used in this specification and claims are used merely to describe specific embodiments and are not intended to limit the invention. The singular expression includes the plural expression unless the context clearly indicates otherwise. In this application, terms such as "comprising" or "having" should be understood as not precluding the existence or addition of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification.

[0040] Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as generally understood by those skilled in the art to which the present invention pertains. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and should not be interpreted in an ideal or overly formal sense unless explicitly defined in this application.

[0041] Furthermore, each component, process, procedure, or method included in each embodiment of the present invention may be shared within a scope that is not technically contradictory. Additionally, components with the same function within the scope of the same concept appearing in the drawings of each embodiment are described using the same reference numerals.

[0043] FIG. 1 is a diagram of a cloud resource scheduling device according to one embodiment of the present invention.

[0045] Figure 1 shows that a cloud resource scheduling device according to one embodiment of the present invention can be network-connected via wired or wireless means to enable information communication with a physical server and a manager device that provide cloud services.

[0047] The cloud resource scheduling device (100) may be a computing device.

[0048] The cloud resource scheduling device (100) can calculate the operation scheduling of future virtual servers based on operation information, which is information generated when the cloud server (V10) is operated from the physical server (200), and request information received from the manager device.

[0049] The cloud resource scheduling device (100) receives a scheduling request from the manager device (300) and / or the cloud server (V10), and can calculate a plan (scheduling) for the resources allocated from the physical server to operate the cloud server, whether the cloud server is operating, and the placement of the cloud server.

[0050] A specific method for the operation of the cloud resource scheduling device (100) will be described later.

[0052] The physical server (200) may be a hardware server on which the cloud server (V10) operates.

[0053] In addition, the physical server (200) itself can be a cloud server. For example, in the case of a Kubernetes-based cloud service, the physical server itself can be a node, and a virtual server within the physical server can be a node.

[0054] The server referred to in the present invention may include input / output devices, a processor, memory, storage, an interface, and a bus, as other components for performing the server environment. The server may include any type of device.

[0055] For example, a server is a digital device and can be a digital device equipped with a processor and memory that has computational capabilities, such as a laptop computer, notebook computer, desktop computer, web pad, or mobile phone.

[0056] For example, the server may be a web server. However, it is not limited to this, and the type of server can be varied to a level obvious to a person of ordinary technical expertise.

[0057] The physical server can communicate information with the cloud resource scheduling device or the manager device.

[0059] The cloud server (V10) may be a server that provides cloud services.

[0060] For example, the cloud server (V10) can be a virtual or physical server.

[0061] For example, when the virtualization is based on a hypervisor installed on a physical server (200), the cloud server may be a virtual machine.

[0062] For example, if the cloud server (V10) is a Kubernetes-based node, the cloud server may be a physical server or a server virtualized from a physical server.

[0063] Operation information generated while the cloud server is operating can be transmitted to the cloud resource scheduling device through the physical server.

[0064] For example, operational information may include information on response time, throughput, availability, CPU utilization, memory utilization, disk I / O, and network bandwidth.

[0065] Hereinafter, in the present invention, a node may be a concept that includes both a hypervisor-based virtual machine and a Kubernetes-based node.

[0066] The cloud server may transmit availability information, which is information about available resources, to the cloud resource scheduling device. Here, the availability information may include both information about the maximum resources available when the cloud service is not provided and information about resources that can be additionally provided when the cloud service is provided.

[0067] Cloud resources refer to computing resources that a cloud server can provide to a task, and may include, for example, compute (CPU, RAM), storage (HHD, SSD), network, database, etc., but the present invention is not limited thereto.

[0069] The manager device (300) may refer to a computing device used by a manager.

[0070] The manager device (300) and the cloud resource scheduling device (100) may be physically one device, but the present invention is not limited thereto, and the manager device may be a device physically separated from the cloud resource scheduling device.

[0071] The manager device (300) is connected to the physical server for information communication and can monitor the status of the cloud server operating on the physical server.

[0072] The manager device (300) can transmit request information including a scheduling request to the cloud resource scheduling device (100).

[0073] The manager device (300) can receive and display the response information produced by the cloud resource scheduling device (100).

[0074] The administrator can manage and operate the cloud server (V10) and / or the cloud resource scheduling device (100) using the administrator device (300).

[0075] The computing device mentioned in the present invention may refer to a device capable of processing information processing operations.

[0076] For example, computing devices may include desktop computers, laptops, smartphones, PDAs (Personal Digital Assistants), PMPs (Portable Multimedia Players), portable terminals, mobile devices and / or smart TVs, servers, etc.

[0077] The computing device is described in detail below.

[0078] Below, a cloud resource scheduling method implemented by a cloud resource scheduling device will be described in detail.

[0080] FIG. 2 is a flowchart of a cloud resource scheduling method according to an embodiment of the present invention.

[0082] Referring to FIG. 2, a cloud resource scheduling method according to one embodiment of the present invention may include the steps of: receiving request information, which is information related to a scheduling request of a cloud server, implemented by a computing device (S100); analyzing the request information using a predetermined analysis method to calculate a quantum dominance index (S200); selecting one of a quantum analysis method and an artificial intelligence analysis method based on the quantum dominance index (S300); and calculating response information, which is a response to a scheduling request, using the selected analysis method (S500).

[0083] Additionally, the cloud resource scheduling method may further include a step (S400) of selecting both the quantum analysis method and the artificial intelligence analysis method without selecting one of the quantum analysis method and the artificial intelligence analysis method when the quantum dominance index is less than the first reference value and greater than the second reference value.

[0084] In addition, the cloud resource scheduling method may further include a step (S600) of displaying the calculated response information.

[0086] In step (S100), the cloud resource scheduling device can receive request information, which is information related to a scheduling request of a cloud server.

[0087] Request information can be received as data in a predetermined format through a predetermined communication method. For example, the predetermined communication method may be a REST API, but the present invention is not limited thereto. For example, the predetermined format may be a JSON format, but the present invention is not limited thereto.

[0088] The request information may include first request information, which is information regarding desired computing resources for each task, an independent resource unit of the cloud, and second request information, which is information regarding constraints on the scheduling of each task. Additionally, the request information may further include time information, which is information regarding the period during which scheduling is calculated. Furthermore, the request information may also include a request from an administrator to schedule cloud resources on a cloud server.

[0089] In the present invention, a task may be defined as an independent unit in which cloud resources are configured. For example, in the case of a hypervisor-based cloud, the task may be a virtual machine. For example, in the case of a Kubernetes-based cloud, the task may be a Pod. In this case, the cloud server may function as a Node. However, this is not limited thereto, and specific examples of tasks may change whenever the form of the cloud service changes.

