Calculation instance scheduling method and device and calculation equipment

Abstract

The invention discloses a computing instance scheduling method and device and computing equipment. When a computing device needs to migrate out of computing examples, a multi-dimensional kernel index of each computing example can be input into a data driving model, and the degradation ratio between the number of the computing examples with QoS degradation values larger than a set threshold value in other computing examples and the total number of the computing examples after each computing example is migrated out is counted; and the calculation instance corresponding to the minimum degradation ratio is migrated out, so that the influence of the calculation instance to be migrated out on other calculation instances can be reduced, and the QoS (Quality of Service) performance of other calculation instances of the calculation equipment can be improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a scheduling method, device and computing equipment for computing instances. Background Art

[0002] Quality of service (QoS) degradation refers to the phenomenon that the QoS performance of tenant services is reduced due to resource competition among multiple virtual machines (VMs) on the same host. Usually, because multiple VMs share the central processing unit (CPU), memory, network bandwidth, cache and other resources in the host, the relevant resources cannot meet the needs of the application load in a timely manner, causing the tenant's service QoS performance indicators to drop significantly.

[0003] In cloud computing scenarios, it is very common for cloud service providers to oversell computing resources. Overselling computing resources can improve hardware utilization, but it can easily cause the application performance of tenants' businesses to degrade. If the QoS degradation of a tenant's computing instance exceeds the set threshold, the cloud service provider needs to take some measures to protect the QoS of the tenant's business in compliance with the service level agreement (SLA). Currently, the commonly used measures are: First, the host can reallocate the resource ratios of different computing instances within the host to ensure the business QoS of tenants with higher priority. Second, the host migrates some low-priority computing instances and releases the resources of low-priority computing instances, thereby alleviating the degradation of other computing instances in the host.

[0004] If the host still cannot alleviate the QoS degradation of tenant services after reallocating computing instance resources, it is necessary to intervene in time to prevent the QoS degradation from exceeding the threshold and select one or more computing instances to perform hot migration. However, the host selects which computing instances to migrate to achieve the greatest overall benefits for tenants and which target host to migrate to to avoid the ping-pong effect, which are the problems that need to be solved. Summary of the invention

[0005] In order to solve the above problems, an embodiment of the present application provides a scheduling method for computing instances, which migrates the computing instances with the least impact on other computing instances and the greatest benefits, so as to improve the QoS performance of other computing instances of the computing device. In addition, the present application also provides a scheduling device and a computing device for computing instances corresponding to the scheduling method for computing instances.

[0006] To this end, the following technical solutions are adopted in the embodiments of the present application:

[0007] In a first aspect, an embodiment of the present application provides a scheduling method for computing instances, the method being executed by a computing device, the computing device comprising multiple computing instances, the method comprising: obtaining multidimensional kernel indicators of the multiple computing instances, the multidimensional kernel indicators comprising one or more of the physical resources of the computing instances, performance indicators, resource utilization, load rate, busyness, and affinity; inputting the multidimensional kernel indicators of the multiple computing instances into a data-driven model to obtain computing instances to be migrated; the computing instances to be migrated are computing instances with the smallest degradation ratio among the multiple computing instances, the degradation ratio of the computing instances being the ratio between the number of computing instances whose quality of service (QoS) degradation values ​​in other computing instances are greater than a set threshold and the total number of computing instances in the computing device after the computing device migrates out one or more computing instances.

[0008] In this implementation, when the computing device needs to migrate out a computing instance, the multi-dimensional kernel indicators of each computing instance can be input into the data-driven model, and the degradation ratio between the number of computing instances in other computing instances whose QoS degradation values ​​are greater than a set threshold and the total number of multiple computing instances after each computing instance is migrated out is calculated, and the computing instance corresponding to the smallest degradation ratio is migrated out. This can reduce the impact of the computing instance to be migrated on other computing instances, thereby improving the QoS performance of other computing instances of the computing device.

[0009] In one embodiment, the method also includes: determining multiple candidate computing devices based on the physical resources of the computing instance to be migrated and the remaining physical resources of other computing devices; inputting the multidimensional kernel indicators of the computing instance to be migrated and the multidimensional kernel indicators of each computing instance in the multiple candidate computing devices into the data-driven model to obtain a target computing device, wherein the target computing device is the candidate computing device with the smallest degradation ratio among the multiple candidate computing devices, and the degradation ratio of the candidate computing device is the ratio between the number of original computing instances in the candidate computing device whose QoS degradation value is greater than a set threshold and the total number of original computing instances in the candidate computing devices after the candidate computing device is migrated to the migrated computing instance; migrating the computing instance to be migrated into the target computing device.

[0010] In this implementation, after the computing device selects the computing instance to be migrated, it can determine multiple candidate computing devices based on the physical resources of the computing instance to be migrated and the remaining physical resources of other computing devices. The computing device then inputs the multi-dimensional kernel indicators of the computing instance to be migrated and the multi-core kernel indicators of each computing instance of each candidate computing device into the data-driven model, and counts the degradation ratio between the number of computing instances whose QoS degradation values ​​in the original computing instances are greater than the set threshold after each candidate computing device migrates to the computing instance to be migrated and the total number of the original multiple computing instances, and uses the candidate computing device corresponding to the smallest degradation ratio as the target computing device, which can reduce the impact of the computing instance to be migrated on the original computing instance of the target computing device, thereby avoiding the ping-pong effect between the computing device and the target computing device.

