Multi-dimensional cloud resource waste identification and specification recommendation method

Abstract

The invention discloses a multi-dimensional cloud resource waste identification and specification recommendation method. The method specifically comprises the following steps: step 1, collecting multi-dimensional resource data of a cloud environment and preprocessing the multi-dimensional resource data; 2, performing resource type identification on the preprocessed multi-dimensional resource data, and dividing the multi-dimensional resource data into load fluctuation type resource data, storage type resource data and special type resource data; and step 3, performing resource waste identification on the load fluctuation type resource data, the storage type resource data and the special type resource data to obtain corresponding resource waste state information, and determining specification recommendation suggestion information of the load fluctuation type resource, the storage type resource and the special type resource based on the resource waste state information. According to the method, the cloud resource waste conditions of different dimensions can be accurately identified, and reasonable change suggestions are given.

Description

Technical Field

[0001] The present invention belongs to the technical field of cloud computing, and in particular relates to a multi-dimensional cloud resource waste identification and specification recommendation method. Background Art

[0002] Cloud computing has become an important tool for enterprises and organizations to achieve flexibility, scalability and efficiency. The cost of cloud services has become an important consideration. Currently, the cost of using cloud resources is mainly reduced by monitoring and analyzing the use of cloud resources and then making corresponding adjustments and optimizations.

[0003] Existing cloud service cost management tools often only focus on the basic usage and costs of cloud resources, without considering the differences between different types of cloud resources, and cannot provide accurate cost forecasts and optimization suggestions. In addition, the optimization methods for cloud resource costs are relatively limited, lacking a comprehensive solution for multi-dimensional cloud resource optimization and cloud resource allocation and scheduling, and cannot effectively improve the efficiency and reliability of cost optimization of different types of cloud resources. Summary of the invention

[0004] The technical problem to be solved by the present invention is to address the deficiencies of the above-mentioned prior art and provide a multi-dimensional cloud resource waste identification and specification recommendation method. By calculating and analyzing historical monitoring data, it can accurately identify cloud resource waste in different dimensions and give reasonable change suggestions.

[0005] In order to achieve the above technical objectives, the technical solution adopted by the present invention is:

[0006] The multi-dimensional cloud resource waste identification and specification recommendation method specifically includes the following steps:

[0007] Step 1: Collect multi-dimensional resource data of the cloud environment and pre-process the multi-dimensional resource data;

[0008] Step 2: Identify the resource types of the preprocessed multi-dimensional resource data, and classify the multi-dimensional resource data into load fluctuation resource data, storage resource data, and special resource data;

[0009] Step 3: Identify resource waste for load fluctuation resource data, storage resource data and special resource data respectively, obtain corresponding resource waste status information, and determine the specification recommendation information for load fluctuation resource, storage resource and special resource respectively based on the resource waste status information.

[0010] To optimize the above technical solutions, the specific measures taken also include:

[0011] The above step 1 collects multi-dimensional resource data of the cloud environment, including:

[0012] Based on the location information of all resource terminals within the cloud environment, all resource terminals are independently monitored to obtain the cloud resource usage data corresponding to each resource terminal;

[0013] The cloud resource usage data is identified and integrated with the resource-side identity information to obtain multi-dimensional resource data.

[0014] The multi-dimensional resource data in step 1 above includes the load of the application instance, resource usage and network status.

[0015] The pre-processing process described in step 1 above is as follows:

[0016] Identify missing values, outliers and invalid information, record the data missing location points, data outlier location points and data invalid location points, and use the data missing location points, data outlier location points and data invalid location points as the basis for data completion, data correction and data deletion.

[0017] The load-fluctuating resource data in step 2 above indicates fluctuations in demand for computing resources due to changes in workload in different time periods, including CPU, memory, network, and floating IP data; the storage resource data indicates resources that provide persistent data storage in a cloud computing environment, including cloud hard disk data; the special resource data indicates resources designed for fixed use cases or service models in cloud computing, including cloud databases, load balancing, object storage, file storage, and middleware data.

[0018] In step 3 above, for load-fluctuating resource data, the average and 95th percentile values ​​of the monitoring data utilization are calculated and combined with the corresponding pre-set thresholds to perform a waste identification strategy; when waste is identified, the waterfall model is used to perform multi-level specification upgrades or downgrades.

