Tenant resource intelligent scheduling method based on load prediction

Abstract

The invention relates to the technical field of computer communication, and discloses a tenant resource intelligent scheduling method based on load prediction, and the method comprises the steps: arranging a plurality of initial monitoring points for each tenant on a cloud computing platform, and collecting a plurality of initial load indexes through the initial monitoring points; dividing each initial monitoring point, and calculating a load prediction coefficient of the tenant; obtaining a historical use condition of the tenant, and determining a historical load collapse value of the tenant according to the historical use condition; adjusting the load prediction coefficient of the tenant according to the historical load collapse value to obtain a target load prediction coefficient, and generating an intelligent resource scheduling strategy according to the target load prediction coefficient, thereby realizing accurate load prediction, generating a differentiated resource scheduling strategy, effectively avoiding insufficient or excessive resource allocation, and improving the resource scheduling efficiency. The resource utilization rate of the cloud computing platform is remarkably improved, the experience stability of the tenants is enhanced, the system crash risk is reduced, and the service quality of each tenant and stable operation of the cloud platform are guaranteed to the maximum extent.

Description

Technical Field

[0001] This invention relates to the field of computer communication technology, and more specifically, to a method for intelligent scheduling of tenant resources based on load prediction. Background Technology

[0002] With the rapid development of cloud computing technology, cloud service models represented by IaaS (Infrastructure as a Service), PaaS (Platform as a Service), and SaaS (Software as a Service) have become the preferred way for enterprises and individuals to obtain computing resources. In a multi-tenant cloud platform environment, how to efficiently, fairly, and intelligently schedule and allocate limited physical resources to meet the dynamically changing needs of numerous tenants is one of the core challenges faced by cloud service providers.

[0003] Traditional resource scheduling methods are mostly based on static allocation or simple dynamic threshold rules. For example, a fixed resource quota is pre-allocated to each tenant, or a fixed threshold is set, triggering a scheduling operation when monitoring data exceeds this threshold. While these methods are simple to implement, they have significant drawbacks: Low resource utilization: Static allocation cannot adapt to the dynamic fluctuations of tenant load. In order to ensure the stability of business peak periods, resources that far exceed their average demand are usually allocated, resulting in a large amount of idle resources during off-peak periods and causing huge waste.

[0004] Response lag and service quality degradation: Dynamic scheduling based on fixed thresholds has a lag effect. When the load increases sharply in a short period of time, the system may experience performance bottlenecks before reaching the threshold and completing resource expansion, leading to slower application response or even service crashes, which seriously affects the service agreement.

[0005] Ignoring individual tenant differences: A uniform scheduling strategy cannot adapt to these differences, and may over-allocate resources to tenants with low demand or stable demand, while failing to provide sufficient protection for tenants with high demand or volatile demand. Summary of the Invention

[0006] This invention provides a method for intelligent tenant resource scheduling based on load prediction. By collecting and analyzing load indicators from initial monitoring points, it achieves accurate load prediction and generates differentiated resource scheduling strategies, effectively avoiding insufficient or excessive resource allocation. This significantly improves the resource utilization of the cloud computing platform, enhances tenant experience stability, reduces the risk of system crashes, and maximizes the service quality for each tenant and the stable operation of the cloud platform.

[0007] To achieve the above objectives, the present invention provides a tenant resource intelligent scheduling method based on load prediction, comprising: On the cloud computing platform, several initial monitoring points are set up for each tenant, the cloud computing platform is continuously run, and several initial load indicators corresponding to several timestamps are collected through the initial monitoring points. The initial load metrics corresponding to each initial monitoring point are analyzed. Based on the analysis results, the initial monitoring points are divided, and the load prediction coefficients corresponding to the tenants are calculated based on the division results. Obtain the historical usage data for each tenant, and determine the historical load failure value corresponding to the tenant based on the historical usage data; The load prediction coefficient for each tenant is adjusted based on the historical load crash value to obtain the target load prediction coefficient, and a resource intelligent scheduling strategy is generated for the tenant based on the target load prediction coefficient.

[0008] Furthermore, before analyzing the initial load metrics corresponding to each initial monitoring point, the following steps are also included: All initial load metrics are parsed and processed, including deleting duplicate metrics, invalid metrics, and erroneous metrics.

