CN111736981A - Container resource configuration method, apparatus, device and storage medium - Google Patents

Abstract

The embodiment of the invention discloses a container resource allocation method, a device, equipment and a storage medium. The method comprises the following steps: acquiring a maximum resource allocation value of a target container, and determining a target private resource and a target standby resource of the target container according to the maximum resource allocation value and a target percentile; determining a shared standby resource corresponding to the target container according to the target standby resource and the rest standby resources of at least one rest container except the target container in the physical machine corresponding to the target container; and configuring the resources of the target container according to the target private resources and the shared standby resources. Through the technical scheme, the container resources are more reasonably configured, and the resources are saved while performance guarantee is provided to a greater extent.

Description

Container resource configuration method, apparatus, device and storage medium

technical field

Embodiments of the present invention relate to computer technologies, and in particular, to a container resource configuration method, apparatus, device, and storage medium.

Background technique

Data centers for platforms such as Internet services, e-commerce services, public clouds, private clouds, and the Internet of Things usually run on servers or server clusters (called physical machines). Currently, the standard architecture of physical machines mostly uses container and microservice technologies. to share server resources. On the container platform, a physical machine can run multiple containers that are isolated from each other and run independently; a business or application is divided into multiple logically independent microservices, and instances of microservices run in containers.

In order to meet service performance and optimize resource utilization, it is necessary to configure the container with resources of appropriate size, such as central processing unit (Central Processing Unit, CPU) resources, internal memory (Memory) space, and disk (Disk) space. Configuring too many resources can guarantee the performance of the container or service, but it will waste resources; conversely, if the resource allocation is insufficient, the performance of the container or service cannot be guaranteed. Therefore, configuring appropriate container resources to ensure performance without wasting resources is a crucial issue for the efficient operation of physical machines, especially for online services with large load changes and strict performance requirements, such as e-commerce services.

At present, the methods for realizing container resource configuration mainly include: first, the container resource configuration based on the maximum value. This solution can estimate the maximum demanded resources required by a container to host services based on the administrator's experience, or analyze the historical resource usage data of a container to determine the maximum used resources of the container, so as to compare the maximum demanded resources or the maximum used resources Conservatively configure resources for this container to ensure that performance can be met under maximum load. Second, percentile-based container resource configuration. The solution is to determine the resource configuration of a container according to the maximum value and a certain percentile (eg, 90% percentile) in the case of determining the maximum value, so as to ensure that 90% of the resource requirements can be satisfied. Third, dynamic container resource configuration. This solution needs to monitor the current load of a container in real time and adjust the maximum value periodically to dynamically adjust the resource configuration of the container.

In the process of implementing the present invention, the inventor found that there are at least the following problems in the prior art: First, although the maximum value-based container resource configuration can guarantee the performance, it is difficult for containers and services whose load and demand are very dynamic, such as Internet services and Online services, many times do not require maximum resources, so it will waste a lot of resources and increase server costs. Second, although the percentile-based container resource configuration can save resources to a certain extent, it cannot guarantee to meet the resource requirements of all loads and requests, which will lead to certain performance conflicts. For important online services, such as e-commerce Business, even a 1% performance issue can cost a lot. Third, dynamic container resource allocation cannot be applied on a large scale due to the immature technology, stability and accuracy problems.

SUMMARY OF THE INVENTION

Embodiments of the present invention provide a container resource configuration method, apparatus, device, and storage medium, so as to realize a more rational configuration of container resources, provide performance assurance, and save resources at the same time.

In a first aspect, an embodiment of the present invention provides a container resource configuration method, including:

Obtain the maximum resource configuration value of the target container, and determine the target private resource and target backup resource of the target container according to the resource configuration maximum value and the target percentile;

Determine the shared backup resource corresponding to the target container according to the target backup resource and the remaining backup resources of at least one other container except the target container in the physical machine corresponding to the target container;

The resources of the target container are configured according to the target private resource and the shared spare resource.

In a second aspect, an embodiment of the present invention further provides an apparatus for configuring container resources, the apparatus including:

a target private resource determination module, configured to obtain the maximum resource configuration value of the target container, and determine the target private resource and target backup resource of the target container according to the resource configuration maximum value and the target percentile;

A shared spare resource determination module, configured to determine the shared spare resource corresponding to the target container according to the target spare resource and the remaining spare resources of at least one other container except the target container in the physical machine corresponding to the target container ;

A container resource configuration module, configured to configure resources of the target container according to the target private resource and the shared backup resource.

In a third aspect, an embodiment of the present invention further provides a device, and the device includes:

one or more processors;

storage means for storing one or more programs,

When the one or more programs are executed by the one or more processors, the one or more processors implement the container resource configuration method provided by any embodiment of the present invention.

In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, implements the container resource configuration method provided by any embodiment of the present invention.