[0019] In the above step 3, for storage resource data, when its storage capacity is higher than 90%, historical monitoring data is used to predict capacity. If the predicted capacity value is higher than 100% of the current capacity, it is identified as efficiency waste. When waste is identified, the predicted capacity value is used to recommend expansion specifications.

[0020] In the above step 3, for special resource data, different data types are processed separately, as follows:

[0021] (1) Cloud database: The actual CPU utilization, actual memory utilization, and actual space utilization are comprehensively considered to identify waste. When waste is identified, CPU and memory specifications are recommended according to the waterfall model, and storage space is directly scaled up or down.

[0022] (2) Load balancing: If the number of LCUs currently purchased is greater than 120% of the number of LCUs expected to be used, the waste is identified and the appropriate instance function version is selected based on business requirements; the number of LCUs expected to be used = max{maximum number of new connections per second / 25, maximum number of concurrent connections per minute / 3000, total amount of data processed per hour / 1GB, number of rule evaluations / 1000};

[0023] (3) Object storage: waste is identified by judging the relationship between the cumulative number of requests and the cumulative number of requests threshold of this gradient. According to different thresholds, storage is classified into standard, low-frequency access, or archive storage solutions. Storage costs are ranked from high to low, and data retrieval and request costs are ranked from low to high.

[0024] (4) File storage: If the total storage volume of small files exceeds 30%, it is considered to be a waste. A small file storage optimization solution is provided to users to merge related small files into large files to reduce the number of storage blocks and metadata storage. The recommended single file size is 10MB to 100MB to reduce metadata overhead.

[0025] (5) Middleware: When the CPU usage of an instance is less than 20% and the memory usage is less than 30%, it is considered a waste and a waterfall model with multi-level downgrading is adopted. When the load is less than 10% for seven consecutive days, it is considered a waste and the instance is automatically dormant. The resources are restored when activated. When the maximum number of connections is more than 150% of the actual demand, it is considered a waste and the maximum number of connections is set to 150%.

[0026] The present invention has the following beneficial effects:

[0027] 1. The present invention collects multi-dimensional resource data of the cloud environment, pre-processes the multi-dimensional resource data, and identifies the types of the pre-processed multi-dimensional resource data, and divides the multi-dimensional resource data into load fluctuation type resource data, storage type resource data, and special type resource data; and also identifies the resource waste of the load fluctuation type resource data, storage type resource data, and special type resource data, respectively, and obtains the corresponding resource waste state information, thereby determining the specification recommendation information for the load fluctuation type resource, storage type resource, and special type resource, which covers the cost optimization needs of various cloud resources in the cloud computing environment, can take care of the needs of the entire cloud computing ecosystem, more accurately analyze the cost waste of multi-dimensional resources, and provide adaptive optimization strategies and specification recommendations for different situations, thereby improving the accuracy and flexibility of cost optimization. It can perform targeted cost optimization on different types of cloud resources under the multi-dimensional resource data, and improve the accuracy of specification recommendations for each type of cloud resource.

[0028] 2. The present invention independently monitors all resource terminals based on their respective location information within the cloud environment, and can accurately and comprehensively grasp the usage status of different types of cloud resources within the cloud computing environment, and obtain cloud resource usage data corresponding to each resource terminal. The cloud resource usage data may be but is not limited to cloud resource usage, etc. The cloud resource usage data is identified and integrated with the resource terminal identity information to obtain multi-dimensional resource data, so that all types of cloud resources involved in the cloud computing environment can be differentiated and identified and summarized, which is convenient for subsequent targeted and differentiated cost optimization of multi-dimensional resource data and the proposal of applicable and accurate specification recommendations.

[0029] 3. The present invention identifies and processes missing data values, abnormal data values ​​and invalid data values ​​for each cloud resource usage data under the multi-dimensional resource data, and obtains the data missing position points, abnormal data position points and invalid data position points within the cloud resource usage data. In this way, the multi-dimensional resource data can be preprocessed with data completion, data correction and data deletion based on the missing data position points, abnormal data position points and invalid data position points, respectively. This can not only improve the accuracy of the data and the credibility of the multi-dimensional resource data, but also effectively retain the key data that users are concerned about and effectively compress the multi-dimensional resource data. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 The figure is a flowchart of the method of the present invention. DETAILED DESCRIPTION

[0031] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is 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 and are not used to limit the present invention.