[0009] Furthermore, when analyzing the initial load metrics corresponding to each initial monitoring point, dividing the initial monitoring points based on the analysis results, and calculating the load prediction coefficients for tenants based on the division results, the process includes: Determine the key initial monitoring points and auxiliary initial monitoring points based on the initial load indicators corresponding to each initial monitoring point; The load prediction coefficient for the tenant is calculated based on all key initial monitoring points.

[0010] Furthermore, when determining key initial monitoring points and auxiliary initial monitoring points based on the initial load metrics corresponding to each initial monitoring point, the following is included: The initial load index corresponding to each initial monitoring point is traversed, and all identical initial load indices are extracted as steady-state initial load indices, resulting in multiple identical steady-state initial load index sequences. The number of first steady-state initial load index sequences in the steady-state initial load index sequence is counted. Extract one steady-state initial load index from each of the steady-state initial load index sequences, and calculate the first steady-state initial load index and its value; Calculate the mean of all steady-state initial load indices, remove all steady-state initial load index sequences that are less than the mean of the indices, and count the number of second steady-state initial load index sequences in the remaining steady-state initial load index sequences. Extract one steady-state initial load index from the remaining steady-state initial load index sequence, and calculate the second steady-state initial load index and value; Calculate the steady-state index change coefficient corresponding to the initial monitoring point based on the number of the first steady-state initial load index sequence, the number of the second steady-state initial load index sequence, the sum of the first steady-state initial load index and the sum of the second steady-state initial load index; Extract the remaining distinct initial load indicators as fluctuating initial load indicators, sort all the changing initial load indicators based on timestamps, and calculate the fluctuation indicator change coefficient corresponding to the initial monitoring point. The weighted summation of the steady-state index change coefficient and the fluctuation index change coefficient yields the comprehensive index change coefficient. Calculate the mean value of the comprehensive indicator change coefficients corresponding to each initial monitoring point, and extract the comprehensive indicator change coefficients that are greater than or equal to the mean value of the coefficients, and take the corresponding initial monitoring points as key initial monitoring points. Use the remaining initial monitoring points as auxiliary initial monitoring points.

[0011] Further, when calculating the steady-state index change coefficient corresponding to the initial monitoring point based on the number of the first steady-state initial load index sequence, the number of the second steady-state initial load index sequence, the sum of the first steady-state initial load index and the sum of the second steady-state initial load index, it includes: The steady-state index variation coefficient is calculated using the following formula: ; Where q is the steady-state index change coefficient, w1 is the number of the first steady-state initial load index sequence, w2 is the number of the second steady-state initial load index sequence, e1 is the sum of the first steady-state initial load index, and e2 is the sum of the second steady-state initial load index.

[0012] Furthermore, when analyzing all the changing initial load indicators based on timestamps and calculating the fluctuation index change coefficient corresponding to the initial monitoring point, the process includes: The coefficient of change of the volatility index is calculated according to the following formula: ; Where p is the fluctuation index change coefficient, u is the number of initial fluctuation load indices, and y i Let y be the initial load index for the i-th fluctuation. i+1 This is the initial load index for the (i+1)th fluctuation.

[0013] Furthermore, when calculating the load prediction coefficients for tenants based on all key initial monitoring points, the following is included: By collecting several real-time resource consumption values ​​for each tenant through resource metering probes, the actual resource gap value can be obtained. The key initial monitoring points are traversed to determine several initial load indicators that have the same timestamp as the actual resource gap value, which are used as standard initial load indicators. Based on the timestamp order, every two standard initial load metrics are combined to obtain multiple standard initial load metric groups; Calculate the absolute value of the difference between the standard initial load indicators for each standard initial load indicator group, and extract the absolute value of the difference between the standard initial load indicators. Determine the maximum and minimum standard initial load indices from all standard initial load indices, and calculate the corresponding extreme standard initial load indices difference. The ratio of the absolute value of the difference between the extreme standard initial load index and the difference between the maximum standard initial load index is used as the load prediction coefficient for the tenant.