The embodiment of the present invention realizes that the resources of a container are divided into two parts by obtaining the maximum resource configuration value of the target container, and determining the target private resources and target backup resources of the target container according to the maximum resource configuration value and the target percentile. The private resource is the resource used independently by the container to meet most of the performance requirements of the container, and the target spare resource will be used to determine the shared spare resource to meet the additional demand beyond the target private resource, which solves the waste of resources and reduces the performance conflicts caused by resources To a certain extent, the resource utilization is optimized and the performance guarantee is taken into account. By determining the shared spare resource corresponding to the target container according to the target spare resource and the remaining spare resources of at least one other container except the target container in the physical machine corresponding to the target container, it is possible to comprehensively consider the multiple resources on the physical machine when configuring the resources of a container. The resource configuration of each container enables multiple containers to share and share spare resources, solves the problem of inability to balance resources and performance caused by configuring a single container resource in isolation, and achieves a more reasonable configuration of container resources, so that the configured container resources are both The technical effect of ensuring performance and saving resources to a greater extent.

Description of drawings

FIG. 1 is a flowchart of a container resource configuration method in Embodiment 1 of the present invention;

2 is a flowchart of a container resource configuration method in Embodiment 2 of the present invention;

3a is a flowchart of a container resource configuration method in Embodiment 3 of the present invention;

FIG. 3b is a schematic diagram of a configuration result of two container resources in Embodiment 3 of the present invention;

3c is a flowchart of another container resource configuration method in Embodiment 3 of the present invention;

FIG. 3d is a schematic diagram of a resource configuration result of sub-container categories in Embodiment 3 of the present invention;

FIG. 3e is a schematic diagram of a comparison of results of configuring container resources based on different container resource configuration methods in Embodiment 3 of the present invention;

4 is a schematic structural diagram of a container resource configuration device according to Embodiment 4 of the present invention;

FIG. 5 is a schematic structural diagram of a device in Embodiment 5 of the present invention.

Detailed ways

The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention, but not to limit the present invention. In addition, it should be noted that, for the convenience of description, the drawings only show some but not all structures related to the present invention.

Example 1

The container resource configuration method provided in this embodiment is applicable to container resource allocation in a server of a data center or a server cluster architecture. The method may be performed by a container resource configuration apparatus, which may be implemented in software and/or hardware, and the apparatus may be integrated in a device based on a container technology architecture, such as a personal computer, a server, a server cluster, or a smart device cluster, etc. . Referring to FIG. 1, the method of this embodiment specifically includes the following steps:

S110. Acquire the maximum resource configuration value of the target container, and determine the target private resources and target backup resources of the target container according to the maximum resource configuration value and the target percentile.

The target container is a container of resources to be configured, which may be one container or multiple containers. The target container can be a newly created container or an already created and running container. The resources here may be at least one of CPU resources, internal memory (memory) resources, graphics card memory (video memory) resources, and disk resources. The maximum resource configuration refers to the maximum resource that needs to be configured for the target container, which can be the maximum resource requirement or the maximum historical resource usage. The maximum resource requirement refers to the maximum resource required by the service carried by the container, for example, how much resources the container needs is determined empirically according to the service that the container needs to carry. The maximum value of historical resource usage refers to the maximum value of the resources actually used by the container during the running process. The target percentile is a predetermined percentile value used to determine the target private resource for the target container. The target percentile can be a predetermined fixed value, such as an empirical 90%, or a dynamic value that changes dynamically according to the running state of the container. Target private resources refer to resources private to the target container, which are directly allocated to the target container and can only be used by the target container. The target spare resource refers to the spare resource calculated for the target container, which is not directly allocated to the target container, but is only a resource at the conceptual level.

In the related art, when determining configuration resources for a container, it is either based on the maximum resource configuration value (which can be a fixed value or a dynamic value), or based on a percentile resource value. Either one will have certain defects and cannot take into account the performance of the container. Therefore, when configuring resources for a target container in this embodiment of the present invention, the configured resources are divided into two parts, that is, target private resources and target backup resources, and the target private resources are used to satisfy most of the container’s requirements. The resource requirements for the performance of the target container are used to ensure most of the performance of the target container; the target spare resources are used as optional resources to provide resource requirements beyond the target private resources to ensure the additional performance of the target container. Such a setting will not allocate too many resources to the target container, but also provide sufficient guarantee for its performance, taking into account resource utilization and performance guarantee to a large extent.

Exemplarily, obtaining the maximum resource configuration of the target container includes: obtaining historical resource usage data corresponding to the target container, and determining the maximum historical resource usage of the target container according to the historical resource usage data as the maximum resource configuration.

Among them, the historical resource usage data refers to the data of the resources actually used by the container in the historical time period before the current moment. resource usage data.

When the maximum value of resource configuration is the maximum value of historical resource usage, it is necessary to obtain resource usage data of the target container for a period of time before the current moment, that is, historical resource usage data. If the target container is a newly created container, the historical resource usage data of the same type of container with the same load characteristics can be obtained as the historical resource usage data of the target container. The load characteristics here refer to CPU, memory, video memory, and disk storage. At this time, the resource configuration process of the target container is the resource configuration process of the created container. If the target container is a container that has already been created and running, you can directly obtain the historical resource usage data of the target container. At this time, the resource configuration process of the target container is the process of updating the container resources.

After obtaining the historical resource usage data of the target container, you can perform statistical analysis on the historical resource usage data to obtain the maximum resource value in the historical resource usage data as the maximum historical resource usage of the target container, that is, to obtain the maximum value of the target container's historical resource usage. Resource configuration maximum value max. The advantage of this setting is that it reduces the dependence of the resource allocation process on human experience, and improves the objectivity and rationality of resource allocation.