[0011] In one embodiment, before obtaining the multi-dimensional kernel indicators of the multiple computing instances, the method further includes: triggering a scheduling task of the computing device, wherein the scheduling task instructs the computing device to migrate out of the computing instance.

[0012] In this implementation, the physical resources of the computing device are oversold, resulting in the degradation of the QoS of the computing instances within the computing device. Therefore, the computing device is required to perform scheduling tasks and migrate out some computing tasks to improve the QoS performance of other computing instances of the computing device.

[0013] In one embodiment, the scheduling task that triggers the computing device, wherein the scheduling task instructs the computing device to migrate out of the computing instance, specifically includes: detecting QoS degradation values ​​of the multiple computing instances; triggering the scheduling task of the computing device when the number of computing instances whose QoS degradation values ​​among the multiple computing instances are greater than a set threshold is greater than a set number, or the QoS degradation value of a specific computing task is greater than a set threshold.

[0014] In this implementation, if the number of computing instances in the computing device whose QoS degradation value is greater than a set threshold is greater than a set number, or the QoS degradation value of a specific computing instance is greater than a set threshold, it indicates that the QoS of the computing instance inside the computing device is seriously degraded, and the computing device needs to execute scheduling tasks and migrate out some computing tasks to improve the QoS performance of other computing instances of the computing device.

[0015] In one embodiment, the multi-dimensional kernel indicators of the multiple computing instances are input into the data-driven model to obtain the computing instances to be migrated, which specifically includes: taking one of the multiple computing instances in turn as the assumed computing instance to be migrated; detecting the QoS degradation values ​​of the computing instances other than the assumed computing instance to be migrated among the multiple computing instances, and judging whether each QoS degradation value is greater than the set threshold; counting the computing instances whose QoS degradation values ​​are greater than the set threshold, and calculating the degradation ratio between the number of computing instances whose QoS degradation values ​​are greater than the set threshold and the total number of the multiple computing instances; taking the assumed computing instance to be migrated corresponding to the smallest degradation ratio as the computing instance to be migrated.

[0016] In one embodiment, there are multiple computing instances with the smallest degradation ratio, and the method further includes: when the computing instances with the smallest degradation ratios are used as the assumed computing instances to be migrated, detecting the QoS degradation value of a specific computing instance; and using the assumed computing instance to be migrated when the QoS degradation value of the specific computing instance is the smallest as the computing instance to be migrated.

[0017] In this implementation, if there are multiple computing instances with the smallest degradation ratio, the data-driven model can detect the QoS degradation value of a specific computing instance when each computing instance with the smallest degradation ratio is used as the assumed computing instance to be migrated. The data-driven model can use the assumed computing instance to be migrated when the QoS degradation value of the specific computing instance is the smallest as the final computing instance to be migrated to ensure the user experience of the tenants of the specific computing instance.

[0018] In the second aspect, an embodiment of the present application provides a scheduling device for computing instances, comprising: an identification unit, used to obtain multidimensional kernel indicators of the multiple computing instances, the multidimensional kernel indicators including one or more of the physical resources, performance indicators, resource utilization, load rate, busyness and affinity of the computing instances; inputting the multidimensional kernel indicators of the multiple computing instances into a data-driven model to obtain the computing instances to be migrated; the computing instances to be migrated are the computing instances with the smallest degradation ratio among the multiple computing instances, and the degradation ratio of the computing instances is the ratio between the number of computing instances whose service quality QoS degradation values ​​in other computing instances are greater than a set threshold and the total number of computing instances in the computing device after the computing device migrates out one or more computing instances.

[0019] In one embodiment, a prediction unit is used to determine multiple candidate computing devices based on the physical resources of the computing instance to be migrated and the remaining physical resources of other computing devices; the multidimensional kernel indicators of the computing instance to be migrated and the multidimensional kernel indicators of each computing instance in the multiple candidate computing devices are input into the data-driven model to obtain a target computing device, wherein the target computing device is the candidate computing device with the smallest degradation ratio among the multiple candidate computing devices, and the degradation ratio of the candidate computing device is the ratio between the number of original computing instances in the candidate computing device whose QoS degradation value is greater than a set threshold and the total number of original computing instances in the candidate computing devices after the candidate computing device is migrated to the migrated computing instance; and a scheduling unit is used to migrate the computing instance to be migrated into the target computing device.

[0020] In one embodiment, the identification unit is further used to trigger a scheduling task of the computing device, wherein the scheduling task instructs the computing device to migrate out of the computing instance.

[0021] In one embodiment, the identification unit is specifically used to detect the QoS degradation values ​​of the multiple computing instances; when the number of computing instances whose QoS degradation values ​​among the multiple computing instances are greater than a set threshold is greater than a set number, or the QoS degradation value of a specific computing task is greater than a set threshold, the scheduling task of the computing device is triggered.