[0014] Furthermore, when calculating the load prediction coefficients for tenants based on all key initial monitoring points, the following is included: By collecting several real-time resource consumption values ​​for each tenant through resource metering probes, the actual resource gap value can be obtained. The key initial monitoring points are traversed to determine several initial load indicators that have the same timestamp as the actual resource gap value, which are used as standard initial load indicators. Based on the timestamp order, every two standard initial load metrics are combined to obtain multiple standard initial load metric groups; Calculate the absolute value of the difference between the standard initial load indicators for each standard initial load indicator group, and extract the absolute value of the difference between the standard initial load indicators. Determine the maximum and minimum standard initial load indices from all standard initial load indices, and calculate the corresponding extreme standard initial load indices difference. The ratio of the absolute value of the difference between the extreme standard initial load index and the difference between the maximum standard initial load index is used as the load prediction coefficient for the tenant.

[0015] Furthermore, when adjusting the load prediction coefficient for each tenant based on the historical load crash value to obtain the target load prediction coefficient, the process includes: Pre-set a first preset historical load crash value and a second preset historical load crash value; The first preset adjustment factor, the second preset adjustment factor, and the third preset adjustment factor are preset. When the historical load failure value is less than the first preset historical load failure value, the first product of the first preset adjustment factor and the load prediction coefficient is calculated as the target load prediction coefficient. When the historical load failure value is greater than or equal to the first preset historical load failure value and less than the second preset historical load failure value, the second product of the second preset adjustment factor and the load prediction coefficient is calculated as the target load prediction coefficient. When the historical load collapse value is greater than or equal to the second preset historical load collapse value, the third product of the third preset adjustment factor and the load prediction coefficient is calculated as the target load prediction coefficient.

[0016] Furthermore, when generating a resource intelligent scheduling strategy for the tenant based on the target load prediction coefficient, the process includes: Set the upper and lower limits of the load threshold for the cloud computing platform; When the target load prediction coefficient is greater than or equal to the upper limit of the load threshold, an intelligent scheduling strategy for expansion resources is generated for the tenant. When the target load prediction coefficient is less than or equal to the lower limit of the load threshold, a shrinkage resource intelligent scheduling strategy is generated for the tenant. When the target load prediction coefficient is greater than the lower limit of the load threshold and less than the upper limit of the load threshold, a resource intelligent scheduling strategy to maintain scaling up is generated for the tenant.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention discloses a method for intelligent scheduling of tenant resources based on load prediction. On a cloud computing platform, several initial monitoring points are deployed for each tenant, and several initial load indicators are collected through these monitoring points. Each initial monitoring point is divided, and the tenant's load prediction coefficient is calculated. Historical usage data of the tenant is obtained, and the tenant's historical load collapse value is determined based on this data. The tenant's load prediction coefficient is adjusted according to the historical load collapse value to obtain a target load prediction coefficient. An intelligent resource scheduling strategy is generated based on the target load prediction coefficient, achieving accurate load prediction and generating differentiated resource scheduling strategies. This effectively avoids insufficient or excessive resource allocation, significantly improves the resource utilization rate of the cloud computing platform, enhances tenant experience stability, reduces system crash risk, and maximizes the guarantee of service quality for each tenant and the stable operation of the cloud platform. Attached Figure Description

[0018] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 The diagram illustrates a flowchart of an intelligent scheduling method for tenant resources based on load prediction, according to an embodiment of the present invention. Detailed Implementation

[0019] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0020] In the description of this application, it should be understood that the terms "center", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0021] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0022] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0023] The following is a description of preferred embodiments of the present invention in conjunction with the accompanying drawings.

[0024] like Figure 1 As shown, an embodiment of the present invention discloses a method for intelligent scheduling of tenant resources based on load prediction, comprising: S110: Deploy several initial monitoring points for each tenant on the cloud computing platform, continuously run the cloud computing platform, and collect several initial load indicators corresponding to several timestamps through the initial monitoring points. In this embodiment, the initial number of monitoring points is preferably 10, but this can be adjusted according to actual needs. They can be evenly distributed on the computer platform.

[0025] In this embodiment, the timestamps are preset, such as the 2nd minute, the 4th minute, the 6th minute, the 8th minute, the 10th minute, the 12th minute, the 14th minute, the 16th minute, the 18th minute, and the 20th minute.

[0026] In this embodiment, the initial load metric is CPU utilization.