Exemplarily, determining the target private resources and target spare resources of the target container according to the resource configuration maximum value and the target percentile includes: determining the target private resource of the target container according to the resource configuration maximum value and the target percentile; and the target private resource to determine the target alternate resource.

The process of determining the target private resource and the target backup resource is as follows: calculate the product of the maximum resource configuration max and the target percentile, that is, determine the resource percentile value Tail-bound of the target container, and use it as the target private resource of the target container. After that, the target backup resource of the target container can be obtained by subtracting the target private resource from the maximum resource configuration value, that is, the target backup resource=max-Tail-bound.

S120. Determine a shared backup resource corresponding to the target container according to the target backup resource and the remaining backup resources of at least one other container except the target container in the physical machine corresponding to the target container.

Wherein, a physical machine is a term for a physical computer relative to a virtual machine, which may be a single computer or a server, or a server cluster or a smart device cluster. The remaining containers are concepts corresponding to the target container, which are containers other than the target container in the physical machine. In the embodiment of the present invention, the resources of each container on the physical machine are set to two parts: private resources and spare resources, and the other spare resources are spare resources of other containers. Shared spare resources refer to spare resources shared by multiple containers, which are actually allocated resource spaces, not conceptual resources, and can be connected to multiple containers at the same time instead of a single container. A shared standby resource is a resource that actually hosts the capabilities of a container beyond the private resource. The shared spare resource on a physical machine can be one, so that all containers on the physical machine share the shared spare resource; it can also be multiple, the first part of the containers on the physical machine share one of the shared spare resources, and the second part of the containers share the second shared spare resource. Shared spare resources, etc. The number of shared spare resources is determined according to business requirements. In this embodiment, it is preferable to set one shared spare resource on one physical machine, so as to maximize the compatibility of resource utilization and performance guarantee.

The determination of the shared spare resource in the embodiment of the present invention needs to depend on the spare resource of the container to which the shared spare resource can be connected, that is, the resource configuration of multiple containers needs to be comprehensively considered. When the shared spare resources on the physical machine are only connected to a part of the containers, determining the shared spare resources corresponding to the target container needs to rely on the target spare resources of the target container and the remaining spare resources of the other containers corresponding to the shared spare resources. When the shared spare resources on the physical machine are connected to all containers, determining the shared spare resources corresponding to the target container needs to depend on the target spare resources of the target container and the rest spare resources of all remaining containers. The comprehensive way of the target spare resource and at least one other spare resource, that is, the way of determining the shared spare resource may be to take the maximum value of each spare resource, or to group (or classify) each spare resource, and take the spare resources of the group (or each category). The mean or maximum value of the sum of resources, etc.

S130. Configure the resources of the target container according to the target private resources and the shared spare resources.

After the target private resource of the target container and its corresponding shared spare resource are determined, resource allocation can be performed according to the target private resource and the shared spare resource, so as to complete the resource configuration of the target container.

In the technical solution of this embodiment, by obtaining the maximum resource configuration of the target container, and determining the target private resources and target spare resources of the target container according to the maximum resource configuration and the target percentile, the resources of one container are divided into two parts. In part, the target private resource is the resource used independently by the container to meet most of the performance requirements of the container, and the target spare resource will be used to determine the shared spare resource to meet the additional demand beyond the target private resource, solving resource waste and reducing resources The problem that leads to performance conflicts optimizes resource utilization to a certain extent and takes into account performance guarantees. By determining the shared spare resource corresponding to the target container according to the target spare resource and the remaining spare resources of at least one other container except the target container in the physical machine corresponding to the target container, it is possible to comprehensively consider the multiple resources on the physical machine when configuring the resources of a container. The resource configuration of each container enables multiple containers to share and share spare resources, solves the problem of inability to balance resources and performance caused by configuring a single container resource in isolation, and achieves a more reasonable configuration of container resources, so that the configured container resources are both The technical effect of ensuring performance and saving resources to a greater extent.

Embodiment 2

This embodiment adds a step of "generating a target percentile" on the basis of the above-mentioned first embodiment. The explanations of terms that are the same as or corresponding to the above embodiments are not repeated here. Referring to FIG. 2, the container resource configuration method provided by this embodiment includes:

S210. Obtain the maximum value of the resource configuration of the target container.

S220. Determine the target resource usage time series data and the initial percentile corresponding to the target container.

The target resource usage time series data is historical resource usage data obtained when the target container or a container of the same type of the target container (when the target container is newly created) configures resources based on the percentile-based container resource configuration method. The initial percentile refers to the percentile on which the target container or its similar containers configure resources based on the percentile before the current operation.

The percentile in the percentile-based container resource configuration method in the related art is usually set empirically, which is subjective. Therefore, in the embodiment of the present invention, before determining the target private resource according to the percentile, the initial percentile is dynamically adjusted according to the resource usage of the target container, so as to obtain a percentile that is more objective and more suitable for the target container, that is, Target percentile.

To generate the target percentile, you need to obtain the initial percentile corresponding to the target container and the time series data of target resource usage. When the target container or similar containers have configured resources according to the percentile container resource configuration method, the historical resource usage time series data of the target container or similar containers and the percentile used for configuring resources can be directly obtained as The target resource corresponding to the target container uses time series data and initial percentiles.