[0022] In one embodiment, the identification unit is specifically used to sequentially select one of the multiple computing instances as a hypothetical computing instance to be migrated; detect the QoS degradation values ​​of the computing instances other than the hypothetical computing instance to be migrated among the multiple computing instances, and determine whether each QoS degradation value is greater than the set threshold; count the computing instances whose QoS degradation values ​​are greater than the set threshold, and calculate the degradation ratio between the number of computing instances whose QoS degradation values ​​are greater than the set threshold and the total number of the multiple computing instances; and select the hypothetical computing instance to be migrated corresponding to the smallest degradation ratio as the computing instance to be migrated.

[0023] In one embodiment, the identification unit is also used to detect the QoS degradation value of a specific computing instance when there are multiple computing instances with the smallest degradation ratio and when each computing instance with the smallest degradation ratio is used as the assumed computing instance to be migrated; and use the assumed computing instance to be migrated when the QoS degradation value of the specific computing instance is the smallest as the computing instance to be migrated.

[0024] In a third aspect, an embodiment of the present application provides a computing device, comprising: at least one memory and at least one processor, wherein the processor is used to execute instructions stored in the memory so that the computing device executes various possible implementation embodiments of the first aspect.

[0025] In a fourth aspect, a computer-readable storage medium is provided in an embodiment of the present application, comprising computer program instructions. When the computer program instructions are executed by a computing device, the computing device executes various possible implementations of the first aspect.

[0026] In a fifth aspect, a computer program product comprising instructions is provided in an embodiment of the present application. The computer program product stores instructions, and when the instructions are executed by a computing device, the computing device implements various possible implementation embodiments of the first aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The following is a brief introduction to the drawings required for describing the embodiments or prior art.

[0028] Figure 1 A schematic diagram of the structure of a computing device provided in an embodiment of the present application;

[0029] Figure 2 A schematic diagram of the structure of the scheduling agent provided in an embodiment of the present application;

[0030] Figure 3 A flowchart of a scheduling method for a computing instance provided in an embodiment of the present application. DETAILED DESCRIPTION

[0031] The technical solutions in the embodiments of the present application will be described below in conjunction with the drawings in the embodiments of the present application.

[0032] The term "and / or" in this article is a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. The symbol " / " in this article indicates that the associated objects are in an or relationship, for example, A / B means A or B.

[0033] The terms "first" and "second" in the specification and claims herein are used to distinguish different objects rather than to describe a specific order of the objects. For example, a first response message and a second response message are used to distinguish different response messages rather than to describe a specific order of the response messages.

[0034] In the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way.

[0035] In the description of the embodiments of the present application, unless otherwise specified, "multiple" means two or more than two. For example, multiple processing units refer to two or more processing units, etc.; multiple elements refer to two or more elements, etc.

[0036] In order to solve the problems existing in the related art, the embodiments of the present application provide a scheduling method for computing instances, a scheduling device for computing instances, and a computing device. In a scenario where computing instances of different priorities are mixed, the computing device can select the computing instance that has the greatest impact on the target tenant to migrate out when the QoS degradation of the target tenant exceeds the set threshold, or select the computing instance that benefits other tenants the most to migrate out, thereby improving the QoS performance of other tenants of the computing device. At the same time, when the computing instance migrates to other computing devices, the computing device can predict the interference of the migrated computing instance on the QoS of the existing computing instance of the embedded computing device, thereby avoiding the ping-pong effect between the computing device and other computing devices. The ping-pong effect refers to the phenomenon that the migration planning of the computing instance is unreasonable, resulting in migration back and forth between two computing devices.

[0037] Figure 1 Schematic diagram of a computing device provided in an embodiment of the present application. Figure 1 As shown, the computing device 100 includes a physical resource layer 110, a virtual resource layer 120 and a service layer 130. The computing device 100 may be a server, a computer, a desktop computer, a portable notebook or other computing function-enabled device.

[0038] The physical resource layer 110 includes processors, cache components, memory components, network components or other types of physical resources. The processor may be a CPU, a graphics processing unit (GPU), a neural processing unit (NPU), a digital signal processor (DSP), an application specific integrated circuit (ASIC), etc. The cache component may be a CPU cache, a hard disk cache, a network card cache, a database cache, etc. The memory component may be a random access memory (RAM), a hard disk drive (HDD), a solid state drive (SSD), etc. The network component may be a network card, a repeater, a network router (NR), a network switch (NS), a network integrated circuit (NIC), a fiber optic device (FOD), etc.

[0039] A host operating system (host OS) can be run in the virtual resource layer 120. Virtualization software runs in the host OS, and the physical resources of the physical resource layer 110 are virtualized into virtual resources through the virtualization software for use by computing instances. In an embodiment of the present application, the host OS can virtualize the processor into N virtual processors, each of which can include at least one processor core. N is a positive integer greater than or equal to 1. The host OS can virtualize the memory component into N virtual memories. For example, the host OS can virtualize a 128GB memory card into 4 32GB virtual memories for use by four computing instances. The host OS can virtualize the network component into N virtual networks. The cache component does not need to be virtualized, and the computing instances can share the cache component. Therefore, the virtual cache inside the virtual resource layer 120 is obtained by virtualizing the cache component of the physical resource layer 110. Optionally, the virtual cache inside the virtual resource layer 120 can be virtualized from the cache component inside other computing devices.