[0090] The first request information is information regarding the computer resources desired for each task, and may include requirements for a processor (CPU), memory, storage, and network bandwidth for each task. For example, the processor requirement may be 2 vCPUs, the memory requirement may be 2 GB, the storage requirement may be 20 GB, and the network bandwidth requirement may be 5 Gbps, but the present invention is not limited thereto.

[0091] The second request information may be information regarding constraints on the scheduling of each task. For example, the second request information may include Node Affinity, Anti-Affinity, Label Selector, and Taint & Toleration.

[0092] Specifically, node affinity may refer to a restriction that ensures a Pod is placed on a specific node or a restriction that ensures a virtual machine is placed on a specific physical server. Anti-affinity may refer to a restriction that ensures a Pod is placed away from a specific node or a virtual machine is placed away from a specific physical server. A label selector may refer to a restriction on labels assigned to a Pod or a restriction on labels assigned to a virtual machine. Taints and tolerations may refer to restrictions that allow only Pods or virtual machines that have been pre-allowed on a node or physical server. Specific examples of the second request information can be varied in many ways to the extent that they are obvious to a person skilled in the art.

[0093] Time information is information about the period during which scheduling is calculated, and may be constraint information that allows calculations to be performed only during the time information period. For example, the time information may be 10 days in the future based on the time of analysis, but the present invention is not limited thereto.

[0095] FIG. 3 is a diagram illustrating vertices and edges in a cloud resource scheduling method according to an embodiment of the present invention.

[0097] The cloud resource scheduling device can calculate a quantum advantage index by analyzing the request information using a predetermined analysis method (S200).

[0098] In step (S200), the step of calculating the quantum superiority index may be a step of calculating a plurality of evaluation indicators that reflect the desired resources and constraints of the cloud server based on the request information, and calculating the quantum superiority index using the evaluation indicators.

[0099] The predetermined analysis method may be a method of calculating a plurality of evaluation indicators that reflect the resources and constraints of the desired cloud server based on the above request information, and calculating the quantum advantage index using the said evaluation indicators.

[0100] Referring to FIG. 3, the predetermined analysis method may be a method of analyzing by using the fairy information (the first request information and the second request information) to represent the relationship of tasks in the form of an edge graph based on vertices representing tasks and constraints.

[0101] The cloud resource scheduling device uses the first request information and the second request information among the request information to form a graph ( It can be converted into ). Here, the first request information is a vertex ( )(10) as the second request information is the edge( It can be expressed as )(20). The first request information is specified for each vertex, and the task can be represented as a single circular vertex on the graph. The second request information is a constraint imposed on the task (vertex), and connections between vertices can be defined according to the constraint. For example, vertices placed on the same node or physical server may be connected to each other by edges, or vertices that exchange data and process computations together may be connected to each other by edges. Additionally, vertices that are constrained from being placed on the same node or physical server may not be connected to each other by edges. However, the present invention is not limited thereto, and the method of connecting vertices and edges can be varied in various ways that are obvious to a person skilled in the art.

[0102] The predetermined analysis method may be a method of analysis utilizing a first evaluation metric that considers the degree of connectivity between adjacent vertices and a second evaluation metric that considers the degree of density among all vertices. Additionally, the predetermined analysis method may be a method of analysis utilizing, in addition to the first and second evaluation metric, a third evaluation metric representing the number of vertices and a fourth evaluation metric that considers time dependency constraints and the time window overlap ratio.

[0104] The first evaluation indicator may be an indicator calculated as the ratio of the actual number of connections between neighboring vertices to the maximum possible number of connections between neighboring vertices. The first evaluation indicator may be an indicator for evaluating the degree of connectivity between neighboring vertices for any vertex. The first evaluation indicator calculates the local constraint density for any vertex and may be calculated for each vertex. Specifically, the method for calculating the first evaluation indicator may follow 'Equation 1'.

[0106] (Mathematical Formula 1)

[0108] Here, n is the number of neighboring vertices, and E is the number of constraints between neighboring vertices.

[0109] Specifically, represents the maximum number of edges that can be connected between adjacent vertices. For example, based on an arbitrary vertex (N11), the adjacent vertices may be vertices (N12, N13, N14) that are actually connected to the arbitrary vertex by an edge. Therefore, the maximum number of edges that can be connected to four vertices may be six. However, referring to FIG. 3, the number of edges actually connected to four vertices may be three. As a result, the first evaluation index for an arbitrary vertex may be 0.5. In this way, the first evaluation index can be calculated for all vertices.

[0111] FIG. 4 is a diagram illustrating a second evaluation metric in a cloud resource scheduling method according to an embodiment of the present invention.

[0113] The second evaluation indicator may be an indicator that considers the degree of density among all vertices. That is, the second evaluation indicator may be an indicator for calculating the bond strength between vertices. The second evaluation indicator may be an indicator calculated based on the weights of the second-1 evaluation indicator, which considers the degree of connectivity among all vertices, and the second-2 evaluation indicator, which considers the degree of grouping of all vertices.

[0114] Evaluation metric 2-1 is an indicator that considers the degree of connectivity between all vertices and may refer to the average constraint density obtained by averaging the local constraint densities for each vertex.

[0115] Evaluation metric 2-2 is an indicator that considers the degree of grouping of all vertices and may represent modularity. The said modularity has a value between -1 and 1, and a higher number indicates a higher degree of grouping of the graph. For example, if it is between 0.3 and 0.7, it may be determined that the graph is grouped to some extent. The method for calculating evaluation metric 2-2 may follow 'Equation 2'.

[0117] (Mathematical Formula 2)

[0119] Here, m represents the total number of vertices, S represents the entire group of the partitioned graph, and s represents a single group within the entire group, , can mean the degree (number of connected edges) of the i-th and j-th vertices. Also, can represent the presence or absence of an edge between the i-th and j-th vertices within the graph inside group s. In this case, if an edge exists, It has '1', and if there is no edge It can have '0'. Here, the graph group can be arbitrarily designated, and the graph group with the highest modularity can be designated through repeated calculations.

[0120] For example, the modularity of the graph in Fig. 4 can be calculated as follows and have a value of 0.36. In this case, the groups can be divided into the first group with three vertices (1, 2, 3) and the second group with the remaining vertices (4, 5, 6).

[0122]

[0123]

[0124]

[0126] Below, a detailed description of modularity may be omitted to the extent that it is obvious to a person skilled in the art.

[0127] For example, the second evaluation indicator may be calculated by adding the value obtained by multiplying the above-mentioned 2-1 evaluation indicator by the first weight and the value obtained by multiplying the above-mentioned 2-2 evaluation indicator by the second weight. For example, the first weight may be 0.6 and the second weight may be 0.4, but the present invention is not limited thereto, and the first weight and the second weight can be varied in various ways that are obvious to a person skilled in the art.

[0129] The third evaluation metric may be an indicator representing the number of peaks.