When the target container or similar containers do not configure resources according to the percentile container resource configuration method, it is necessary to configure initial resources for the target container according to the percentile container resource configuration method. The above-mentioned initial resources may be referred to as initial private resources. The percentile on which this is based is the initial percentile. Exemplarily, determining the target resource usage time series data and the initial percentile corresponding to the target container includes: determining the initial private resource of the target container according to the resource configuration maximum value and the initial percentile; when the target container runs with the initial private resource, obtaining Time series data of the target resource usage of the target container within a preset time period. During specific implementation, the initial private resource of the target container is first determined according to the product of the maximum resource configuration value of the target container and the initial percentile. After that, trigger the target container to run for at least a preset time period under the condition of the initial private resource, and use the resource usage time series data within the preset time period as the target resource usage time series data of the target container.

S230. Determine the proportion of the time duration in which the continuous resource usage value is greater than the initial private resource in the target resource usage time series data.

The continuous resource usage value refers to a set number of resource usage values for which the time difference between every two resource usage values is less than a preset time threshold, and the preset time threshold can be manually set in advance, for example, 3 minutes. The initial private resource is the private resource of the target container corresponding to the initial percentile.

The basis for judging whether the initial percentile matches the performance requirements of the target container is mainly based on whether the resources allocated by the initial percentile can guarantee most of the container performance, that is, the resources caused by the allocated resources during the running of the container. Whether the conflict between performance and performance (allocation of resources cannot guarantee service performance, referred to as performance conflict) satisfies the user's tolerance for the degree of performance conflict, the degree of performance conflict can be characterized by the proportion of the duration of performance conflict in the entire running time. When characterizing the performance conflict degree, the persistent performance conflict has a greater impact on the performance of the container. Therefore, in order to improve the efficiency of determining the performance conflict degree, a continuous resource usage value is used instead of a single resource usage value in this embodiment.

A more intuitive way of characterizing performance conflicts is when the resource usage value exceeds the initial private resources allocated by the container. Then, when determining the performance conflict degree of the target container, the ratio between the time duration in which the continuous resource usage value in the target resource usage time series data is greater than the initial private resource and the total duration time in the target resource usage time series data is calculated, that is, the duration ratio is determined.

Exemplarily, S230 includes: determining that each resource usage peak value in the target resource usage time sequence data is greater than the peak length of the initial private resource; and generating first peak time sequence data according to each peak length, a preset peak length threshold, and the target resource usage time sequence data ; Determine the duration ratio of consecutive peaks in the first peak time series data.

The resource usage peak refers to the maximum value of resource usage within a certain time interval, and in the curve corresponding to the resource usage time series data, the resource usage peak is the peak value of each curve. The peak length refers to the value of the resource usage peak larger than the initial private resource. The preset peak length threshold refers to a preset peak length, which is used to filter all peak lengths to filter out data with smaller peak lengths. The first peak time series data refers to time series data that only retains resource usage data whose peak length exceeds a preset peak length threshold. A continuous peak refers to a set number of resource usage peaks where the time difference between every two resource usage peaks is less than a preset time threshold.

The above process of determining the duration ratio can be further improved as follows: determine each resource usage peak in the target resource usage time series data, calculate the difference between each resource usage peak and the initial private resource, and determine the peak length corresponding to each resource usage peak . Then, in the target resource usage time series data, each resource usage value whose peak length is greater than a preset peak length threshold is retained to generate first peak time series data. Finally, the ratio between the duration occupied by the continuous peaks in the first peak time series data and the total duration of the first peak time series data is counted, that is, the duration ratio is determined. The advantage of this setting is that, on the one hand, by filtering the resource usage value, the amount of data is reduced, the efficiency of determining the duration ratio is improved, and the adjustment efficiency of percentiles is further improved; The resource usage value can reduce the sensitivity of percentile adjustment, reduce the percentile adjustment operation with too small percentile adjustment range to a certain extent, and enhance the actual value of the percentile adjustment operation.

S240. Adjust the initial percentile according to the duration ratio, the preset ratio threshold, and the demand ratio to generate a target percentile.

The preset proportion threshold refers to a preset time proportion, such as 10%, which is used to represent the user's tolerance to the degree of performance conflict. The demand ratio is a preset duration ratio, which is used to represent the performance conflict degree of the container expected by the user. The demand ratio can be equal to the preset ratio threshold, such as 10%, or it can be unequal.

Determine whether and how to adjust the initial percentage by comparing the duration ratio with the preset ratio threshold. If the proportion of time duration is greater than the preset proportion threshold, the initial percentile is increased according to the proportion of demand, and the target percentile is generated. This situation shows that the private resources currently allocated for the target container cannot guarantee most of the performance of the target container. It is necessary to increase the configuration of the target private resources according to the relationship between the demand ratio and the duration ratio, so as to ensure the running performance of integration and services. If the proportion of time duration is less than the preset proportion threshold, the initial percentile is reduced according to the proportion of demand, and the target percentile is generated. This situation shows that there are many private resources allocated to the target container at present, which is wasted on the basis of ensuring most of the performance of the target container. It is necessary to reduce the configuration of the target private resources according to the relationship between the demand ratio and the duration ratio, so that Further saving resources.