[0040] The business layer 130 includes multiple computing instances. A computing instance refers to a virtual computer provided by a cloud service provider and available for rent to users. A computing instance is one of the core components of a cloud computing infrastructure and can provide a complete computing environment, including physical resources such as processors, memory, cache, operating system, and network components. In an embodiment of the present application, when the business layer 130 creates each computing instance, it will occupy a portion of the virtual resources of the virtual resource layer 120, such as virtual processors, virtual caches, virtual memory, virtual networks, etc. That is, each computing instance will occupy a portion of the computing resources such as processors, cache components, memory components, and network components in the physical resource layer 110.

[0041] A computing instance can be a computing unit for running a business, such as a VM, a container, or a process. If the computing instance is a VM, the computing resource can run a guest operating system (guest OS) so that tenants can run their own businesses. If the computing instance is a container or a process, tenants can run their own businesses directly on the computing instance.

[0042] The business layer 130 also includes a scheduling agent, which can collect parameters of various types of physical resources and parameters of physical resources of each computing instance through the host OS, calculate the QoS of the computing device 100 and the QoS of each computing instance, and determine whether the QoS of the computing device 100 and each computing instance exceeds a set threshold. When the scheduling agent determines that the QoS of the computing device 100 and each computing instance exceeds the set threshold, the computing instance to be migrated can be selected.

[0043] Figure 2 Schematic diagram of the structure of the scheduling agent provided in the embodiment of the present application. Figure 2 As shown, the scheduling agent 200 includes an identification unit 210, a prediction unit 220 and a scheduling unit 230. The identification unit 210, the prediction unit 220 and the scheduling unit 230 can be implemented by software, or by hardware, or by a combination of software and hardware.

[0044] Typically, the physical resources of the computing device 100 are oversold, resulting in degradation of the QoS of the computing instances within the computing device 100. If the QoS of each computing instance of the computing device 100 is degraded, or the QoS of a specific computing instance is degraded, the scheduling agent 200 needs to perform a scheduling task to resolve the QoS degradation of the computing instance.

[0045] The identification unit 210 can detect the degradation value of the QoS of the computing instance and determine whether the degradation value of the QoS of the computing instance is greater than a set threshold. If the number of computing instances in the computing device 100 whose degradation value of QoS is greater than the set threshold is greater than the set number, or the degradation value of QoS of a specific computing instance is greater than the set threshold, the identification unit 210 can collect multi-dimensional kernel indicators of each computing instance, and input the multi-dimensional kernel indicators of each computing instance into the data-driven model to predict the computing instance to be migrated. Among them, the multi-dimensional kernel indicators refer to the physical resources of the computing instance, performance indicators, resource utilization, load rate, busyness, affinity and other parameters.

[0046] A data-driven model is a model built based on data and machine learning techniques. It is used to learn from data, discover patterns and regularities, and make predictions, classifications, or decisions based on patterns and regularities. Data-driven models can automatically discover hidden relationships and regularities in data by training and learning from large amounts of data, thereby generating models that can predict and classify new data. The data-driven model training process can be divided into data collection, data preprocessing, feature selection and extraction, model training, model evaluation, and model application. The specific process is as follows:

[0047] Data collection: Collect relevant data, which can be structured data (such as tabular data in a database, etc.) or unstructured data (such as text, images, audio, etc.). Data preprocessing: Clean, organize and transform the data to remove noise, handle missing values, handle outliers, and other operations to ensure the quality and consistency of the data. Feature selection and extraction: According to the definition and objectives of the problem, select features related to prediction or classification, and perform appropriate transformation and extraction on the features. Model training: According to the type of problem, select appropriate algorithms, such as decision trees, support vector machines, neural networks, etc., and use machine learning algorithms to input preprocessed data into the model for training. Model evaluation: Use the reserved test data set to evaluate the trained model to measure the performance and generalization ability of the trained model on unknown data. Model application: The trained and evaluated model can be applied to actual scenarios to predict, classify, and make decisions on newly input data.

[0048] In an embodiment of the present application, the identification unit 210 can customize the training goal so that the computing instance that is migrated has the greatest benefit to other computing instances. After the data-driven model receives the multi-dimensional kernel indicators of each computing instance, one of the multiple computing instances can be used as the computing instance to be migrated. Then, the data-driven model can simulate the environment in which the computing device 100 migrates out the computing instance to be migrated, such as releasing the physical resources of the computing instance to be migrated, and allocating the released physical resources to other computing instances. Next, the data-driven model can detect the QoS degradation values ​​of other computing instances in the computing device 100, and determine whether the QoS degradation values ​​of other computing instances exceed the set threshold. The data-driven model counts the computing instances whose QoS degradation values ​​exceed the set threshold, and calculates the ratio of the number of computing instances whose QoS degradation values ​​exceed the set threshold to the total number of computing instances in the computing device 100 as the degradation ratio of the computing instance to be migrated. By analogy, the data-driven model takes multiple computing instances as the computing instances to be migrated in turn, and calculates the degradation ratio of each computing instance. Among them, the smaller the degradation ratio of the computing instance, the smaller the proportion of computing instances in the computing device 100 whose QoS degradation value exceeds the set threshold, indicating that the impact of the assumed computing instance to be migrated on other computing instances is smaller and the benefit is maximized.