[0032] Although the steps in the present invention are arranged with numbers, they are not used to limit the order of the steps. Unless the order of the steps is clearly stated or the execution of a certain step requires other steps as a basis, the relative order of the steps can be adjusted. It is understood that the term "and / or" used in this article involves and covers any and all possible combinations of one or more of the associated listed items.

[0033] The present invention proposes a multi-dimensional cost optimization method in the field of cloud computing, including waste identification and specification recommendation algorithms, such as Figure 1 As shown, the specific implementation steps are as follows:

[0034] Step 1: Collect multi-dimensional resource data of the cloud environment and pre-process the multi-dimensional resource data;

[0035] In the embodiment, the cloud computing environment contains different types of cloud computing services such as IaaS and PaaS. These cloud computing services need to call different cloud resources during execution. Each type of cloud resource comes from the corresponding resource end. These resource ends can allocate or recycle the corresponding type of cloud resources according to the execution status of the cloud computing service during the execution of the cloud computing service. In order to accurately and comprehensively grasp the usage status of different types of cloud resources in the cloud computing environment, all resource ends are independently monitored based on the location information of all resource ends in the cloud environment to obtain the cloud resource usage data corresponding to each resource end. The cloud resource usage data can be but not limited to the cloud resource usage status. Then, the resource end identity information is identified and integrated for the cloud resource usage data, so as to obtain multi-dimensional resource data, so that all types of cloud resources involved in the cloud computing environment can be differentiated and identified and summarized, so as to facilitate the subsequent targeted and differentiated cost optimization of the multi-dimensional resource data and the proposal of applicable and accurate specification recommendations.

[0036] Specifically, it collects and monitors multi-dimensional data in the cloud environment, including the load of application instances, resource usage, network status, etc.

[0037] The collected data is preprocessed to process the missing values ​​and abnormal values ​​of the original monitoring data and to clean up the invalid information, so as to calculate the key data concerned by the present invention, such as the average value, 95th percentile value, maximum and minimum values ​​of the monitoring data.

[0038] During the monitoring and formation process of multi-dimensional resource data, it cannot be guaranteed that all data within the multi-dimensional resource data are complete and accurate. In order to improve the credibility of the multi-dimensional resource data, the missing values, data anomaly values ​​and data invalid values ​​are identified and processed for each cloud resource usage data under the multi-dimensional resource data, and the data missing location points, data anomaly location points and data invalid location points within the cloud resource usage data are obtained. In this way, the multi-dimensional resource data can be preprocessed with data completion, data correction and data deletion based on the missing data, anomaly data and invalid data. This can not only improve the accuracy of the data, but also effectively retain the key data that users are concerned about and effectively compress the multi-dimensional resource data.

[0039] Step 2: Identify the resource types of the preprocessed multi-dimensional resource data, and classify the multi-dimensional resource data into load fluctuation resource data, storage resource data, and special resource data;

[0040] In the embodiment, the multi-dimensional resource types include the following:

[0041] IaaS layer: CPU, memory, cloud disk, network, floating IP

[0042] PaaS layer: cloud database, load balancing, object storage, file storage, middleware

[0043] The present invention divides the multi-dimensional resource types into three types and adopts different cost optimization methods: load fluctuation type (different utilization rates at different times), storage type (only increases but never decreases), and special type (a separate solution is required).

[0044] For the included multidimensional data types, they are as follows:

[0045] Load Fluctuation: This type of data load fluctuation resource refers to the situation where the demand for computing resources fluctuates due to changes in workload in different time periods. This type of cloud resource is characterized by periodic or non-periodic fluctuations in resource usage, and requires rapid performance adjustment based on changes in workload, and the ability to quickly expand or shrink to cope with load changes, such as: CPU, memory, network, floating IP, etc.

[0046] Storage type: This type of data is a resource that provides persistent data storage in a cloud computing environment, and the amount of data only increases and never decreases. They are usually used to store data that needs to be retained for a long time, such as cloud hard drives.

[0047] Special type: This type of data in cloud computing refers to resources designed for specific use cases or service models. These resources usually have unique characteristics and optimization points to meet the needs of different business scenarios. This type of cloud resources is designed for specific application scenarios or service models. For example: cloud database, load balancing, object storage, file storage, middleware, etc.