S250. Determine the target private resource and the target spare resource of the target container according to the maximum resource configuration value and the target percentile.

S260: Determine a shared backup resource corresponding to the target container according to the target backup resource and the remaining backup resources of at least one other container except the target container in the physical machine corresponding to the target container.

S270. Configure the resources of the target container according to the target private resources and the shared spare resources.

In the technical solution of this embodiment, the time-series data of the target resource usage corresponding to the target container and the initial percentile are used to determine the percentage of the time duration in which the continuous resource usage value is greater than the initial private resource in the target resource usage time-series data, and according to the duration percentage, preset The proportion threshold and demand proportion, adjust the initial percentile, and generate the target percentile. The dynamic determination of the target percentile that is more objective and more suitable for the target container is realized, so that the determination of the target private resources and the target spare resources can also be more objective and more suitable for the target container, which further improves the flexibility of the resource configuration of the target container. properties, thereby further improving the resource utilization of the target container and the compatibility of performance guarantees.

Embodiment 3

On the basis of the above-mentioned first embodiment, this embodiment performs the steps of "determining the shared spare resources corresponding to the target container according to the target spare resources and the remaining spare resources of at least one other container except the target container in the physical machine corresponding to the target container". for further optimization. The explanations of terms that are the same as or corresponding to the above embodiments are not repeated here. Referring to FIG. 3a, the container resource configuration method provided by this embodiment includes:

S310. Obtain the maximum resource configuration value of the target container, and determine the target private resources and target backup resources of the target container according to the maximum resource configuration value and the target percentile.

S320. Determine the resource requirement correlation between the two containers according to the historical resource usage time series data corresponding to the target container and each of the remaining containers.

The resource requirement dependency refers to the dependency of the resource requirement change trend between two containers. If the resource requirements of two containers are highly correlated, it means that their resource requirements are relatively consistent. Conversely, if the resource requirements of two containers are relatively low, it means that their resource requirements are not correlated or anti-correlated, for example, when the resource requirements of one container are high, the resource requirements of the other container are low.

The determination of the shared spare resources needs to comprehensively consider the spare resources of the corresponding containers. In this embodiment, the comprehensive manner of multiple spare resources is determined based on the resource requirement correlation between the containers. Then, it is necessary to determine the resource requirement dependencies between the containers before determining the shared spare resources. Since the historical resource usage time series data can reflect the resource requirements of the container within a period of time, the resource requirement correlation in this embodiment can be calculated directly based on the historical resource usage time series data of the target container and any two of the other containers, For example, the covariance of formula (1) can be used to calculate:

分别表示相应容器的历史资源使用时序数据的资源使用均值，n表示历史资源使用时序数据中资源使用值的数量。Among them, COV(x, y) represents the covariance of the two containers, and x and y represent the historical resource usage time series data of the two containers, respectively.

Respectively represent the resource usage average of the historical resource usage time series data of the corresponding container, and n represents the number of resource usage values in the historical resource usage time series data.

The resource requirement correlation can also be calculated by using some feature data in the historical resource usage time series data. Since the peak usage rate (probability of peak occurrence) in the historical resource usage data can more clearly reflect the resource requirements of the container, the peak data in the historical resource usage time series data corresponding to the two containers can be used to calculate the difference between the two containers. Dependency between resource requirements.

Exemplarily, S320 includes: respectively generating corresponding second peak time series data according to the historical resource usage time series data corresponding to the target container and each remaining container; Resource requirement dependencies between two containers.

The second peak time series data refers to the time series data that retains the resource usage peak in the historical resource usage time series data.

Using peak data to determine resource demand correlation requires first obtaining the second peak time series data corresponding to each container. Specifically, filtering the container's historical resource usage time series data to eliminate non-peak resource usage values. In order to further improve the efficiency of determining the correlation of subsequent resource requirements and improve the reference value of the correlation of resource requirements, it is possible to further eliminate the resource usage peaks that do not exceed the target private resource among all resource usage peaks, because the resource usage that exceeds the target private resource The peak value is more indicative of the container's usage of spare resources. After that, the similarity between every two second peak time series data is calculated using the Jaccard coefficient such as formula (2), and the similarity can represent the resource requirement correlation between the two containers.

Among them, J(A, B) represents the resource demand correlation between the two containers, and A and B respectively represent the second peak time series data of the two containers.

The advantage of this setting is that it can improve the efficiency of determining the correlation of resource requirements and improve the reference value of the correlation of resource requirements.

S330. Classify the target container and each other container according to the correlation of each resource requirement and based on a clustering algorithm to obtain at least one container category.

According to the above characteristics of resource demand correlation, it can be seen that each container with consistent resource demand correlation has a higher probability of reaching peak performance in the same time period, while each container with inconsistent resource demand correlation has a lower probability of reaching performance peak in the same time period. , so the target container and each other container can be classified according to the correlation of resource requirements, so that the containers with consistent correlation are in the same container category, and the containers with inconsistent correlation are in different container categories, and then according to the corresponding container categories Each spare resource is used to determine the shared spare resource, so as to ensure that each container corresponding to the shared spare resource will not reach the peak performance at the same time, so that the shared spare resource has a high probability of being less than the sum of the spare resources of all containers, so as to greatly reduce the peak value. Resource requirements, reduce resource allocation as much as possible, and ensure the performance of related containers, minimizing container performance conflicts.