[0049] After the data-driven model has counted the degradation ratios of each computing instance of the computing device 100, the computing instance with the smallest degradation ratio can be used as the computing instance to be migrated, and the computing instance to be migrated can be output to the prediction unit 220 and the scheduling unit 230. In one case, if there is only one computing instance with the smallest degradation ratio, the data-driven model can directly use the computing instance with the smallest degradation ratio as the computing instance to be migrated. In another case, if there are multiple computing instances with the smallest degradation ratio, the data-driven model can detect the QoS degradation value of a specific computing instance when each computing instance with the smallest degradation ratio is used as the computing instance to be migrated. The data-driven model can use the computing instance to be migrated when the QoS degradation value of the specific computing instance is the smallest as the final computing instance to be migrated. Among them, the specific computing instance can be a computing instance with a high priority, a computing instance of a paying tenant, a computing instance of an important tenant, or other types of computing instances.

[0050] After obtaining the computing instance to be migrated, the prediction unit 210 can select one or more candidate computing devices to which the computing instance to be migrated can be migrated according to the physical resources of the computing instance to be migrated and the physical resources available to other computing devices. The prediction unit 210 can collect the multi-dimensional kernel index of each computing instance in each candidate computing device, and input the multi-dimensional kernel index of the computing instance to be migrated and the multi-dimensional kernel index of each computing instance in each candidate computing device into the data-driven model to predict the target computing device to which the computing instance to be migrated can be migrated.

[0051] In the embodiment of the present application, after receiving the physical resources of the computing instance to be migrated and the physical resources of each computing instance in each candidate computing device, the data-driven model can simulate the environment when each candidate computing device runs the computing instance to be migrated, such as reallocating the physical resources of each candidate computing device, so that each candidate computing device allocates physical resources for the computing instance to be migrated. The data-driven model can detect the QoS degradation values ​​of each original computing instance in each candidate computing device, and determine whether the QoS degradation value of each computing instance exceeds the set threshold. The data-driven model counts the computing instances whose QoS degradation values ​​in each candidate computing device exceed the set threshold, and calculates the ratio of the number of computing instances whose QoS degradation values ​​in each candidate computing device exceed the set threshold to the total number of computing instances in the corresponding candidate computing device, as the degradation ratio of the candidate computing device. Among them, the smaller the degradation ratio of the candidate computing device, the smaller the proportion of the original computing devices whose QoS degradation values ​​in the candidate computing device exceed the set threshold, indicating that the computing instance to be migrated has a smaller impact on the original computing instance of the candidate computing device, and the maximum benefit. After the data-driven model calculates the degradation ratios of the candidate computing devices, the candidate computing device with the smallest degradation ratio may be used as the target computing device, and the target computing device may be output to the scheduling unit 230 .

[0052] After obtaining the computing instance to be migrated and the target computing device, the scheduling unit 230 can generate a scheduling policy. The scheduling policy can migrate the computing instance to be migrated to the target computing device. After completing the scheduling task, the scheduling unit 230 can send a detection instruction to the identification unit 210, so that the identification unit 210 can detect again whether the number of computing instances whose QoS degradation value is greater than the set threshold in the computing device 100 is greater than the set number, or the QoS degradation value of a specific computing instance is greater than the set threshold.

[0053] In the embodiment of the present application, when the number of computing instances whose QoS degradation value is greater than a set threshold value in the computing device 100 is greater than a set number, or the QoS degradation value of a specific computing instance is greater than a set threshold value, the computing instance with the smallest impact on other computing instances and the largest benefit can be migrated out to improve the QoS performance of other computing instances of the computing device 100. In addition, when the computing device 100 migrates the computing instance to be migrated to other computing devices, the impact of the computing instance to be migrated on each computing instance in the other computing devices is detected, and the computing instance to be migrated can be migrated to the computing device with the smallest impact on the QoS degradation value of each computing instance, so as to avoid the ping-pong effect between the computing device 100 and other computing devices.

[0054] Figure 3 Schematic diagram of the flow of the scheduling method of the computing instance provided in the embodiment of the present application. Figure 3 As shown, the method is executed by the computing device 100, and the specific implementation process is as follows:

[0055] S301, the computing device 100 determines whether to execute the scheduled task. In one case, when the computing device 100 determines that the scheduled task needs to be executed, step S301 is executed. In one case, when the computing device 100 determines that the scheduled task does not need to be executed, the process ends directly.