[0130] For example, referring to FIG. 3, the number of vertices may be 15, but the present invention is not limited thereto.

[0132] The fourth evaluation metric may be a metric that considers time-dependent constraints and the time window overlap ratio. Specifically, the fourth evaluation metric may refer to the complexity between the number of time-dependent constraints and the time window overlap ratio.

[0133] Time-dependent constraints may be constraints that change over time. For example, they may refer to conditions that change over time, such as a constraint that prevents placement at a specific node only at a specific time. However, the present invention is not limited thereto. As the number of time-dependent constraints on edges connecting two vertices increases, the complexity may further increase.

[0134] The time window overlap ratio can refer to the ratio of the number of vertices (branches) reachable from a single vertex in a task process involving movement from one vertex to another. For vertices operating within the same timeframe, the complexity of the time window overlap ratio can increase if there are many paths available to move from one vertex to another. Furthermore, the complexity of the time window overlap ratio can increase further as the number of movements (depth) from one vertex to another increases.

[0135] The fourth evaluation metric, complexity (C) (computational amount), can be determined according to 'Equation 3' below.

[0137] (Mathematical Formula 3)

[0139] Here, is a function that calculates complexity related to the number of constraints, and is a function that calculates the time window overlap ratio, and d can represent the depth of the path.

[0140] However, this is not limited to this, and the specific formula for calculating complexity can be varied in many ways to the extent that it is obvious to a person skilled in the art.

[0142] FIG. 5 is a diagram illustrating the process of calculating a quantum dominance index in a cloud resource scheduling method according to an embodiment of the present invention.

[0144] A cloud resource scheduling device can calculate a quantum advantage index by inputting the first to fourth evaluation indicators into a pre-trained artificial intelligence model (S200).

[0145] Referring to FIG. 5, in step (200), the first evaluation indicator (P11), the second evaluation indicator (P12), the third evaluation indicator (P13), and the fourth evaluation indicator (P14), which are calculated based on the request information (R10), can be input into a prediction model (M10), which is a pre-trained artificial intelligence model. The prediction model (M10) can calculate a quantum superiority index (Y10) consisting of numbers from 0 to 1.

[0146] Here, the prediction model (M10) may be a model produced through supervised learning based on past first to fourth evaluation indicators and training data labeled with quantum dominance indices corresponding to those evaluation indicators. For example, supervised learning may utilize algorithms such as linear regression, logistic regression, K-nearest neighbors, and random forest, but the present invention is not limited thereto, and specific supervised learning algorithms can be modified in various ways that are obvious to a person skilled in the art.

[0147] For example, the larger the first evaluation metric, the larger the quantum superiority index can be, and the smaller the first evaluation metric, the smaller the quantum superiority index can be. Additionally, the larger the second evaluation metric, the larger the quantum superiority index can be, and the smaller the second evaluation metric, the smaller the quantum superiority index can be. Furthermore, the larger the third evaluation metric, the larger the quantum superiority index can be, and the smaller the third evaluation metric, the smaller the quantum superiority index can be. Also, the larger the fourth evaluation metric, the larger the quantum superiority index can be, and the smaller the fourth evaluation metric, the smaller the quantum superiority index can be.

[0148] In other words, as the number of vertices in the entire graph increases, as there are many edges connecting the vertices, and as the complexity increases, the likelihood of it being an NP-Hard problem increases, so the quantum advantage index can increase.

[0150] FIG. 6 is a diagram illustrating the process of selecting an analysis method in a cloud resource scheduling method according to an embodiment of the present invention.

[0152] Referring to FIG. 6, the cloud resource scheduling device can select one of a quantum analysis method and an artificial intelligence analysis method based on a quantum advantage index (S300).

[0153] In step (S300), the step of selecting one of the quantum analysis method and the artificial intelligence analysis method may be a step of selecting the quantum analysis method when the quantum advantage index is greater than or equal to a first reference value (S320), and selecting the artificial intelligence analysis method when the quantum advantage index is less than a second reference value (S310).

[0154] When the quantum dominance index is greater than or equal to the first threshold value, the quantum analysis method may be more effective than the artificial intelligence analysis method because the relationships between the vertices are very entangled. Additionally, when the quantum dominance index is less than the second threshold value, the quantum analysis method may be more effective because the relationships between the vertices are not entangled.

[0155] For example, the first reference value may be 0.7, but the present invention is not limited thereto, and the specific numerical value of the first reference value can be varied in various ways that are obvious to a person skilled in the art.

[0156] For example, the second reference value may be 0.3, but the present invention is not limited thereto, and the specific numerical value of the second reference value can be varied in various ways that are obvious to a person skilled in the art.

[0158] In another embodiment of the present invention, the cloud resource scheduling device may select one of a plurality of predefined scheduling analysis methods according to the quantum dominance index.

[0159] Here, a plurality of predetermined scheduling analysis methods may include quantum analysis methods, artificial intelligence analysis methods, analysis methods based on predetermined regulations, etc. Specific descriptions of the quantum analysis methods and artificial intelligence analysis methods may be omitted to the extent that they overlap with the contents described in this specification.

[0160] An analysis method based on predetermined regulations may be a method for calculating response information using predicted resource demand, the available information, and the request information according to a predetermined algorithm. For example, the analysis method based on predetermined regulations may be a method for calculating response information that satisfies the request information and the predicted resource demand, while the present invention is not limited thereto, and specific examples of the analysis method based on predetermined regulations can be varied in various ways that are obvious to a person skilled in the art.

[0162] When the quantum advantage index is less than the first reference value and greater than the second reference value, the cloud resource scheduling device may select a hybrid analysis method that utilizes both the quantum analysis method and the artificial intelligence analysis method without selecting one of the quantum analysis method and the artificial intelligence analysis method (S400).

[0163] In step (S400), if the quantum superiority index calculated by the prediction model is less than the first reference value and greater than or equal to the second reference value, it may not be clear which analysis method is clearly superior. Consequently, if the quantum superiority index calculated by the prediction model is less than the first reference value and greater than or equal to the second reference value, the hybrid analysis method utilizing both the quantum analysis method and the artificial intelligence analysis method may be selected.

[0164] In step (S400), the hybrid analysis method may be an analysis method that divides all vertices into a minimum number of groups through a partitioning algorithm, and then utilizes the quantum analysis method or the artificial intelligence analysis method for each group according to the degree of coupling between vertices within the group.

[0166] FIG. 7 is a diagram illustrating a partitioning algorithm in a cloud resource scheduling method according to an embodiment of the present invention.

[0168] In step (S400), the entire graph may be divided into several groups by a partitioning algorithm. The partitioning algorithm may be an algorithm that divides the entire graph into several groups as a minimum partition. For example, the partitioning algorithm may be the METIS algorithm, but the present invention is not limited thereto.