[0048] Step 3: Identify resource waste for load fluctuation resource data, storage resource data, and special resource data respectively, obtain corresponding resource waste status information, and determine the specification recommendation information for load fluctuation resources, storage resources, and special resources based on the resource waste status information. This covers the cost optimization needs of various cloud resources in the cloud computing environment, can take care of the needs of the entire cloud computing ecosystem, more accurately analyze the cost waste of multi-dimensional resources, and provide adaptive optimization strategies and specification recommendations for different situations, thereby improving the accuracy and flexibility of cost optimization.

[0049] In the embodiment, the method of performing waste identification and specification recommendation is as follows:

[0050] 1. For load fluctuation type:

[0051] Waste identification: Calculate key data such as the average and 95th percentile of monitoring data utilization, and set thresholds based on the actual situation of the resource.

[0052] Taking the CPU as an example, the specification can be set to an average CPU utilization of less than or equal to 10% and a CPU utilization P95 of less than 80%. The reason for using P95 is that the maximum value is an arbitrary and extreme number, and P95 can better express most situations. At the same time, a threshold is set. When the average utilization is higher than this value, the specification is upgraded. For example, if the average CPU utilization is higher than 90%, it is recommended to upgrade; when the average CPU utilization is lower, it is recommended to downgrade.

[0053] For memory, a similar waste identification strategy is used as for CPU.

[0054] Cloud resources that cannot meet actual usage are also defined as waste, which is a waste of efficiency.

[0055] Specification recommendation: Specification recommendation adopts waterfall model, multi-level downgrade. Taking the 95th percentile as an example, if the 95th percentile is greater than 25 and less than 50, it is recommended to downgrade the virtual machine level by 1 level; if the 95th percentile is greater than 12.5 and less than 25, it is recommended to downgrade the virtual machine level by 2 levels; and so on. For example, a virtual machine with 4 cores and 4GB of memory is downgraded by one level, it will be downgraded to 2 cores and 2GB of memory. If it is downgraded by two levels, it will remain at 1 core and 1GB of memory. According to the actual existing specifications, the number of CPU cores can be reduced to 1 at most, and the number of CPU cores and the size of memory GB meet the ratio of 1:1, 1:2, and 1:4.

[0056] 2. For storage type:

[0057] Waste identification: Set a threshold and make recommendations for capacity expansion when the storage capacity exceeds the threshold. Taking cloud hard disks as an example, when the capacity is higher than 90%, historical monitoring data is used to predict the capacity. If the predicted value is higher than 100% of the current capacity, it is identified as waste (efficiency waste).

[0058] Specification recommendation: Use capacity forecast values ​​to make specification recommendations.

[0059] 3. For special types:

[0060] Since the specifications of this type of data, the types of services provided, and billing methods vary greatly among different cloud service providers, different data types are handled differently.

[0061] (1) Cloud database:

[0062] Waste identification: Cloud database waste identification is a combination of load fluctuation type and storage type. The configuration specifications of the database also include CPU and memory. The storage of the cloud database is also a cloud hard disk, so the two can be combined. In the embodiment, the waste identification of the cloud database is related to the actual CPU utilization, actual memory utilization and actual space utilization. The previous idea of ​​handling load fluctuation can be followed to set a threshold such as 80%. When the threshold is exceeded, the configuration is upgraded, and when the utilization is low, the configuration is downgraded.

[0063] (2) Load balancing:

[0064] Different load balancing products have different specification configuration methods, which are roughly divided into the following four types: the number of new connections processed per second, the number of concurrent connections per minute, the amount of data processed for HTTP(S) requests and responses, and the product of the total number of rules processed and the number of requests per second. Select load balancing products with different specifications based on the usage of these four indicators.

[0065] In the embodiment, waste identification: the unit price of the load balancing instance fee is related to the instance function version, which is divided into basic version, standard version and enhanced version. The basic version supports the ALB function based on forwarding rules such as domain names and paths, and the standard version supports the ALB function based on forwarding rules such as HTTP request methods and source IP. The enhanced version adds web application security protection functions on the basis of the standard version. The instance fee will be charged according to the version. The unit prices of the three versions are 0.049 yuan / hour, 0.147 yuan / hour and 0.245 yuan / hour respectively. The version needs to be selected according to business needs.