[0056] Normally, the physical resources of the computing device 100 are oversold, resulting in degradation of the QoS of the computing instances within the computing device 100. If the QoS of each computing instance of the computing device 100 is degraded, or the QoS of a specific computing instance is degraded, the computing device 100 needs to execute a scheduling task to resolve the QoS degradation of the computing instance. In an embodiment of the present application, the computing device 100 can detect the degradation value of the QoS of the computing instance, and determine whether the degradation value of the QoS of the computing instance is greater than a set threshold. If the number of computing instances in the computing device 100 whose QoS degradation value is greater than the set threshold is greater than the set number, or the QoS degradation value of a specific computing instance is greater than the set threshold, the scheduling task is executed.

[0057] S302, the computing device 100 obtains multi-dimensional kernel indicators of multiple computing instances of itself.

[0058] S303, the computing device 100 inputs the multi-dimensional kernel indicators of multiple computing instances into the data-driven model to determine the computing instances to be migrated.

[0059] The computing device 100 can collect multi-dimensional kernel indicators of each computing instance, and input the multi-dimensional kernel indicators of each computing instance into the data-driven model to predict the computing instance to be migrated. The multi-dimensional kernel indicators refer to the physical resources of the computing instance, performance indicators, resource utilization, load rate, busyness, affinity and other parameters.

[0060] After receiving the multi-dimensional kernel indicators of each computing instance, the data-driven model can use one of the multiple computing instances as the computing instance to be migrated out. Then, the data-driven model can simulate the environment in which the computing device 100 migrates out the computing instance to be migrated out, such as releasing the physical resources of the computing instance to be migrated out, and allocating the released physical resources to other computing instances. Next, the data-driven model can detect the QoS degradation values ​​of other computing instances in the computing device 100, and determine whether the QoS degradation values ​​of other computing instances exceed the set threshold. The data-driven model counts the computing instances whose QoS degradation values ​​exceed the set threshold, and calculates the ratio of the number of computing instances whose QoS degradation values ​​exceed the set threshold to the total number of computing instances in the computing device 100 as the degradation ratio of the computing instance to be migrated out. By analogy, the data-driven model uses multiple computing instances as the computing instances to be migrated out in turn, and calculates the degradation ratio of each computing instance. After the data-driven model counts the degradation ratios of each computing instance of the computing device 100, the computing instance with the smallest degradation ratio can be used as the computing instance to be migrated out.

[0061] In one case, if there is only one computing instance with the smallest degradation ratio, the data-driven model can directly use the computing instance with the smallest degradation ratio as the computing instance to be migrated.

[0062] In another case, if there are multiple computing instances with the smallest degradation ratio, the data-driven model can detect the QoS degradation value of a specific computing instance when each computing instance with the smallest degradation ratio is used as the computing instance to be migrated. The data-driven model can use the computing instance to be migrated when the QoS degradation value of the specific computing instance is the smallest as the final computing instance to be migrated. The specific computing instance can be a computing instance with a high priority, a computing instance of a paying tenant, a computing instance of an important tenant, or other types of computing instances.

[0063] S304, the computing device 100 determines a plurality of candidate computing devices according to the physical resources of the computing instance to be migrated and the remaining physical resources of other computing devices.

[0064] S305, the computing device 100 inputs the multi-dimensional kernel indicators of the computing instance to be migrated and the multi-dimensional kernel indicators of each computing instance of each candidate computing device into the data-driven model to determine the target computing device.

[0065] After the computing device 100 obtains the computing instance to be migrated, it can select one or more candidate computing devices to which the computing instance to be migrated can be migrated based on the physical resources of the computing instance to be migrated and the physical resources available to other computing devices. The computing device 100 can collect the multi-dimensional kernel index of each computing instance in each candidate computing device, and input the multi-dimensional kernel index of the computing instance to be migrated and the multi-dimensional kernel index of each computing instance in each candidate computing device into the data-driven model to predict the target computing device to which the computing instance to be migrated can be migrated.

[0066] After receiving the physical resources of the computing instance to be migrated and the physical resources of each computing instance in each candidate computing device, the data-driven model can simulate the environment of each candidate computing device when running the computing instance to be migrated, such as reallocating the physical resources of each candidate computing device and allowing each candidate computing device to allocate physical resources for the computing instance to be migrated. The data-driven model can detect the QoS degradation values ​​of each original computing instance in each candidate computing device, and determine whether the QoS degradation value of each computing instance exceeds the set threshold. The data-driven model counts the computing instances whose QoS degradation values ​​exceed the set threshold in each candidate computing device, and calculates the ratio of the number of computing instances whose QoS degradation values ​​exceed the set threshold in each candidate computing device to the total number of computing instances in the corresponding candidate computing device as the degradation ratio of the candidate computing device. After the data-driven model counts the degradation ratios of each candidate computing device, the candidate computing device with the smallest degradation ratio can be used as the target computing device.

[0067] S306, the computing device 100 generates a scheduling strategy according to the computing instance to be migrated and the target computing device.

[0068] S307, the computing device 100 migrates the computing instance to be migrated to the target computing device according to the scheduling policy.