[0169] For example, referring to FIG. 7, in the overall graph, three groups can be formed as a first division (C11), a second division (C12), and a third division (C13).

[0171] In step (S400), the hybrid analysis method may be a method of calculating a quantum superiority index for each group separated by a partitioning algorithm, and selecting one of the quantum analysis method or the artificial intelligence analysis method based on the quantum superiority index calculated for each group. To this end, a first to fourth evaluation indicator is calculated for each group, and the calculated first to fourth evaluation indicators are input into a prediction model so that a quantum superiority index can be calculated for each group.

[0172] The hybrid analysis method may be a method in which groups whose quantum dominance index is greater than or equal to a predetermined set value are analyzed by a quantum analysis method. Additionally, the hybrid analysis method may be a method in which groups whose quantum dominance index is less than a predetermined set value are analyzed by an artificial intelligence analysis method. In this way, the hybrid analysis method can efficiently produce a solution (response information) to a scheduling request by implementing different analysis methods according to the quantum dominance index for each group.

[0173] For example, the above-mentioned predetermined setting value may be 0.5, but the present invention is not limited thereto, and the specific value of the above-mentioned predetermined setting value can be varied in various ways to the extent that it is obvious to a person skilled in the art.

[0174] To explain in detail, in the hybrid analysis method, a quantum dominance index can be calculated for each group according to the method described above. Here, when calculating the quantum dominance index, edges connecting groups may be excluded. If the quantum dominance index of any group is greater than or equal to the aforementioned predetermined setting value, a quantum analysis method may be selected for that group to generate preliminary response information. Furthermore, if the quantum dominance index of any group is less than the aforementioned predetermined setting value, an artificial intelligence analysis method may be selected for that group to generate preliminary response information. All possible combinations of the preliminary response information generated for each group can be combined to generate (candidate) response information.

[0175] For example, referring to FIG. 7, the entire graph can be divided into a first group (G11), a second group (G12), and a third group (G13) by a partitioning algorithm. In addition, the quantum superiority index of the first group (G11) may be 0.7, the quantum superiority index of the second group (G12) may be 0.8, and the quantum superiority index of the third group (G13) may be 0.3. In this case, preliminary response information can be calculated for the first group (G11) and the second group (G12) through a quantum analysis method. Additionally, preliminary response information can be calculated for the third group (G13) through an artificial intelligence analysis method. It can be assumed that for the first group (G11), the first preliminary response information is produced, for the second group (G12), the second preliminary response information and the third preliminary response information are produced, and for the third group (G13), the fourth preliminary response information and the fifth preliminary response information are produced. At this time, the (candidate) response information may be produced as the first (candidate) response information combined with the first preliminary response information, the second preliminary response information, and the fourth preliminary response information, the second (candidate) response information combined with the first preliminary response information, the second preliminary response information, and the fifth preliminary response information, the third (candidate) response information combined with the first preliminary response information, the third preliminary response information, and the fourth preliminary response information, and the fourth (candidate) response information combined with the first preliminary response information, the third preliminary response information, and the fifth preliminary response information.

[0177] Hereinafter, the quantum analysis method and artificial intelligence analysis method implemented in the present invention will be described in detail.

[0179] FIG. 8 is a diagram illustrating an artificial intelligence analysis method and a quantum analysis method in a cloud resource scheduling method according to an embodiment of the present invention.

[0181] The artificial intelligence analysis method may be a method that produces the above response information by sequentially utilizing simulated annealing techniques and genetic algorithms.

[0182] Specifically, referring to FIG. 8, the artificial intelligence analysis method may be an analysis method that predicts the resource demand required for a future cloud server using a resource prediction model (T110) based on operation information received from the cloud server, calculates a solution to a schedule request using a heuristic engine (T120) based on the predicted resource demand, the availability information, and the request information, and calculates (candidate) response information by excluding solutions from the solutions that do not satisfy the constraint condition (T130). Here, the first request information may refer to the minimum cloud resources that each task must possess, and the predicted resource demand may refer to the maximum cloud resources that each task must possess.

[0183] The resource prediction model (T110) can predict the resources of the cloud server for a future time by analyzing the workload pattern of a predetermined past time. To this end, operational information received from the cloud server may be input into the resource prediction model (T110). In addition, the resource prediction model (T110) can predict the resources required in the future for each task. For example, the predetermined past time may be 24 hours and the future time may be 6 hours, but the present invention is not limited thereto.

[0184] For example, the resource prediction model (T110) may be a transformer-based prediction engine, but the present invention is not limited thereto, and specific examples of the resource prediction model (T110) can be modified in various ways to the extent that it is obvious to a person skilled in the art. For example, the resource prediction model (T110) may utilize a known resource prediction model.

[0185] The heuristic engine (T120) can search for a first candidate solution based on predetermined rules, select a second candidate solution from the first candidate solution using a simulated annealing technique, and then generate a final solution from the second candidate solution using a genetic algorithm.

[0186] The predetermined rules may refer to rules predefined by an administrator. For example, the aforementioned predetermined rules may include a rule to assign tasks with predicted resource demand exceeding a certain threshold to cloud servers meeting a predetermined standard, a rule to assign tasks with a specific label to a single cloud server, or a rule to assign tasks so that the cloud server operates at a rate below a predetermined percentage of the maximum available resources. However, the rules are not limited thereto, and specific examples of the aforementioned predetermined rules can be varied in various ways that are obvious to a person skilled in the art.

[0187] The heuristic engine (T120) can select a second candidate solution from among the first candidate solutions by utilizing a simulation annealing technique. The simulation annealing technique may be a probabilistic algorithm designed to find the 'Global Optimum' within a complex structure. The simulation annealing technique may be an algorithm that finds the correct answer by gradually modifying a single solution. Here, when the simulation annealing technique is utilized, a solution among the first candidate solutions that has a high probability of satisfying the predicted resource demand, the availability information, and the request information can be selected as the second candidate solution. A detailed description of the simulation annealing technique may be omitted within the scope of known technology.

[0188] The heuristic engine (T120) can generate a final solution among the second candidate solutions by utilizing a genetic algorithm. A genetic algorithm may refer to an algorithm that crosses multiple solutions with one another and induces mutations to find a superior solution. That is, the genetic algorithm may be an algorithm that crosses the second candidate solutions with one another, induces changes (mutations) to generate new candidate solutions, and calculates a final solution by checking whether the candidate solutions can satisfy the requested information, the available information, and the predicted resource demand. A detailed description of the genetic algorithm may be omitted within the scope of known technology.

[0189] Among the final answers produced through the heuristic engine (T120), answers that do not satisfy the constraint (T130) can be excluded to produce (candidate) response information. Here, the constraint may be the second request information mentioned above, but the present invention is not limited thereto and may be a constraint pre-specified to the administrator.