[0027] S120: Analyze the initial load index corresponding to each initial monitoring point, divide the initial monitoring points based on the analysis results, and calculate the load prediction coefficient corresponding to the tenant based on the division results. In some embodiments of this application, before analyzing the initial load metrics corresponding to each initial monitoring point, the following steps are also included: All initial load metrics are parsed and processed, including deleting duplicate metrics, invalid metrics, and erroneous metrics.

[0028] The beneficial effects of the above technical solution are: the present invention can delete duplicate indicators, invalid indicators and erroneous indicators, thus ensuring the basic accuracy of the data.

[0029] In some embodiments of this application, when analyzing the initial load index corresponding to each initial monitoring point, dividing the initial monitoring points based on the analysis results, and calculating the load prediction coefficient corresponding to the tenant based on the division results, the process includes: Determine the key initial monitoring points and auxiliary initial monitoring points based on the initial load indicators corresponding to each initial monitoring point; The load prediction coefficient for the tenant is calculated based on all key initial monitoring points.

[0030] In some embodiments of this application, determining key initial monitoring points and auxiliary initial monitoring points based on the initial load metrics corresponding to each initial monitoring point includes: The initial load index corresponding to each initial monitoring point is traversed, and all identical initial load indices are extracted as steady-state initial load indices, resulting in multiple identical steady-state initial load index sequences. The number of first steady-state initial load index sequences in the steady-state initial load index sequence is counted. Extract one steady-state initial load index from each of the steady-state initial load index sequences, and calculate the first steady-state initial load index and its value; Calculate the mean of all steady-state initial load indices, remove all steady-state initial load index sequences that are less than the mean of the indices, and count the number of second steady-state initial load index sequences in the remaining steady-state initial load index sequences. Extract one steady-state initial load index from the remaining steady-state initial load index sequence, and calculate the second steady-state initial load index and value; Calculate the steady-state index change coefficient corresponding to the initial monitoring point based on the number of the first steady-state initial load index sequence, the number of the second steady-state initial load index sequence, the sum of the first steady-state initial load index and the sum of the second steady-state initial load index; Extract the remaining distinct initial load indicators as fluctuating initial load indicators, sort all the changing initial load indicators based on timestamps, and calculate the fluctuation indicator change coefficient corresponding to the initial monitoring point. The weighted summation of the steady-state index change coefficient and the fluctuation index change coefficient yields the comprehensive index change coefficient. Calculate the mean value of the comprehensive indicator change coefficients corresponding to each initial monitoring point, and extract the comprehensive indicator change coefficients that are greater than or equal to the mean value of the coefficients, and take the corresponding initial monitoring points as key initial monitoring points. Use the remaining initial monitoring points as auxiliary initial monitoring points.

[0031] In this embodiment, for ease of understanding, an example is given. For instance, the initial load index corresponding to the initial monitoring point is {0.2, 0.2, 0.3, 0.4, 0.5, 0.5, 0.5, 0.6, 0.6, 0.7}. The same steady-state initial load index sequence is {0.2, 0.2}, {0.5, 0.5, 0.5}, and {0.6, 0.6}. The first steady-state initial load index sequence has 3 values, and one steady-state initial load index is extracted from each of these sequences, specifically 0.2, 0.5, and 0.6. The average value of all steady-state initial load indices is 0.45. Therefore, the remaining steady-state initial load index sequences are {0.5, 0.5, 0.5} and {0.6, 0.6}. The second steady-state initial load index sequence has 2 values, and one steady-state initial load index is extracted from each of these sequences, specifically 0.5 and 0.6. This is an example and not a specific limitation.

[0032] In this embodiment, the remaining different initial load indices are 0.3, 0.4, and 0.7.

[0033] In this embodiment, the steady-state index change coefficient and the fluctuation index change coefficient can be weighted based on the subjective weighting method or the objective weighting method. Here, the weight of the steady-state index change coefficient is preferably 0.7, and the weight of the fluctuation index change coefficient is preferably 0.3.

[0034] In this embodiment, after the above calculation, the comprehensive index change coefficient corresponding to each initial monitoring point can be obtained, and then the mean value can be calculated as the coefficient mean value.