In a specific implementation, a clustering algorithm such as K-Means can be used to classify the target container and each remaining container based on the resource requirement correlation of each container to obtain at least one container category. The process of container classification based on the K-Means algorithm can be as follows:

1) Randomly select N containers from the containers as the initial "center points" of each category.

2) Calculate the Euclidean distance from each container to the N central points, and the distance is defined as the covariance COV of the historical resource usage time series data of the container and this central point container.

3) According to the Euclidean distance, the containers are classified into the closest categories, and each category is a container that is related to each other.

4) Update the center of each category, which is defined as the average of the historical resource usage time series data of all containers in the same category.

5) Continue with 2) to iterate until a certain number of times or specified error requirements are met.

S340. Determine the shared backup resource according to the sum of the backup resources corresponding to each container category.

The sum of the spare resources is the sum of the spare resources of each container in the corresponding container category.

First, determine the sum of the spare resources of all containers in each container class, then the sum of spare resources consistent with the number of container classes can be obtained. If the target container and the rest of the containers are in the container class, the sum of the spares is the sum of the target spare and each of the rest, and if there are no target containers in the container class, the sum of the spares is the sum of the rest of the spares. Then, determine the maximum value in the total sum of each spare resource, and determine the maximum value of the total sum of the spare resources as the shared spare resource of the target container and the other related containers.

If the number of container classes is at least two, the shared spare is less than the sum of the target spare and each of the remaining spares; if the number of container classes is one, the shared spare is equal to the sum of the target spare and each of the remaining spares sum.

Taking two containers as an example, if the resource requirements of the two containers have a low correlation (irrelevance or negative correlation), then their shared spare resources are equal to the maximum value of the respective spare resources of the two containers, that is, shared spare resources = MAX ( max ₁ –Tail_bound ₁ ,max ₂ –Tail_bound ₂ ), where max ₁ and max ₂ are the maximum resource configuration values of container 1 and container 2, respectively, and Tail_bound ₁ and Tail_bound ₂ are the resource percentile values of container 1 and container 2, respectively . If the resource requirements of two containers are highly correlated, then their shared spare resources are equal to the sum of their respective spare resources, ie shared spare resources=SUM(max ₁ -Tail_bound ₁ ,max ₂ -Tail_bound ₂ ).

According to the method of this embodiment, the resource configuration of two containers 1 and 2 with low resource requirements correlation can obtain the resource configuration result shown in FIG. 3b and the resource usage time series data curve of the two containers.

S350. Configure the resources of the target container according to the target private resources and the shared spare resources.

It should be noted that, the determination of the target percentile in this embodiment may also adopt the corresponding operations in the second embodiment.

Referring to Figure 3c, if resources are configured for all containers on a physical machine, the process of resource configuration based on the maximum historical resource usage and target percentile is roughly as follows:

A. Statistically analyze the historical resource usage data of a single container on the physical machine, and determine the maximum historical resource usage of a single container;

B. Configure the private resources of a single container. The determination of the private resources can be calculated according to the maximum historical resource usage and the target percentile;

C. Calculate the spare resources of a single container, and use the maximum historical resource usage and private resources of a single container to calculate its spare resources;

D. Statistically analyze the resource requirement correlation between all containers on the physical machine;

E. Perform container clustering analysis according to the resource requirement correlation between each two containers to obtain each container category;

共享备用资源是各容器类别的备用资源总和的最大值，即

F. According to the spare resources of each container and each container category, calculate and configure the shared spare resources of all containers on the physical machine. Here, the sum of spare resources for a container class is the sum of spare resources for each container in that container class, i.e.

The shared spare resource is the maximum value of the sum of spare resources of each container category, that is,

Based on the above description and Fig. 3d, when resources are configured for the 9 containers divided into 3 container categories, the result is that the total private resources configured by the 9 containers are C1+C2+...+C9, and the shared spare resources are max( H1+H2+H3, H4+H5+H6+H7, H8+H9).

Referring to Figure 3e, using the traditional percentile (90%)-based container resource allocation method (method 1), maximum value-based container resource allocation method (method 2), and average-based container resource allocation method (method 3) And the resource configuration method (method 4) of the present invention performs resource configuration of 18 containers, and compares the total amount of configuration resources (with the total amount of resource configuration at the 90% percentile as a standardized reference) and performance conflicts (the duration of performance conflicts). percentage) were compared, and the two comparison indicators are that the smaller the value, the better. Among them, although the total amount of resource allocation of method 4 is not the least, the total amount of resource allocation of method 4 is far less than that of method 2, and the performance conflict of method 4 is also small. Taken together, it is difficult for other methods to optimize resource usage while satisfying performance, and method 4 is the best in terms of ensuring container performance and saving resources at the same time.

In the technical solution of this embodiment, by determining the resource requirement correlation between two containers, and classifying the containers according to the resource requirement correlation, it is obtained that the resource requirement correlation of each container within the container category is consistent, and the resource requirement correlation between different container categories is obtained. The resource requirements of each container are inconsistent in each container category, and the shared spare resources are determined according to the sum of the spare resources corresponding to each container category, and the arrival time of the peak performance of each container is staggered. By determining the shared spare resources, the configuration of the shared spare resources can be further optimized by utilizing the resource requirement correlation of each container, so that the shared spare resources are smaller than the sum of the spare resources of each container, thereby ensuring the performance of each container to a greater extent and saving resources.