[0190] However, not limited thereto, the artificial intelligence analysis method is a method for calculating an optimal solution for predicted resource demand by utilizing an artificial intelligence model, and may be calculated using other known algorithms, and a detailed explanation thereof may be omitted from the scope of the known technology.

[0192] The quantum analysis method (T200) may be a method for generating response information by utilizing a quantum computing process. The quantum analysis method may be a method for generating response information by utilizing request information, availability information, and operation information by utilizing a quantum computing process. To this end, a cloud resource scheduling device may transmit data required by an external quantum computer and request that response information be generated through the quantum analysis method, and then receive the request information, but the present invention is not limited thereto.

[0193] For example, to explain quantum analysis methods, scheduling decisions are binary variables It can be converted by encoding. Here, If it is a task, it means that a task (i) receiving support from desired computer resources is assigned to a cloud server (j), and If so, it may mean that the task (i) receiving the desired computer resources is not assigned to the cloud server (j). Next, the objective function ( ) can be constructed in the form of Quadratic Unconstrained Binary Optimization (QUBO) (T210) through binary variables. Here, Q can be a symmetric matrix encoded with the objective function and constraints. The diagonal elements of Q may list the objective encoding information, and the non-diagonal elements of Q may list the constraint encoding information.

[0194] And, the objective function (T210), which is a QUBO problem, can be transformed into an Ising model (T220). That is, the binary variable x of the QUBO problem is a spin variable By mapping to, the Ising Hamiltonian ( ) can be constructed. Here, Coefficients and The coefficients can be derived from the QUBO matrix Q. That is, The formula of, and, The formula can be utilized.

[0195] The converted Ising Hamiltonian can be further processed into a form suitable for a quantum circuit or a quantum annealer. In the case of a gate-based quantum computer (T230), the Ising Hamiltonian is converted into a QAOA circuit (T240), and in the case of a quantum annealer (T250), the Ising Hamiltonian can be mapped to a chimera or pegasus graph structure (T260). Then, response information for a scheduling request can be produced through the QAOA circuit (T240). Alternatively, response information for a scheduling request can be produced through a mapping process based on a chimera or pegasus graph structure.

[0196] The quantum algorithm QAOA(T240) can initialize all qubits to a uniform superposition state. Here, QAOA can consist of p layers. A circuit structure of p layers can alternately apply two operations to find the correct answer. One of the two operations is the cost Hamiltonian ( ) and another operation is the mixer Hamiltonian( It can be. By repeating the two operations p times, the quantum state can be induced toward the direction with a higher probability of being the correct answer. That is, by executing a variational parameter optimization loop and using a classical optimization algorithm (COBYLA), the parameters ( , While adjusting ) the expected value of the cost function( ...can be minimized. In each iteration, the quantum circuit can be executed and measurements performed. Then, the final execution is performed with optimal parameters and the results are collected, and the most frequently observed bit string can be produced as response information. The above-described process is a general description of the quantum algorithm QAOA, and more specific details may be omitted within the scope of known technology.

[0197] A method for generating response information through a mapping process based on a chimera or pegasus graph structure (T240) is a known quantum computing process method, and a detailed description thereof may be omitted within the scope of known technology.

[0198] Here, the QAOA-based quantum analysis method can yield a single solution, while the chimera or pegasus graph structure quantum analysis method can yield multiple solutions.

[0200] A cloud resource scheduling device can produce response information that is a response to a scheduling request using a selected analysis method (S500).

[0201] Here, the response information may include information on specific plans regarding how tasks are deployed to cloud servers, how cloud resources are allocated to tasks, and at what point each task operates for future time.

[0203] FIG. 9 is a detailed flowchart of the step of calculating response information in a cloud resource scheduling method according to one embodiment of the present invention.

[0205] Referring to FIG. 9, the step of calculating response information (S500) may be a step of calculating multiple candidate response information using a selected analysis method (S510) and a step of calculating a final response information among the candidate response information based on a reliability index calculated through a predetermined reliability analysis method (S520).

[0206] Here, the response information may include the placement or allocation of cloud resources using a selected analysis method, but the present invention is not limited thereto. For example, the response information may further include whether a task is running, whether a cloud server is running, etc., but the present invention is not limited thereto.

[0208] The cloud resource scheduling device can generate multiple candidate response information using a selected analysis method (S510).

[0209] In step (S510), if the quantum advantage index is greater than or equal to the first threshold value, a quantum analysis method is selected, and at least one candidate response information can be produced through the quantum analysis method. If the quantum advantage index is less than the second threshold value, an artificial intelligence analysis method is selected, and at least one candidate response information can be produced through the artificial intelligence analysis method. If the quantum advantage index is less than the first threshold value but greater than or equal to the second threshold value, a hybrid analysis method is selected, and at least one candidate response information can be produced through the hybrid analysis method.

[0211] The cloud resource scheduling device can calculate the final response information among candidate response information based on the reliability index calculated through a predetermined reliability analysis method (S520).

[0212] In step (S520), the predetermined reliability analysis method may be a method of analysis that considers a first reliability index, which is an index for determining whether a constraint is satisfied; a second reliability index, which is an index for determining whether a predetermined objective function is satisfied; and a third reliability index, which is an index for determining whether past performance is inferred. The predetermined reliability analysis method may be a method of calculating a final reliability index that considers the first reliability index, the second reliability index, and the third reliability index for each candidate response information.

[0213] The final confidence index is the above-mentioned first confidence index ( ) the first weight ( The value multiplied by ), the above second confidence index ( ) is the second weight ( The value multiplied by ) and the above third confidence index ( ) third weight( It can be calculated as a value multiplied by ) and can follow the mathematical formula 4 below.

[0215] (Mathematical Formula 4)

[0217] The first, second, and third weights can be dynamically adjusted by the reinforcement learning agent. Additionally, the combination of weights may be determined according to the policy network of the cloud server, but the present invention is not limited thereto.

[0219] The first confidence index may be an index regarding how well the candidate response information meets the constraints of the second request information of the request information. The higher the degree to which the candidate response information satisfies the constraints, the higher the first confidence index may be. Conversely, the lower the degree to which the candidate response information satisfies the constraints, the lower the first confidence index may be. For example, the first confidence index may be calculated based on the ratio of satisfied constraints among all constraints. For example, the first confidence index may be calculated by converting it into a number between 0 and 1, but the present invention is not limited thereto.

[0221] The second reliability index may be a reliability index that determines how well candidate response information aligns with a predetermined purpose. For example, the aforementioned predetermined purpose may be the purpose of minimizing resource usage and the purpose of maximizing performance, but the present invention is not limited thereto. The second reliability index may be calculated as the average value of the 2-1 reliability index and the 2-2 reliability index described below.