[0035] The beneficial effects of the above technical solution are as follows: This invention calculates the steady-state index change coefficient corresponding to the initial monitoring point based on the number of first steady-state initial load index sequences, the number of second steady-state initial load index sequences, the sum of the first steady-state initial load index and the sum of the second steady-state initial load index, thus realizing the stability analysis of the initial load index. Based on the timestamp, all the changing initial load indices are sorted, and the fluctuation index change coefficient corresponding to the initial monitoring point is calculated, thus realizing the volatility analysis of the initial load index. The steady-state index change coefficient and the fluctuation index change coefficient are weighted and summed to obtain the comprehensive index change coefficient. Then, based on the comprehensive index change coefficient, key initial monitoring points and auxiliary initial monitoring points are divided, which can lay the foundation for tenant load analysis, remove initial monitoring points with large noise, and ensure the accuracy of tenant load analysis.

[0036] In some embodiments of this application, when calculating the steady-state index change coefficient corresponding to the initial monitoring point based on the number of the first steady-state initial load index sequence, the number of the second steady-state initial load index sequence, the sum of the first steady-state initial load index and the sum of the second steady-state initial load index, the calculation includes: The steady-state index variation coefficient is calculated using the following formula: ; Where q is the steady-state index change coefficient, w1 is the number of the first steady-state initial load index sequence, w2 is the number of the second steady-state initial load index sequence, e1 is the sum of the first steady-state initial load index, and e2 is the sum of the second steady-state initial load index.

[0037] In some embodiments of this application, when analyzing all changing initial load indicators based on timestamps and calculating the fluctuation index change coefficient corresponding to the initial monitoring point, the following steps are included: The coefficient of change of the volatility index is calculated according to the following formula: ; Where p is the fluctuation index change coefficient, u is the number of initial fluctuation load indices, and y i Let y be the initial load index for the i-th fluctuation. i+1 This is the initial load index for the (i+1)th fluctuation.

[0038] In some embodiments of this application, the calculation of the load prediction coefficient corresponding to a tenant based on all key initial monitoring points includes: By collecting several real-time resource consumption values ​​for each tenant through resource metering probes, the actual resource gap value can be obtained. The key initial monitoring points are traversed to determine several initial load indicators that have the same timestamp as the actual resource gap value, which are used as standard initial load indicators. Based on the timestamp order, every two standard initial load metrics are combined to obtain multiple standard initial load metric groups; Calculate the absolute value of the difference between the standard initial load indicators for each standard initial load indicator group, and extract the absolute value of the difference between the standard initial load indicators. Determine the maximum and minimum standard initial load indices from all standard initial load indices, and calculate the corresponding extreme standard initial load indices difference. The ratio of the absolute value of the difference between the extreme standard initial load index and the difference between the maximum standard initial load index is used as the load prediction coefficient for the tenant.

[0039] In this embodiment, the resource metering probe refers to a lightweight data collection agent program that resides on the cloud computing platform.

[0040] In this embodiment, the real-time resource consumption value refers to the CPU cores per second (vCPU·s) actually occupied by the tenant at the current sampling time, such as a real-time resource consumption value of 0.8vCPU·s. The actual resource gap value refers to the difference between the CPU cores per second required to ensure the tenant's service quality and the real-time resource consumption value at the same sampling time. Actual resource gap value = CPU cores per second required for service quality minus real-time resource consumption value. Example: If the CPU cores per second required for service quality is 1.0vCPU·s and the real-time resource consumption value is 0.8vCPU·s, then the actual resource gap value is 0.2.

[0041] In this embodiment, in reality, there may be a situation where the actual resource gap value is 0. In this case, the CPU cores per second required for service quality is equal to the real-time resource consumption value. Therefore, to avoid errors, several initial load indicators corresponding to the timestamp of the actual resource gap value of 0 are not counted. When the actual resource gap value is 0.2, several initial load indicators with the same timestamp as the actual resource gap value are determined as standard initial load indicators.

[0042] In this embodiment, when combining any two standard initial load indicators, they can be combined sequentially based on the timestamp order. If there is a single standard initial load indicator that is not combined, it can be deleted and will not be included in the calculation.

[0043] In this embodiment, the absolute value of the difference between standard initial load indicators refers to the absolute value of the difference between two standard initial load indicators in each standard initial load indicator group.