Embodiment 4

This embodiment provides an apparatus for configuring container resources. Referring to FIG. 4 , the apparatus specifically includes:

The target private resource determination module 410 is configured to obtain the maximum resource configuration value of the target container, and determine the target private resource and target backup resource of the target container according to the resource configuration maximum value and the target percentile;

The shared spare resource determination module 420 is configured to determine the shared spare resource corresponding to the target container according to the target spare resource and the remaining spare resources of at least one other container except the target container in the physical machine corresponding to the target container;

The container resource configuration module 430 is configured to configure the resources of the target container according to the target private resources and the shared spare resources.

Optionally, the target private resource determination module 410 includes:

The target private resource determination submodule is used to determine the target private resource of the target container according to the resource configuration maximum value and the target percentile;

The target backup resource determination sub-module is used for determining the target backup resource according to the maximum value of the resource configuration and the target private resource.

Optionally, on the basis of the above device, the device also includes a target percentile generation module for:

Before determining the target private resources and target spare resources of the target container according to the maximum resource configuration value and the target percentile, determine the target resource usage time series data and the initial percentile corresponding to the target container;

Determine the proportion of the time duration in which the continuous resource usage value in the target resource usage time series data is greater than the initial private resource, where the initial private resource is the private resource of the target container corresponding to the initial percentile;

Adjust the initial percentile to generate the target percentile according to the duration ratio, the preset ratio threshold, and the demand ratio.

Further, the target percentile generation module is specifically used for:

Determine the peak length of each resource usage peak in the target resource usage time series data that is greater than the initial private resource;

generating first peak time series data according to each peak length, preset peak length threshold and target resource usage time series data;

Determine the duration ratio of consecutive peaks in the first peak time series data.

Optionally, the shared spare resource determination module 420 is specifically configured to:

Determine the resource requirement correlation between the two containers according to the historical resource usage time series data corresponding to the target container and each other container;

According to the correlation of each resource requirement, based on the clustering algorithm, classify the target container and each other container, and obtain at least one container category;

The shared spare resources are determined according to the sum of spare resources corresponding to each container category, wherein the sum of spare resources is the sum of spare resources of each container in the corresponding container category.

Further, the shared spare resource determination module 420 is also specifically used for:

generating corresponding second peak time series data according to the target container and the historical resource usage time series data corresponding to each of the remaining containers respectively;

The similarity between every two second peak time series data is respectively determined as the resource requirement correlation between the corresponding two containers.

Optionally, the target private resource determination module 410 is specifically configured to:

Obtain the historical resource usage data corresponding to the target container, and determine the historical resource usage maximum value of the target container based on the historical resource usage data as the resource configuration maximum value.

Through the container resource configuration device according to the fourth embodiment of the present invention, a more rational configuration of container resources is realized, and resources are saved while providing performance assurance to a greater extent.

The container resource configuration device provided by the embodiment of the present invention can execute the container resource configuration method provided by any embodiment of the present invention, and has functional modules and beneficial effects corresponding to the execution method.

It is worth noting that in the above embodiments of the container resource configuration device, the units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be realized; in addition, The specific names of the functional units are only for the convenience of distinguishing from each other, and are not used to limit the protection scope of the present invention.

Embodiment 5

Referring to FIG. 5 , this embodiment provides a device, which includes: one or more processors 520 ; and a storage device 510 for storing one or more programs, when the one or more programs are processed by the one or more processors 520 executes, so that one or more processors 520 implement the container resource configuration method provided by the embodiment of the present invention, including:

Obtain the maximum resource configuration of the target container, and determine the target private resources and target spare resources of the target container according to the maximum resource configuration and target percentile;

Determine the shared spare resource corresponding to the target container according to the target spare resource and the remaining spare resources of at least one other container except the target container in the physical machine corresponding to the target container;

Configure the resources of the target container according to the target private resources and shared spare resources.

Of course, those skilled in the art can understand that the processor 520 may also implement the technical solution of the container resource configuration method provided by any embodiment of the present invention.

The device shown in FIG. 5 is only an example, and should not impose any limitation on the function and scope of use of the embodiments of the present invention. As shown in FIG. 5, the device includes a processor 520 and a storage device 510; the number of processors 520 in the device may be one or more, and one processor 520 is taken as an example in FIG. 5; the processor 520 in the device and the storage device The devices may be connected through a bus or in other ways. In FIG. 5 , the connection through the bus 530 is taken as an example.

As a computer-readable storage medium, the storage device 510 can be used to store software programs, computer-executable programs, and modules, such as program instructions/modules corresponding to the container resource configuration method in the embodiment of the present invention (for example, in the container resource configuration device). target private resource determination module, shared spare resource determination module and container resource configuration module).