[0222] For example, in order to calculate the second reliability index, when tasks are operated on a cloud server according to candidate response information, the average resource utilization rate of all cloud servers is calculated, and the higher the average resource utilization rate, the lower the second-1 reliability index, which is the reliability index for the purpose of minimizing resource usage. Conversely, the lower the average resource utilization rate, the higher the second-1 reliability index. Here, the second-1 reliability index can be calculated by converting it into a number between 0 and 1. Here, the standard for the average resource utilization rate may be based on the processor (CPU), but the present invention is not limited thereto. In addition, an open-source simulation program may be utilized to calculate the average resource utilization rate.

[0223] In addition, for the second reliability index to be calculated, when tasks are operated on a cloud server according to candidate response information, whether there is a violation of a Service Level Agreement (SLA) is verified, and the more frequent the SLA violations, the lower the reliability index for the purpose of performance maximization, which is the second-2 reliability index. Conversely, the fewer the SLA violations, the higher the second-2 reliability index may be. Here, the second-2 reliability index can be calculated by converting it into a number between 0 and 1. The SLA may be predetermined by an administrator, but the present invention is not limited thereto.

[0224] Here, the specific method for calculating the 2-1 reliability index and the 2-2 reliability index can be varied in various ways at a level obvious to a person skilled in the art, and likewise, the specific method for calculating the 2 reliability index can be varied in various ways at a level obvious to a person skilled in the art.

[0226] The third confidence index may be a confidence index calculated based on a performance model trained on past training data. The third confidence index may be calculated based on the service performance derived by inputting candidate response information into the performance model.

[0227] The performance model may be a model trained through deep learning with labeled information regarding tasks previously deployed on cloud servers—such as task deployment, resources allocated to the task, and the type of task—information regarding the physical server where the task is deployed, and operational information of the cloud server. In other words, the performance model may be a model that simulates how the cloud server operates depending on the situation of the task. This may be trained internally, or a publicly available simulation model may be supplied from an external server and utilized. A detailed description of the performance model may be omitted within the scope of the publicly available technology.

[0228] Here, deep learning may utilize, but is not limited to, a back propagation algorithm, which is an algorithm that updates the weights of a neural network using labeled data from the output layer. Furthermore, since deep neural networks and back propagation algorithms are as previously known, a detailed description thereof may be omitted.

[0229] The performance of the cloud server utilized in the third reliability index may be at least one of response time and throughput, but the present invention is not limited thereto. The lower the average response time of the cloud server and the higher the average throughput, the higher the third reliability index may be. Conversely, the higher the average response time of the cloud server and the lower the average throughput, the lower the third reliability index may be. The third reliability index may be converted into a number between 0 and 1 and calculated, but the present invention is not limited thereto.

[0231] In step (S520), the final confidence indices calculated for each candidate response information are compared with each other, and the candidate response information having the highest final confidence indices can be calculated as the final response information.

[0232] If there are candidate response information with the same final confidence index, the candidate response information having the largest value among the following: the absolute difference between the first criterion value and the quantum advantage index in the case of the quantum advantage analysis method; the absolute difference between the second criterion value and the quantum advantage index in the case of the artificial intelligence analysis method; and the absolute difference between the quantum advantage index of the entire graph and the first criterion value or the quantum advantage index of the entire graph and the second criterion value in the case of the hybrid analysis method, may be selected as the final response information. This may be because the nature of the graph is distinct, and the candidate response information that follows the analysis method appropriate to that nature may be the best response information.

[0234] The cloud resource scheduling device can display the calculated (final) response information through a display unit or transmit it to the manager device (S600).

[0235] Through this, the administrator can examine the final response information displayed via the display unit or administrator device and establish future management strategies for the cloud server.

[0237] FIG. 10 is a configuration diagram of a cloud resource scheduling device according to one embodiment of the present invention.

[0239] Referring to FIG. 10, a cloud resource scheduling device (100) according to one embodiment of the present invention is a computing device and may include an input / output device (101, I / O), a processor (102, processor), a memory (103, memory), a storage (104, storage), an interface (105, interface), and a bus (106, bus).

[0240] In addition, the cloud resource scheduling device (100) may further include a display unit (not shown) that displays a user interface.

[0241] An input / output device (101), a processor (102), a memory (103), a storage (104), an interface (105), a display unit (not shown) and / or a communication unit (not shown) can be combined and connected to each other via a bus (106). Here, the bus (106) may correspond to a path through which data travels.

[0242] Specifically, the input / output device (101) may include at least one of a keypad, a keyboard, a touchscreen, and a display device.

[0243] Additionally, the processor (102) may include at least one of a CPU (Central Processing Unit), MPU (Micro Processor Unit), MCU (Micro Controller Unit), GPU (Graphic Processing Unit), microprocessor, digital signal processor, microcontroller, application processor (AP), and logic elements capable of performing similar functions.

[0244] The memory (103) can load data and / or programs, etc. At this time, the memory (103) is an operational memory for enhancing the operation of the processor (101) and may include high-speed DRAM and / or SRAM, etc. The memory (103) may include one or more volatile memory devices such as DDR SDRAM (Double Data Rate Static DRAM) and SDR SDRAM (Single Data Rate SDRAM) and / or one or more non-volatile memory devices such as EEPROM (Electrical Erasable Programmable ROM) and flash memory.

[0245] Storage (104) (storage device) can store and retain data and / or programs, etc. Storage (104) may include one or more non-volatile memory devices such as a solid-state drive (SSD), a hard drive, and a flash memory. In the present invention, storage (104) can store a computer program consisting of instructions for performing a cloud resource scheduling method.

[0246] The interface (105) can perform the function of transmitting data to a communication network or receiving data from a communication network. The interface (105) may be wired or wireless. For example, the interface (105) may include an antenna or a wired / wireless transceiver. The interface (150) may be configured to transmit and receive data with a device such as a server or other electronic device. The interface (150) may be configured to transmit and receive data with a physical server (200), a cloud server (V10), and / or a manager device (300). The interface (105) may include a Bluetooth communication unit, a BLE (Bluetooth Low Energy) communication unit, a near field communication unit, a WLAN (Wi-Fi) communication unit, a Zigbee communication unit, an infrared (IrDA, infrared Data Association) communication unit, a WFD (Wi-Fi Direct) communication unit, an UWB (ultra wideband) communication unit, an Ant+ communication unit, a mobile communication network, or a wired Ethernet network.