[0044] The beneficial effects of the above technical solution are: the present invention uses the ratio of the absolute value of the difference between the extreme standard initial load index and the maximum standard initial load index as the load prediction coefficient of the tenant, which ensures the accuracy of the determination of the load prediction coefficient, thereby providing a reliable scheduling basis for intelligent scheduling of tenant resources and ensuring the accuracy of intelligent scheduling of tenant resources.

[0045] S130: Obtain the historical usage information for each tenant, and determine the historical load failure value corresponding to the tenant based on the historical usage information; In some embodiments of this application, when obtaining the historical usage information of each tenant and determining the historical load collapse value corresponding to the tenant based on the historical usage information, the process includes: The historical usage data is analyzed to determine the number of historical crashes and the total number of historical usages of the cloud computing platform; The ratio of the number of historical crashes to the total number of historical uses is used as the historical load crash value for the tenant.

[0046] In this embodiment, if a tenant uses a cloud computing platform and the platform crashes due to load conditions, it is counted as one crash.

[0047] The beneficial effects of the above technical solution are: historical load failure values ​​can reflect situations where tenants experienced severe load discrepancies in their historical usage, providing reliable technical support for intelligent tenant scheduling.

[0048] S140: Adjust the load prediction coefficient of each tenant according to the historical load crash value to obtain the target load prediction coefficient, and generate a resource intelligent scheduling strategy for the tenant according to the target load prediction coefficient.

[0049] In some embodiments of this application, adjusting the load prediction coefficient for each tenant based on the historical load crash value to obtain the target load prediction coefficient includes: Pre-set a first preset historical load crash value and a second preset historical load crash value; The first preset adjustment factor, the second preset adjustment factor, and the third preset adjustment factor are preset. When the historical load failure value is less than the first preset historical load failure value, the first product of the first preset adjustment factor and the load prediction coefficient is calculated as the target load prediction coefficient. When the historical load failure value is greater than or equal to the first preset historical load failure value and less than the second preset historical load failure value, the second product of the second preset adjustment factor and the load prediction coefficient is calculated as the target load prediction coefficient. When the historical load collapse value is greater than or equal to the second preset historical load collapse value, the third product of the third preset adjustment factor and the load prediction coefficient is calculated as the target load prediction coefficient.

[0050] In this embodiment, the first preset historical load crash value is preferably 0.2, and the second preset historical load crash value is preferably 0.6. The specific values ​​can be adjusted adaptively according to the actual situation.

[0051] In this embodiment, the first preset adjustment factor is preferably 0.95, the second preset adjustment factor is preferably 1.15, and the third preset adjustment factor is preferably 1.25. The specific values ​​can be adjusted adaptively according to the actual situation.

[0052] The beneficial effects of the above technical solution are as follows: Based on the relationship between historical load collapse value, first preset historical load collapse value and second preset historical load collapse value, the present invention selects a corresponding preset adjustment factor to adjust the load prediction coefficient, thereby realizing the dynamic adjustment of the load prediction coefficient and obtaining the target load prediction coefficient. This ensures the accuracy and comprehensiveness of the target load prediction coefficient, further providing accurate scheduling basis for intelligent scheduling of tenant resources, and ensuring the accuracy and efficiency of intelligent scheduling of tenant resources.

[0053] In some embodiments of this application, generating a resource intelligent scheduling strategy for the tenant based on the target load prediction coefficient includes: Set the upper and lower limits of the load threshold for the cloud computing platform; When the target load prediction coefficient is greater than or equal to the upper limit of the load threshold, an intelligent scheduling strategy for expansion resources is generated for the tenant. When the target load prediction coefficient is less than or equal to the lower limit of the load threshold, a shrinkage resource intelligent scheduling strategy is generated for the tenant. When the target load prediction coefficient is greater than the lower limit of the load threshold and less than the upper limit of the load threshold, a resource intelligent scheduling strategy to maintain scaling up is generated for the tenant.

[0054] In this embodiment, the upper limit of the load threshold is preferably 6, and the lower limit of the load threshold is preferably 3. The specific values ​​can be adjusted according to actual needs.