The storage device 510 may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function; the storage data area may store data created according to the use of the terminal, and the like. Additionally, storage device 510 may include high-speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other non-volatile solid-state storage device. In some instances, storage device 510 may further include memory located remotely from processor 520, which may be connected to the device through a network. Examples of such networks include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

Embodiment 6

This embodiment provides a storage medium containing computer-executable instructions. The computer-executable instructions are used to execute a container resource configuration method when executed by a computer processor, and the method includes:

Obtain the maximum resource configuration of the target container, and determine the target private resources and target spare resources of the target container according to the maximum resource configuration and target percentile;

Determine the shared spare resource corresponding to the target container according to the target spare resource and the remaining spare resources of at least one other container except the target container in the physical machine corresponding to the target container;

Configure the resources of the target container according to the target private resources and shared spare resources.

Of course, a storage medium containing computer-executable instructions provided by an embodiment of the present invention, the computer-executable instructions of the storage medium are not limited to the above method operations, and can also execute the related container resource configuration methods provided by any embodiment of the present invention. operate.

From the above description of the embodiments, those skilled in the art can clearly understand that the present invention can be realized by software and necessary general-purpose hardware, and of course can also be realized by hardware, but in many cases the former is a better embodiment . Based on such understanding, the technical solutions of the present invention can be embodied in the form of software products in essence or the parts that make contributions to the prior art, and the computer software products can be stored in a computer-readable storage medium, such as a floppy disk of a computer , Read-Only Memory (ROM), Random Access Memory (RAM), Flash Memory (FLASH), hard disk or CD, etc., including several instructions to make a device (which can be a server, or a network equipment, etc.) to execute the container resource configuration methods provided by the various embodiments of the present invention.

Note that the above are only preferred embodiments of the present invention and applied technical principles. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the protection scope of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments, and can also include more other equivalent embodiments without departing from the concept of the present invention. The scope is determined by the scope of the appended claims.

Claims (10)

1. A container resource configuration method, comprising: Obtain the maximum resource configuration value of the target container, and determine the target private resource and target backup resource of the target container according to the resource configuration maximum value and the target percentile; Determine the shared backup resource corresponding to the target container according to the target backup resource and the remaining backup resources of at least one other container except the target container in the physical machine corresponding to the target container; The resources of the target container are configured according to the target private resource and the shared spare resource. 2 . The method according to claim 1 , wherein determining the target private resources and target spare resources of the target container according to the resource configuration maximum value and target percentile comprises: 2 . Determine the target private resource of the target container according to the resource configuration maximum value and the target percentile; The target spare resource is determined according to the maximum resource configuration value and the target private resource. 3. The method according to claim 1, wherein before the determining the target private resource and the target spare resource of the target container according to the resource configuration maximum value and the target percentile, the method further comprises: Determine the target resource usage time series data and the initial percentile corresponding to the target container; Determining the proportion of the duration of the continuous resource usage value greater than the initial private resource in the target resource usage time series data, wherein the initial private resource is the private resource of the target container corresponding to the initial percentile; The initial percentile is adjusted according to the duration ratio, the preset ratio threshold and the demand ratio to generate the target percentile. 4. The method according to claim 3, wherein the determining the proportion of the duration of the continuous resource usage value in the target resource usage timing data that is greater than the initial private resource comprises: determining that each resource usage peak value in the target resource usage time series data is greater than the peak length of the initial private resource; generating first peak time series data according to each of the peak lengths, the preset peak length thresholds, and the target resource usage time series data; Determine the duration ratio of continuous peaks in the first peak time series data. 5 . The method according to claim 1 , wherein the shared resource corresponding to the target container is determined according to the target backup resource and the remaining backup resources of at least one remaining container in the physical machine corresponding to the target container. 6 . Alternate resources include: According to the historical resource usage time series data corresponding to the target container and each of the remaining containers, determine the resource requirement correlation between the two containers; Classifying the target container and each of the remaining containers according to the correlation of each of the resource requirements and based on a clustering algorithm to obtain at least one container category; The shared spare resources are determined according to the sum of spare resources corresponding to each of the container classes, wherein the sum of spare resources is the sum of spare resources of each container in the corresponding container class. 6. The method according to claim 5, wherein, according to the historical resource usage time series data corresponding to the target container and each of the remaining containers, determining the resource requirement correlation between two containers comprises: generating corresponding second peak time series data according to the historical resource usage time series data corresponding to the target container and each of the remaining containers respectively; The similarity between each two of the second peak time series data is respectively determined as the resource requirement correlation between the corresponding pairs of containers. 7. The method according to claim 1, wherein the obtaining the maximum value of the resource configuration of the target container comprises: Acquire historical resource usage data corresponding to the target container, and determine a historical resource usage maximum value of the target container according to the historical resource usage data as the resource configuration maximum value. 8. An apparatus for configuring container resources, comprising: a target private resource determination module, configured to obtain the maximum resource configuration value of the target container, and determine the target private resource and target backup resource of the target container according to the resource configuration maximum value and the target percentile; A shared spare resource determination module, configured to determine the shared spare resource corresponding to the target container according to the target spare resource and the remaining spare resources of at least one other container except the target container in the physical machine corresponding to the target container ; A container resource configuration module, configured to configure resources of the target container according to the target private resource and the shared backup resource. 9. A device, characterized in that the device comprises: one or more processors; storage means for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the container resource configuration method according to any one of claims 1-7. 10. A computer-readable storage medium on which a computer program is stored, wherein the computer program implements the container resource configuration method according to any one of claims 1-7 when the computer program is executed by a processor.

Images