[0069] After obtaining the computing instance to be migrated and the target computing device, the computing device 100 can generate a scheduling policy. The scheduling policy can migrate the computing instance to be migrated to the target computing device. After completing the scheduling task, the computing device 100 again determines whether to execute the scheduling task until the number of computing instances whose QoS degradation value among the remaining computing instances is greater than the set threshold is less than or equal to the set number, or the QoS degradation value of a specific computing instance is less than or equal to the set threshold.

[0070] In the embodiment of the present application, when the number of computing instances whose QoS degradation value is greater than a set threshold value in the computing device 100 is greater than a set number, or the QoS degradation value of a specific computing instance is greater than a set threshold value, the computing instance with the smallest impact on other computing instances and the largest benefit can be migrated out to improve the QoS performance of other computing instances of the computing device 100. In addition, when the computing device 100 migrates the computing instance to be migrated to other computing devices, the impact of the computing instance to be migrated on each computing instance in the other computing devices is detected, and the computing instance to be migrated can be migrated to the computing device with the smallest impact on the QoS degradation value of each computing instance, so as to avoid the ping-pong effect between the computing device 100 and other computing devices.

[0071] The present application also provides a computing device, which includes at least one memory and at least one processor, and the processor can execute the following Figure 1-Figure 3 And the above corresponding protected technical solutions, so that the computing device has the technical effect of the above protected technical solutions.

[0072] In an embodiment of the present application, a computer-readable storage medium is further provided, including computer program instructions. When the computer program instructions are executed by a computing device, the computing device executes the above Figure 1-Figure 3 and any method described in the corresponding description.

[0073] In an embodiment of the present application, a computer program product comprising instructions is also provided, wherein the computer program product stores instructions, and when the instructions are executed by a computing device, the computing device implements the above-mentioned Figure 1-Figure 3 and any method described in the corresponding description.

[0074] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the embodiments of the present application.

[0075] In addition, various aspects or features of the embodiments of the present application can be implemented as methods, devices or products using standard programming and / or engineering techniques. The term "product" used in this application covers computer programs that can be accessed from any computer-readable device, carrier or medium. For example, computer-readable media may include, but are not limited to: magnetic storage devices (e.g., hard disks, floppy disks or tapes, etc.), optical disks (e.g., compact discs (CDs), digital versatile discs (DVDs), etc.), smart cards and flash memory devices (e.g., erasable programmable read-only memory (EPROM), cards, sticks or key drives, etc.). In addition, the various storage media described herein may represent one or more devices and / or other machine-readable media for storing information. The term "machine-readable medium" may include, but is not limited to, wireless channels and various other media capable of storing, containing and / or carrying instructions and / or data.

[0076] In the above embodiment, Figure 2 The scheduling agent 200 in the embodiment can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented by software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) mode. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media integrated. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., an SSD), etc.

[0077] It should be understood that in various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0078] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0079] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0080] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0081] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present application is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or an access network device, etc.) to perform all or part of the steps of the method described in each embodiment of the embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program code.

[0082] The above is only a specific implementation of the embodiments of the present application, but the protection scope of the embodiments of the present application is not limited to this. Any technician familiar with the technical field can easily think of changes or replacements within the technical scope disclosed in the embodiments of the present application, which should be covered within the protection scope of the embodiments of the present application.

Claims

1. A method for scheduling a computing instance, characterized in that: The method is performed by a computing device, the computing device including a plurality of computing instances, and the method includes: Acquire multi-dimensional kernel indicators of the plurality of computing instances, the multi-dimensional kernel indicators including one or more of physical resources, performance indicators, resource utilization, load rate, busyness, and affinity of the computing instances; The multi-dimensional kernel indicators of the multiple computing instances are input into the data-driven model to obtain the computing instances to be migrated; the computing instances to be migrated are the computing instances with the smallest degradation ratio among the multiple computing instances, and the degradation ratio of the computing instances is the ratio between the number of computing instances in which the quality of service QoS degradation value in other computing instances is greater than the set threshold and the total number of computing instances in the computing device after the computing device migrates out one or more computing instances.

2. The method according to claim 1, characterized in that The method further comprises: Determine a plurality of candidate computing devices according to the physical resources of the computing instance to be migrated and the remaining physical resources of other computing devices; Inputting the multidimensional kernel index of the computing instance to be migrated and the multidimensional kernel index of each computing instance in the multiple candidate computing devices into the data-driven model, obtaining a target computing device, wherein the target computing device is a candidate computing device with the smallest degradation ratio among the multiple candidate computing devices, and the degradation ratio of the candidate computing device is the ratio between the number of original computing instances in the candidate computing device whose QoS degradation value is greater than a set threshold and the total number of original computing instances in the candidate computing device after the candidate computing device migrates to the migrated computing instance; Migrate the computing instance to be migrated to the target computing device.

3. The method according to claim 1 or 2, characterized in that: Before obtaining the multi-dimensional kernel indicators of the multiple computing instances, the method further includes: A scheduling task of the computing device is triggered, wherein the scheduling task instructs the computing device to migrate out of the computing instance.