[0055] The beneficial effects of the above technical solution are as follows: This invention generates intelligent resource scheduling strategies based on target load prediction coefficients, achieving accurate load prediction, generating differentiated resource scheduling strategies, significantly improving the resource utilization of the cloud computing platform, enhancing tenant experience stability, reducing the risk of system crashes, and maximizing the guarantee of service quality for each tenant and the stable operation of the cloud platform. At the same time, it avoids the problem of static allocation scheduling, which avoids over-allocating resources to tenants with low or stable demand while failing to provide sufficient guarantees for tenants with high demand or volatile demand.

[0056] In the description of the above embodiments, specific features, structures, materials, or characteristics may be combined in any suitable manner in one or more embodiments or examples.

[0057] Although the invention has been described above with reference to embodiments, various modifications can be made and components can be replaced with equivalents without departing from the scope of the invention. In particular, as long as there is no structural conflict, the features in the embodiments disclosed in this invention can be combined with each other in any way. The fact that not all of these combinations are described in this specification is merely for the sake of brevity and resource conservation.

[0058] It will be understood by those skilled in the art that the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for intelligent scheduling of tenant resources based on load prediction, characterized in that, include: On the cloud computing platform, several initial monitoring points are set up for each tenant, the cloud computing platform is continuously run, and several initial load indicators corresponding to several timestamps are collected through the initial monitoring points. The initial load metrics corresponding to each initial monitoring point are analyzed. Based on the analysis results, the initial monitoring points are divided, and the load prediction coefficients corresponding to the tenants are calculated based on the division results. Obtain the historical usage data for each tenant, and determine the historical load failure value corresponding to the tenant based on the historical usage data; The load prediction coefficient for each tenant is adjusted based on the historical load crash value to obtain the target load prediction coefficient, and a resource intelligent scheduling strategy is generated for the tenant based on the target load prediction coefficient.

2. The intelligent scheduling method for tenant resources based on load prediction according to claim 1, characterized in that, Before analyzing the initial load metrics corresponding to each initial monitoring point, the following is also included: All initial load metrics are parsed and processed, including deleting duplicate metrics, invalid metrics, and erroneous metrics.

3. The intelligent scheduling method for tenant resources based on load prediction according to claim 1, characterized in that, When analyzing the initial load metrics corresponding to each initial monitoring point, dividing the initial monitoring points based on the analysis results, and calculating the load prediction coefficient for each tenant based on the division results, the process includes: Determine the key initial monitoring points and auxiliary initial monitoring points based on the initial load indicators corresponding to each initial monitoring point; The load prediction coefficient for the tenant is calculated based on all key initial monitoring points.

4. The intelligent scheduling method for tenant resources based on load prediction according to claim 3, characterized in that, When determining key initial monitoring points and auxiliary initial monitoring points based on the initial load metrics corresponding to each initial monitoring point, the following are included: The initial load index corresponding to each initial monitoring point is traversed, and all identical initial load indices are extracted as steady-state initial load indices, resulting in multiple identical steady-state initial load index sequences. The number of first steady-state initial load index sequences in the steady-state initial load index sequence is counted. Extract one steady-state initial load index from each of the steady-state initial load index sequences, and calculate the first steady-state initial load index and its value; Calculate the mean of all steady-state initial load indices, remove all steady-state initial load index sequences that are less than the mean of the indices, and count the number of second steady-state initial load index sequences in the remaining steady-state initial load index sequences. Extract one steady-state initial load index from the remaining steady-state initial load index sequence, and calculate the second steady-state initial load index and value; Calculate the steady-state index change coefficient corresponding to the initial monitoring point based on the number of the first steady-state initial load index sequence, the number of the second steady-state initial load index sequence, the sum of the first steady-state initial load index and the sum of the second steady-state initial load index; Extract the remaining distinct initial load indicators as fluctuating initial load indicators, sort all the changing initial load indicators based on timestamps, and calculate the fluctuation indicator change coefficient corresponding to the initial monitoring point. The weighted summation of the steady-state index change coefficient and the fluctuation index change coefficient yields the comprehensive index change coefficient. Calculate the mean value of the comprehensive indicator change coefficients corresponding to each initial monitoring point, and extract the comprehensive indicator change coefficients that are greater than or equal to the mean value of the coefficients, and take the corresponding initial monitoring points as key initial monitoring points. Use the remaining initial monitoring points as auxiliary initial monitoring points.