[0066] The performance capacity unit LCU (Loadbalancer Capacity Unit) is the smallest unit used to measure the resources consumed by load balancing. The number of LCUs = max{maximum number of new connections (seconds) ÷ LCU coefficient (25), maximum number of concurrent connections (minutes) ÷ LCU coefficient (3000), total processed data volume (hours) ÷ LCU coefficient (1GB), number of rule evaluations (none) ÷ LCU coefficient (1000)}.

[0067] If the number of LCUs currently purchased is greater than 120% of the number of LCUs expected to be used, the number of connections purchased is reduced to 120% of the number expected to be used.

[0068] When waste is identified, the specification recommendation is: select the appropriate instance function version according to business needs. If a lower version can be selected, select the lower version. At the same time, if the number of LCUs exceeds 120% of the expected number of usage, reduce the number of LCUs to achieve the effect of the specification recommendation.

[0069] (3) Object storage: Provides a tiered storage solution based on the access frequency of storage objects. Storage objects with low access frequency are stored in low-frequency space to reduce costs. It can be divided into standard storage, low-frequency storage, archive storage and other types.

[0070] In the embodiment, waste identification: object storage is a special type of cloud resource. A tiered storage solution is provided according to the access frequency of the storage object. Storage objects with low access frequency are stored in a low-frequency space to reduce costs. It can be divided into standard storage, low-frequency storage, archive storage and other types.

[0071] Set a threshold for the number of cumulative requests and whether data retrieval is required over a long period of time. The threshold is related to pricing. For example, it can be set to 20,000 times (the time can be set to 1 year, etc.), and different storage objects can be classified in a gradient manner.

[0072] When waste is identified, the specification recommends: Classify storage into different storage solutions such as standard, infrequent access, and archive based on different thresholds, with storage costs from high to low and data retrieval and request costs from low to high.

[0073] (4) File storage: Provide users with a small file storage optimization solution to merge related small files into large files to reduce the number of storage blocks and metadata storage, thereby improving storage efficiency. This is a weak recommendation.

[0074] In the embodiment, waste identification: if the total storage capacity of small files (less than 1MB) accounts for more than 30%, it is determined to be waste;

[0075] When waste is identified, the specification recommends: Provide users with a small file storage optimization solution to merge related small files into large files to reduce the number of storage blocks and metadata storage; recommend a single file size of 10MB to 100MB to reduce metadata overhead;

[0076] (5) Middleware: Different middlewares have different billing methods. Bandwidth-related billing is handled in the same way as the network billing. Scheduling-related billing is handled in the same way as load balancing billing.

[0077] In the embodiment, waste identification: when the CPU usage of the instance is less than 20% and the memory usage is less than 30%, it is considered waste; when the load is less than 10% for 7 consecutive days (such as no active connections or requests), it is considered waste; when the maximum number of connections is limited to more than 150% of the actual demand, it is considered waste;

[0078] When waste is identified, the specification recommends:

[0079] Using the waterfall model and multi-level downgrading, if the 95th percentile value is greater than 25 and less than 50, it is recommended to downgrade the middleware level by 1 level; if the 95th percentile value is greater than 12.5 and less than 25, it is recommended to downgrade the middleware level by 2 levels; and so on.

[0080] For long periods of low load (for example, load less than 10% within 7 days), an automatic sleep option is provided, and resources are restored when activated to further save costs.

[0081] If the maximum connection limit is more than 150% of the actual demand, set it to 150%.

[0082] The glossary of terms in this article is explained as follows:

[0083] The IaaS (Infrastructure as a Service) layer is part of cloud computing and provides basic computing resources such as virtual machines, storage, networks, and operating systems.

[0084] The PaaS (Platform as a Service) layer is part of cloud computing and provides a complete development and deployment environment that supports users to build, test, deploy and manage applications without having to deal with the complexity of the underlying infrastructure. The PaaS layer mainly includes components such as databases, object storage, middleware and runtime environments.

[0085] P95 (95th percentile) is a metric in statistics that describes the characteristics of data distribution. In performance testing and monitoring, P95 is often used to measure the performance of response time or latency. Specifically, P95 means that 95% of the values ​​in the data set are less than or equal to this value, and only 5% of the values ​​are greater than it.