4. The method according to claim 3, characterized in that The triggering of the scheduling task of the computing device, wherein the scheduling task instructs the computing device to migrate out of the computing instance, specifically includes: Detecting QoS degradation values ​​of the multiple computing instances; When the number of computing instances whose QoS degradation values ​​among the multiple computing instances are greater than a set threshold is greater than a set number, or the QoS degradation value of a specific computing task is greater than a set threshold, the scheduling task of the computing device is triggered.

5. The method according to any one of claims 1 to 4, characterized in that: The step of inputting the multi-dimensional kernel indicators of the plurality of computing instances into the data-driven model to obtain the computing instances to be migrated specifically includes: sequentially taking one of the multiple computing instances as a hypothetical computing instance to be migrated; Detecting QoS degradation values ​​of computing instances other than the computing instance assumed to be migrated out among the multiple computing instances, and determining whether each QoS degradation value is greater than the set threshold; Counting the computing instances whose QoS degradation values ​​are greater than the set threshold, and calculating the degradation ratio between the number of computing instances whose QoS degradation values ​​are greater than the set threshold and the total number of the multiple computing instances; The assumed to-be-eliminated calculation instance corresponding to the minimum degradation ratio is used as the to-be-eliminated calculation instance.

6. The method according to any one of claims 1 to 5, characterized in that: The number of the calculation instances with the smallest degradation ratio is multiple, and the method further includes: When the computing instances with the smallest degradation ratios are used as the computing instances to be migrated, detecting the QoS degradation value of the specific computing instances; The assumed computing instance to be migrated when the QoS degradation value of the specific computing instance is the smallest is used as the computing instance to be migrated.

7. A scheduling device for computing instances, characterized in that: include: An identification unit, configured to obtain multi-dimensional kernel indicators of the plurality of computing instances, wherein the multi-dimensional kernel indicators include one or more of physical resources, performance indicators, resource utilization, load rate, busyness, and affinity of the computing instances; The multi-dimensional kernel indicators of the multiple computing instances are input into the data-driven model to obtain the computing instances to be migrated; the computing instances to be migrated are the computing instances with the smallest degradation ratio among the multiple computing instances, and the degradation ratio of the computing instances is the ratio between the number of computing instances in which the quality of service QoS degradation value in other computing instances is greater than the set threshold and the total number of computing instances in the computing device after the computing device migrates out one or more computing instances.

8. The device according to claim 7, characterized in that A prediction unit, configured to determine a plurality of candidate computing devices according to the physical resources of the computing instance to be migrated and the remaining physical resources of other computing devices; Inputting the multidimensional kernel index of the computing instance to be migrated and the multidimensional kernel index of each computing instance in the multiple candidate computing devices into the data-driven model, obtaining a target computing device, wherein the target computing device is a candidate computing device with the smallest degradation ratio among the multiple candidate computing devices, and the degradation ratio of the candidate computing device is the ratio between the number of original computing instances in the candidate computing device whose QoS degradation value is greater than a set threshold and the total number of original computing instances in the candidate computing device after the candidate computing device migrates to the migrated computing instance; The scheduling unit is used to migrate the computing instance to be migrated to the target computing device.

9. The device according to claim 7 or 8, characterized in that The identification unit is further used to trigger a scheduling task of the computing device, wherein the scheduling task instructs the computing device to migrate out of the computing instance.

10. The device according to claim 9, characterized in that The identification unit is specifically used to detect the QoS degradation values ​​of the multiple computing instances; When the number of computing instances whose QoS degradation values ​​among the multiple computing instances are greater than a set threshold is greater than a set number, or the QoS degradation value of a specific computing task is greater than a set threshold, the scheduling task of the computing device is triggered.

11. The device according to any one of claims 7 to 10, characterized in that: The identification unit is specifically configured to sequentially select one of the plurality of computing instances as a hypothetical computing instance to be migrated; Detecting QoS degradation values ​​of computing instances other than the computing instance assumed to be migrated out among the multiple computing instances, and determining whether each QoS degradation value is greater than the set threshold; Counting the computing instances whose QoS degradation values ​​are greater than the set threshold, and calculating the degradation ratio between the number of computing instances whose QoS degradation values ​​are greater than the set threshold and the total number of the multiple computing instances; The assumed to-be-eliminated calculation instance corresponding to the minimum degradation ratio is used as the to-be-eliminated calculation instance.

12. The device according to any one of claims 8 to 11, characterized in that: The identification unit is further configured to detect the QoS degradation value of a specific computing instance when the number of the computing instances with the smallest degradation ratio is multiple and each computing instance with the smallest degradation ratio is used as the computing instance to be migrated; The assumed computing instance to be migrated when the QoS degradation value of the specific computing instance is the smallest is used as the computing instance to be migrated.

13. A computing device, characterized in that: include: at least one memory; At least one processor, wherein the processor is configured to execute instructions stored in the memory so that the computing device executes the method according to any one of claims 1 to 6.

14. A computer-readable storage medium, characterized in that: The method comprises computer program instructions, and when the computer program instructions are executed by a computing device, the computing device performs the method according to any one of claims 1 to 6.

15. A computer program product comprising instructions, characterized in that The computer program product stores instructions, which, when executed by a computing device, enable the computing device to implement the method according to any one of claims 1 to 6.

Images