Method, medium and device for constructing a multi-dimensional energy consumption quantification analysis model of a data center

Abstract

This invention discloses a method, medium, and equipment for constructing a multidimensional energy consumption quantitative analysis model for data centers. First, it uses multinomial regression to fit the relationship between the utilization rate of physical CPUs, memory, and disks and their power consumption. The power consumption of physical hosts, virtual hosts, containers, and computing tasks is considered as the sum of the power consumption generated by the CPU, memory, and disk. CPU utilization, memory utilization, and disk I / O throughput are used to represent the power consumption of physical hosts, virtual hosts, containers, and computing tasks. Finally, the energy consumption over a specified time period is calculated by integrating the power consumption over time. Compared with existing technologies, this invention studies the hierarchical relationship of data centers, comprehensively reflecting the overall energy consumption of data centers, and has the advantage of reflecting data center energy consumption from multiple levels and dimensions.

Description

Technical Field

[0001] This invention belongs to the field of computer data processing technology and relates to data center energy consumption modeling, providing a method for constructing a multi-dimensional energy consumption quantitative analysis model for data centers. Background Technology

[0002] With the rapid growth in the number and scale of data centers, their energy demand is also increasing rapidly. In order to support the future development needs of data centers towards greater intensification and scale, and to form a new data center development model that is green, low-carbon, and whose computing power scale is compatible with the growth of the digital economy, it is urgent to study cloud data center energy consumption models to provide a basis for energy-saving optimization of cloud data centers.

[0003] Traditional data center energy consumption models often consider only a single level or dimension, modeling energy consumption for physical hosts or single applications. They fail to reflect the correlation between the energy consumption of physical hosts and that of virtual hosts, containers, and computing tasks. They cannot establish an energy consumption model from the perspective of the entire data center; they only model each component individually, while these components are actually dynamically interconnected and collaborative. Independent modeling makes comprehensive data center analysis difficult. Therefore, to address these problems, this invention proposes a multi-level, multi-dimensional energy consumption quantification analysis technique. It analyzes the relationships between physical hosts, virtual hosts, containers, and computing tasks at four levels and establishes an energy consumption quantification analysis model to solve the problem of data center energy consumption quantification analysis. Summary of the Invention

[0004] The problem this invention aims to solve is that existing technologies model the energy consumption of data centers in a single level or dimension, which cannot fully reflect the overall energy consumption of data centers. Due to the complexity of the relationships between physical hosts, virtual hosts, containers, and computing tasks in data centers, it is not possible to simply and directly correlate the energy consumption of physical hosts, virtual hosts, containers, and computing tasks, which is not conducive to the accurate analysis of the energy consumption of data centers.

[0005] The technical solution of this invention is as follows: a method for constructing a multi-dimensional energy consumption quantitative analysis model for data centers, which divides the energy consumption of data centers into four levels: physical hosts, virtual hosts, containers, and computing tasks, analyzes the relationships between each level, and establishes a hierarchical and multi-dimensional energy consumption quantitative analysis model for data centers, realizing a data center energy consumption model from coarse-grained to fine-grained, including the following steps:

[0006] Step 1: The data center monitoring module collects system monitoring information, including the resource utilization and real-time power consumption of physical hosts, as well as the physical resource allocation and load of virtual hosts, containers, and computing tasks.

[0007] Step 2: Build a data center physical host energy consumption quantitative analysis model based on a single physical host: obtain the power consumption data of the physical host, as well as the physical CPU utilization, physical memory utilization and disk I / O throughput information at this time, perform multiple sampling to obtain a monitoring dataset, use the dataset to build a power consumption model, and use the power consumption model to calculate and analyze the energy consumption of the physical host in real time.

[0008] Step 3: Based on the model data from Step 2, analyze the usage of physical resources by virtual hosts, analyze the proportion of various physical resources occupied by virtual hosts in the total usage, and construct a quantitative analysis model for the energy consumption of data center virtual hosts.