[0086] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

[0087] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

Claims

1. A multi-dimensional cloud resource waste identification and specification recommendation method, characterized in that: The specific steps include: Step 1: Collect multi-dimensional resource data of the cloud environment and pre-process the multi-dimensional resource data; Step 2: Identify the resource types of the preprocessed multi-dimensional resource data, and classify the multi-dimensional resource data into load fluctuation resource data, storage resource data, and special resource data; Step 3: Identify resource waste for load fluctuation resource data, storage resource data and special resource data respectively, obtain corresponding resource waste status information, and determine the specification recommendation information for load fluctuation resources, storage resources and special resources based on the resource waste status information.

2. The multi-dimensional cloud resource waste identification and specification recommendation method according to claim 1, characterized in that: The step 1 collects multi-dimensional resource data of the cloud environment, including: Based on the location information of all resource terminals within the cloud environment, all resource terminals are independently monitored to obtain the cloud resource usage data corresponding to each resource terminal; The cloud resource usage data is identified and integrated with the resource-side identity information to obtain multi-dimensional resource data.

3. The multi-dimensional cloud resource waste identification and specification recommendation method according to claim 1, characterized in that: The multi-dimensional resource data in step 1 includes the load of the application instance, resource usage and network status.

4. The multi-dimensional cloud resource waste identification and specification recommendation method according to claim 1, characterized in that: The pre-processing process described in step 1 is as follows: Identify missing values, outliers and invalid information, record the data missing location points, data outlier location points and data invalid location points, and use the data missing location points, data outlier location points and data invalid location points as the basis for data completion, data correction and data deletion.

5. The multi-dimensional cloud resource waste identification and specification recommendation method according to claim 1, characterized in that: The load fluctuation resource data in step 2 indicates the fluctuation of the demand for computing resources due to changes in workload in different time periods, including CPU, memory, network, and floating IP data; The storage-type resource data represents resources that provide persistent data storage in a cloud computing environment, including data on cloud hard disks; the special-type resource data represents resources designed for fixed use cases or service models in cloud computing, including cloud databases, load balancing, object storage, file storage, and middleware data.

6. The multi-dimensional cloud resource waste identification and specification recommendation method according to claim 1, characterized in that: In step 3, for load-fluctuating resource data, the average value and 95th percentile of the monitoring data utilization are calculated and a waste identification strategy is performed in combination with the corresponding pre-set thresholds of the resources; when waste is identified, a waterfall model is used to perform multi-level specification upgrades or downgrades.

7. The multi-dimensional cloud resource waste identification and specification recommendation method according to claim 1, characterized in that: In step 3, for storage resource data, when its storage capacity is higher than 90%, historical monitoring data is used to predict capacity. If the predicted capacity value is higher than 100% of the current capacity, it is identified as efficiency waste. When waste is identified, the predicted capacity value is used to recommend expansion specifications.

8. The multi-dimensional cloud resource waste identification and specification recommendation method according to claim 1, characterized in that: In step 3, special resource data is processed according to different data types, as follows: (1) Cloud database: The actual CPU utilization, actual memory utilization, and actual space utilization are comprehensively considered to identify waste. When waste is identified, CPU and memory specifications are recommended according to the waterfall model, and storage space is directly scaled up or down. (2) Load balancing: If the number of LCUs currently purchased is greater than 120% of the number of LCUs expected to be used, the waste is identified and the appropriate instance function version is selected based on business requirements; the number of LCUs expected to be used = max{maximum number of new connections per second / 25, maximum number of concurrent connections per minute / 3000, total amount of data processed per hour / 1GB, number of rule evaluations / 1000}; (3) Object storage: waste is identified by judging the relationship between the cumulative number of requests and the cumulative number of requests threshold of this gradient. According to different thresholds, storage is classified into standard, low-frequency access, or archive storage solutions. Storage costs are ranked from high to low, and data retrieval and request costs are ranked from low to high. (4) File storage: If the total storage volume of small files exceeds 30%, it is considered to be a waste and a small file storage optimization solution is provided to the user to merge related small files into large files to reduce the number of storage blocks and metadata storage; The recommended single file size is 10MB to 100MB to reduce metadata overhead; (5) Middleware: When the CPU usage of an instance is less than 20% and the memory usage is less than 30%, it is considered a waste and a waterfall model with multi-level downgrading is adopted. When the load is less than 10% for seven consecutive days, it is considered a waste and the instance is automatically dormant. The resources are restored when activated. When the maximum number of connections is more than 150% of the actual demand, it is considered a waste and the maximum number of connections is set to 150%.

Images