[0247] A display unit (not shown) may include any device capable of displaying an image, such as a display device, a screen device, or a beam projector. A display (e.g., a display) may include a panel, a holographic device, a projector, or a control circuit for controlling these. The panel may be implemented, for example, in a flexible, transparent, or wearable manner. The panel may be composed of a touch panel and one or more modules. According to one embodiment, the panel may include a pressure sensor (or force sensor) capable of measuring the intensity of pressure in response to a user's touch. The pressure sensor may be implemented integrally with the touch panel or as one or more sensors separate from the touch panel. A holographic device may display a three-dimensional image in mid-air using light interference. A projector may display an image by projecting light onto a screen. The screen may be located, for example, inside or outside an electronic device.

[0248] A server according to embodiments of the present invention may be a system formed by connecting a plurality of computing devices to each other through a network. In such cases, each module or combination of modules may be implemented as a computing device. However, the present embodiment is not limited thereto.

[0249] Additionally, the server may be implemented as at least one of a workstation, a data center, an internet data center (IDC), a direct attached storage (DAS) system, a storage area network (SAN) system, a network attached storage (NAS) system, a redundant array of inexpensive disks, or a redundant array of independent disks (RAID) system, and an Electronic Document Management (EDMS) system, but the present embodiment is not limited thereto.

[0250] In some embodiments of the present invention, when a program is executed by the at least one processor, it may perform the operation of receiving request information, which is information related to a scheduling request of a cloud server; the operation of calculating a quantum advantage index by analyzing the request information using a predetermined analysis method; the operation of selecting one of a quantum analysis method and an artificial intelligence analysis method based on the quantum advantage index; and the operation of calculating response information, which is a response to the scheduling request, using the selected analysis method.

[0251] The cloud resource scheduling method according to the embodiment may also be implemented in the form of a computer-readable medium that stores instructions and data executable by a computer.

[0252] In this case, instructions and data may be stored in the form of program code, and when executed by a processor, they may generate a specific program module to perform a specific operation. Additionally, a computer-readable medium may be any available medium accessible by a computer, and includes both volatile and non-volatile media, as well as removable and inremovable media. Furthermore, a computer-readable medium may be a computer recording medium, which may include both volatile and non-volatile, removable and inremovable media implemented by any method or technique for storing information such as computer-readable instructions, data structures, program modules, or other data. For example, a computer recording medium may be a magnetic storage medium such as HDDs and SSDs, an optical recording medium such as CDs, DVDs, and Blu-ray discs, or memory contained in a server accessible via a network.

[0253] In addition, the cloud resource scheduling method according to the embodiment may be implemented as a computer program (or computer program product) comprising instructions executable by a computer. The computer program includes programmable machine instructions processed by a processor and may be implemented in a high-level programming language, an object-oriented programming language, assembly language, or machine language, etc. Additionally, the computer program may be recorded on a tangible computer-readable recording medium (e.g., memory, hard disk, magnetic / optical medium, or SSD (Solid-State Drive), etc.).

[0254] Accordingly, the cloud resource scheduling method according to the embodiment can be implemented by executing a computer program as described above by a computing device. The computing device may include at least some of a processor, memory, storage, a high-speed interface connected to the memory and a high-speed expansion port, and a low-speed interface connected to the low-speed bus and storage. Each of these components is connected to one another using various buses and may be mounted on a common motherboard or mounted in other suitable ways.

[0255] Here, the processor can process instructions within the computing device, such as instructions stored in memory or storage devices to display graphic information for providing a Graphic User Interface (GUI) on external input and output devices, such as a display connected to a high-speed interface. In another embodiment, a plurality of processors and / or a plurality of buses may be utilized together with a plurality of memories and memory types. Additionally, the processor may be implemented as a chipset comprising chips including a plurality of independent analog and / or digital processors.

[0256] In addition, memory stores information within a computing device. For example, memory may consist of volatile memory units or a set thereof. As another example, memory may consist of non-volatile memory units or a set thereof. Furthermore, memory may be other forms of computer-readable media, such as magnetic or optical discs.

[0257] And storage can provide a large amount of storage space to computing devices. A storage device may be a computer-readable medium or a configuration containing such a medium, and may include, for example, devices or other configurations within a Storage Area Network (SAN), and may be a floppy disk device, a hard disk device, an optical disk device, or a tape device, flash memory, or other similar semiconductor memory device or device array.

[0258] Additionally, a cloud resource scheduling device according to some embodiments of the present invention comprises a processor, a memory for loading a computer program executed by the processor, and a storage for storing the computer program, wherein the computer program may include an operation of receiving request information, which is information related to a scheduling request of a cloud server; an operation of calculating a quantum advantage index by analyzing the request information using a predetermined analysis method; an operation of selecting one of a quantum analysis method and an artificial intelligence analysis method based on the quantum advantage index; and an operation of calculating response information, which is a response to the scheduling request, using the selected analysis method.

[0259] In addition, a computer program according to some embodiments of the present invention may be combined with a computer device which is hardware and stored on a recording medium readable by a computer device capable of performing the cloud resource scheduling method described above.

[0261] The manager device is a computing device, and a detailed description may be omitted within the same scope as the description of the computing device described above.

[0262] A physical server is a server or computing device, and a detailed description may be omitted within the same scope as the description of the server or computing device described above.

[0264] Some embodiments of the present invention can be utilized for managing and operating heterogeneous private cloud servers and public cloud servers. Through this, an administrator can manage and integrally control physical servers, heterogeneous private clouds, and public clouds from a single portal by utilizing response information received from a cloud resource scheduling device.

[0265] That is, the administrator device can maximize cloud service efficiency by scheduling the placement or cloud resource allocation of the physical server, the cloud server operating on the physical server, and each task operating on the cloud server based on the above final response information.

[0267] In order to more clearly express the technical concept of the present invention, the attached drawings briefly depict or omit configurations that are unrelated to or have little relevance to the technical concept of the present invention.

[0268] The above description is merely an illustrative explanation of the technical concept of the present embodiment, and a person skilled in the art to which the present embodiment belongs would be able to make various modifications and variations within the scope of the essential characteristics of the present embodiment. Accordingly, the present embodiments are intended to explain, not limit, the technical concept of the present embodiment, and the scope of the technical concept of the present embodiment is not limited by these embodiments. The scope of protection of the present embodiment shall be interpreted by the claims below, and all technical concepts within an equivalent scope shall be interpreted as being included within the scope of rights of the present embodiment. Explanation of the symbols

[0270] 100 : Cloud resource scheduling unit 200 : Physical server V10 : Cloud Server 300 : Management Device