[0009] Step 4: The container runs on a physical host or a virtual host. Depending on the location of the container, based on the model data from Steps 2 and 3, analyze the proportion of the usage of various resources of the physical host used by the container in the total usage of various resources of the physical host, or the proportion of the usage of various resources of the virtual host used by the container in the total usage of various resources of the virtual host, and build a data center container energy consumption quantitative analysis model.

[0010] Step 5: The computing task runs on a physical host, virtual host, or container; Based on the model data from Steps 2, 3, and 4, if the computing task is located on a physical host or virtual host, analyze the usage of various resources of the physical host or virtual host used by the computing task; if the computing task runs in a container, the overall power consumption of the container is regarded as the power consumption of the computing task running in that container; construct a quantitative analysis model for the energy consumption of data center computing tasks.

[0011] Step 6: The constructed quantitative analysis model forms a multi-dimensional energy consumption quantitative analysis model for data centers, which is used to perform multi-level and multi-dimensional quantitative analysis of data center energy consumption.

[0012] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed, implements the above-described method for constructing a multidimensional energy consumption quantitative analysis model for a data center, thereby constructing the multidimensional energy consumption quantitative analysis model for a data center.

[0013] The present invention also provides an electronic device, including a memory and a processor. The memory stores a computer program, and the computer program implements the above-described method for constructing a multidimensional energy consumption quantitative analysis model for data centers when the processor is running, thereby obtaining a multidimensional energy consumption quantitative analysis model for data centers, which is used to quantitatively analyze the energy consumption of data centers.

[0014] The beneficial effects of this invention are as follows:

[0015] (1) The present invention provides a method for constructing a multi-dimensional energy consumption quantitative analysis model for data centers. By analyzing the relationship between physical hosts, virtual hosts, containers, and computing tasks, it creatively predicts the proportion of energy consumption generated by virtual hosts, containers, and computing tasks during their operation to the total energy consumption of their parent system according to the proportion of resources they use. Thus, it links the power consumption of virtual hosts, containers, and computing tasks with the power consumption of their parent system, solving the problem that existing energy consumption models can only establish energy consumption models for a specific dimension of physical hosts, virtual hosts, containers, and computing tasks, and there is no correlation between different dimension models. It establishes a multi-dimensional and multi-level energy consumption quantitative analysis model, which can predict both coarse-grained energy consumption such as physical hosts and fine-grained energy consumption such as virtual hosts, containers, and computing tasks, and can reflect the proportion of energy consumption of virtual hosts, containers, and computing tasks on physical nodes in real time.

[0016] (2) The present invention provides a method for constructing a multidimensional energy consumption quantitative analysis model for data centers. First, energy consumption data, CPU usage information, and memory are obtained through the data center monitoring module. Multivariate linear regression is used to construct the energy consumption relationship between CPU and its utilization rate, the energy consumption relationship between memory and its utilization rate, and the energy consumption relationship between disk I / O throughput. In this way, the energy consumption of physical hosts, virtual hosts, containers, and computing tasks are all represented by the energy consumption generated by CPU, memory, and disk I / O operations, which has good interpretability. Attached Figure Description

[0017] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0018] Existing technologies model energy consumption analysis of physical hosts or single applications in data centers, which is difficult to intuitively reflect the overall energy consumption of the data center. This invention studies and analyzes the relationship between physical hosts, virtual hosts, containers, and computing tasks in a data center, and proposes a method for constructing a multi-dimensional energy consumption quantitative analysis model for data centers. This model considers multiple layers of the data center and can reflect the energy consumption of physical hosts, virtual hosts, containers, and computing tasks, as well as their interrelationships and influences. It provides an overall reflection of the energy consumption of the data center and is beneficial for the quantitative analysis of the energy consumption of the entire data center.