5. The intelligent scheduling method for tenant resources based on load prediction according to claim 4, characterized in that, When calculating the steady-state index change coefficient corresponding to the initial monitoring point based on the number of the first steady-state initial load index sequence, the number of the second steady-state initial load index sequence, the sum of the first steady-state initial load index and the sum of the second steady-state initial load index, the calculation includes: The steady-state index variation coefficient is calculated using the following formula: ； Where q is the steady-state index change coefficient, w1 is the number of the first steady-state initial load index sequence, w2 is the number of the second steady-state initial load index sequence, e1 is the sum of the first steady-state initial load index, and e2 is the sum of the second steady-state initial load index.

6. The intelligent scheduling method for tenant resources based on load prediction according to claim 4, characterized in that, When analyzing all changing initial load indicators based on timestamps and calculating the fluctuation index change coefficient corresponding to the initial monitoring point, the following is included: The coefficient of change of the volatility index is calculated according to the following formula: ； Where p is the fluctuation index change coefficient, u is the number of initial fluctuation load indices, and y i Let y be the initial load index for the i-th fluctuation. i+1 This is the initial load index for the (i+1)th fluctuation.

7. The intelligent scheduling method for tenant resources based on load prediction according to claim 3, characterized in that, When calculating the load prediction coefficient for a tenant based on all key initial monitoring points, the following is included: By collecting several real-time resource consumption values ​​for each tenant through resource metering probes, the actual resource gap value can be obtained. The key initial monitoring points are traversed to determine several initial load indicators that have the same timestamp as the actual resource gap value, which are used as standard initial load indicators. Based on the timestamp order, every two standard initial load metrics are combined to obtain multiple standard initial load metric groups; Calculate the absolute value of the difference between the standard initial load indicators for each standard initial load indicator group, and extract the absolute value of the difference between the standard initial load indicators. Determine the maximum and minimum standard initial load indices from all standard initial load indices, and calculate the corresponding extreme standard initial load indices difference. The ratio of the absolute value of the difference between the extreme standard initial load index and the difference between the maximum standard initial load index is used as the load prediction coefficient for the tenant.

8. The intelligent scheduling method for tenant resources based on load prediction according to claim 1, characterized in that, When acquiring the historical usage data for each tenant and determining the historical load collapse value corresponding to that tenant based on the historical usage data, the process includes: The historical usage data is analyzed to determine the number of historical crashes and the total number of historical usages of the cloud computing platform; The ratio of the number of historical crashes to the total number of historical uses is used as the historical load crash value for the tenant.

9. The intelligent scheduling method for tenant resources based on load prediction according to claim 1, characterized in that, When adjusting the load prediction coefficient for each tenant based on the historical load crash values ​​to obtain the target load prediction coefficient, the following steps are included: Pre-set a first preset historical load crash value and a second preset historical load crash value; The first preset adjustment factor, the second preset adjustment factor, and the third preset adjustment factor are preset. When the historical load failure value is less than the first preset historical load failure value, the first product of the first preset adjustment factor and the load prediction coefficient is calculated as the target load prediction coefficient. When the historical load failure value is greater than or equal to the first preset historical load failure value and less than the second preset historical load failure value, the second product of the second preset adjustment factor and the load prediction coefficient is calculated as the target load prediction coefficient. When the historical load collapse value is greater than or equal to the second preset historical load collapse value, the third product of the third preset adjustment factor and the load prediction coefficient is calculated as the target load prediction coefficient.

10. The intelligent scheduling method for tenant resources based on load prediction according to claim 1, characterized in that, When generating a resource intelligent scheduling strategy for the tenant based on the target load prediction coefficient, the following steps are included: Set the upper and lower limits of the load threshold for the cloud computing platform; When the target load prediction coefficient is greater than or equal to the upper limit of the load threshold, an intelligent scheduling strategy for expansion resources is generated for the tenant. When the target load prediction coefficient is less than or equal to the lower limit of the load threshold, a shrinkage resource intelligent scheduling strategy is generated for the tenant. When the target load prediction coefficient is greater than the lower limit of the load threshold and less than the upper limit of the load threshold, a resource intelligent scheduling strategy to maintain scaling up is generated for the tenant.

Images