Claims

Claim 1 A scheduling method implemented by a computing device comprises: receiving request information, which is information related to a scheduling request of a cloud server; analyzing the request information using a predetermined analysis method to calculate a quantum superiority index; and selecting one of a quantum analysis method and an artificial intelligence analysis method based on the quantum superiority index. A scheduling method comprising the step of producing response information, which is a response to a scheduling request, using a selected analysis method; wherein the request information comprises a first request information, which is information about a desired computing resource for each task, which is an independent resource unit of the cloud, and a second request information, which is information about constraints on the scheduling of each task; wherein the predetermined analysis method is a method of analyzing by utilizing the first request information and the second request information to represent the relationship between tasks in the form of an edge graph based on a vertex representing a task and constraints, and then utilizing a first evaluation indicator that considers the degree of connectivity between neighboring vertices and a second evaluation indicator that considers the degree of density between all vertices, and wherein the first evaluation indicator is an indicator calculated as the ratio of the number of actual connections between neighboring vertices to the maximum number of possible connections between neighboring vertices. Claim 2 In paragraph 1, the first evaluation indicator is calculated by mathematical formula 1, and the mathematical formula 1 is, A scheduling method in which n is the number of adjacent vertices and E represents the number of constraints between adjacent vertices. Claim 3 A scheduling method implemented by a computing device comprises: receiving request information, which is information related to a scheduling request of a cloud server; analyzing the request information using a predetermined analysis method to calculate a quantum superiority index; and selecting one of a quantum analysis method and an artificial intelligence analysis method based on the quantum superiority index. A scheduling method comprising the step of calculating response information, which is a response to a scheduling request, using a selected analysis method; wherein the request information comprises a first request information, which is information about a desired computing resource for each task, which is an independent resource unit of the cloud, and a second request information, which is information about constraints on the scheduling of each task; wherein the predetermined analysis method is a method of analyzing by utilizing the first request information and the second request information to represent the relationship of tasks in the form of an edge graph based on a vertex representing a task and constraints, and then utilizing a first evaluation indicator that considers the degree of connectivity between neighboring vertices and a second evaluation indicator that considers the degree of density between all vertices, and wherein the second evaluation indicator is an indicator calculated based on the weights of a 2-1 evaluation indicator that considers the degree of connectivity between all vertices and a 2-2 evaluation indicator that considers the degree of grouping of all vertices. Claim 4 In paragraph 3, the above 2-1 evaluation indicator is a scheduling method that means the average constraint density obtained by averaging the local constraint densities for each vertex. Claim 5 In paragraph 3, the above 2-2 evaluation indicator is a scheduling method signifying modularity. Claim 6 A scheduling method implemented by a computing device comprises: receiving request information, which is information related to a scheduling request of a cloud server; analyzing the request information using a predetermined analysis method to calculate a quantum advantage index; selecting one of a quantum analysis method and an artificial intelligence analysis method based on the quantum advantage index; and calculating response information, which is a response to a scheduling request, using the selected analysis method; wherein the request information comprises a first request information, which is information regarding a desired computing resource for each task, which is an independent resource unit of the cloud, and a second request information, which is information regarding constraints on the scheduling of each task; and wherein the predetermined analysis method is a method of analyzing by utilizing the first request information and the second request information to represent the relationship of tasks in the form of an edge graph based on a vertex representing a task and constraints, and then utilizing a first evaluation metric that considers the degree of connectivity between neighboring vertices, a second evaluation metric that considers the degree of density between all vertices, a third evaluation metric that represents the number of vertices, and a fourth evaluation metric that considers time dependency constraints and the time window overlap ratio. Claim 7 In paragraph 6, the above-mentioned fourth evaluation metric is a scheduling method calculated by the complexity between the number of time-dependent constraints and the time window overlap ratio. Claim 8 A scheduling method according to claim 6, wherein the step of calculating the quantum superiority index is the step of inputting the first to fourth evaluation indicators into a pre-trained artificial intelligence model to calculate the quantum superiority index. Claim 9 A scheduling method implemented by a computing device, comprising: receiving request information, which is information related to a scheduling request of a cloud server; analyzing the request information using a predetermined analysis method to calculate a quantum advantage index; selecting one of a quantum analysis method and an artificial intelligence analysis method based on the quantum advantage index; if the quantum advantage index is less than a first reference value and greater than or equal to a second reference value, selecting a hybrid analysis method that utilizes both the quantum analysis method and the artificial intelligence analysis method without selecting one of the quantum analysis method and the artificial intelligence analysis method; and calculating response information, which is a response to a scheduling request, using the selected analysis method. Claim 10 A scheduling method according to claim 9, wherein the step of selecting one of the quantum analysis method and the artificial intelligence analysis method is to select the quantum analysis method when the quantum advantage index is greater than or equal to the first reference value, and to select the artificial intelligence analysis method when the quantum advantage index is less than the second reference value. Claim 11 In claim 9, the above-determined analysis method is a method of analysis utilizing an edge-shaped graph that expresses the relationship of tasks based on vertices representing tasks and constraints using the above-determined request information, and the above-determined hybrid analysis method is a scheduling method that utilizes the above-determined quantum analysis method or the above-determined artificial intelligence analysis method for each group according to the degree of coupling between vertices within the group after dividing all vertices into a minimum group through a partitioning algorithm. Claim 12 In any one of claims 1, 3, 6 and 9, the artificial intelligence analysis method is a scheduling method that produces the response information by sequentially utilizing a simulated annealing technique and a genetic algorithm. Claim 13 A scheduling method, wherein in any one of claims 1, 3, 6 and 9, the step of calculating the response information is a step of calculating the response information including cloud resource placement or allocation using a selected analysis method. Claim 14 A scheduling method according to any one of paragraphs 1, 3, 6 and 9, wherein the step of calculating the response information comprises: a step of calculating a plurality of candidate response information using a selected analysis method; and a step of calculating a final response information among the candidate response information based on a reliability index calculated through a predetermined reliability analysis method. Claim 15 In paragraph 14, the above-mentioned predetermined reliability analysis method is a scheduling method that analyzes by considering a first reliability index, which is an index for determining whether a constraint condition is satisfied; a second reliability index, which is an index for determining whether a predetermined objective function is satisfied; and a third reliability index, which is an index for determining by inferring past performance. Claim 16 A scheduling device comprising: a processor; a memory for loading a computer program executed by the processor; and a storage for storing the computer program, wherein the computer program comprises: an operation of receiving request information, which is information related to a scheduling request of a cloud server; an operation of calculating a quantum advantage index by analyzing the request information using a predetermined analysis method; an operation of selecting one of a quantum analysis method and an artificial intelligence analysis method based on the quantum advantage index; an operation of selecting a hybrid analysis method that utilizes both the quantum analysis method and the artificial intelligence analysis method without selecting one of the quantum analysis method and the artificial intelligence analysis method when the quantum advantage index is less than a first reference value and greater than or equal to a second reference value; and an operation of calculating response information, which is a response to a scheduling request, using the selected analysis method. Claim 17 A computer program stored in a computer-readable storage medium to execute a scheduling method according to any one of claims 1 to 11, combined with a computer device which is hardware.

Images