[0019] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0020] like Figure 1As shown, a method for constructing a multi-dimensional energy consumption quantitative analysis model for data centers is proposed. This method divides data center energy consumption into four levels: physical hosts, virtual hosts, containers, and computing tasks. By analyzing the relationships between physical hosts, virtual hosts, containers, and computing tasks, a new modeling scheme is proposed to reasonably unify the energy consumption models of these four levels. This overcomes the difficulty in establishing connections between energy consumption models of different dimensions, and establishes a hierarchical, multi-dimensional energy consumption quantitative analysis model for data centers, achieving a data center energy consumption model from coarse-grained to fine-grained. The specific steps of this invention are as follows.

[0021] Step 1: The data center monitoring module collects system monitoring information, including the resource utilization and real-time power consumption of physical hosts. It also collects information on the physical resource allocation and load of virtual hosts, containers, and computing tasks. The data center monitoring module directly collects initial data such as physical host resource utilization, energy consumption, and load. This data is then preprocessed and converted into usable data before being used. The data center monitoring module collects information on a per-physical-host basis, including the following:

[0022] 1-1: Sample the CPU core utilization (U) of the current physical host using the API provided by the Linux operating system. cpui Memory usage U mem and disk I / O throughput U I / O , where i is the CPU core number. Assuming the current physical host has N CPU cores, the value of i ranges from 1 to N.

[0023] 1-2: Using the API provided by libvirt, the CPU and memory usage of virtual hosts, as well as disk I / O throughput, are obtained. The data center monitoring module preprocesses this raw data, calculating the utilization rate of CPU time, memory usage, and disk I / O throughput to obtain the utilization rate of each virtual host for each CPU core, memory usage, and disk I / O throughput. These data are denoted as U. cpui-vmj U mem-vmj U I / o-vmj , where i is the CPU core number and j is the virtual host number.

[0024] 1-3: The system hosting the Docker container may be a physical host or a virtual host. The Docker API is used to obtain the CPU, memory usage, and disk I / O throughput of the Docker container on the system. The data center monitoring module preprocesses this data to determine the Docker container's utilization per CPU core, memory utilization, and disk I / O throughput. These data are denoted as U... cpui-docker Umem-docker U I / o-docker , where i is the CPU core number.

[0025] 1-4: The system where each process resides may be a physical host or a virtual host. The memory usage and disk I / O throughput of each process on its host system can be obtained through the Linux API, denoted as U. mem-process U I / o-process It can also obtain the CPU cores where the process is located and their utilization, denoted as U. cpui-process , where i is the CPU core number.

[0026] 1-5: From the perspective of the physical host, regarding the utilization rate of CPU core numbered i, let...

[0027]

[0028] Regarding memory usage,

[0029]

[0030] For disk I / O throughput, let

[0031]

[0032] Where V represents the number of virtual hosts on the physical host, D represents the number of containers on the physical host, K represents the number of processes on the physical host, r represents the container number on the physical host, and s represents the process number on the physical host.

[0033] 1-6: From the perspective of the virtual host, for the utilization rate of CPU core numbered i, let:

[0034]

[0035] Regarding memory utilization, let:

[0036]

[0037] Regarding disk I / O utilization, let:

[0038]

[0039] D v K represents the number of containers on the current virtual host. v This indicates the number of processes currently running on the virtual host, where y is the container number on the virtual host and q is the process number on the virtual host.

[0040] 1-7: Obtain real-time power consumption data of physical hosts through data center environment monitoring systems or power meters.

[0041] Step 2: Construct a quantitative analysis model of energy consumption for data center physical hosts, using a single physical host as the unit.

[0042] Step 2-1: Obtain the power consumption data of the physical host, as well as the physical CPU utilization, physical memory utilization and disk I / O throughput information at this time through the data center monitoring module. Perform multiple samplings to obtain the monitoring dataset, and use the dataset to build a power consumption model.

[0043] Step 2-2: Consider the power consumption of the physical host as the sum of the power consumption of the physical CPU, the power consumption of the physical memory, and the power consumption generated by disk I / O operations. The model is as follows:

[0044] P = P cpu +P mem +P I / o #(7)

[0045] Where P represents the power consumption of the physical host, P cpu P represents the power consumption of the physical CPU. mem P represents the power consumption of physical memory. I / o This represents the power consumption generated by disk I / O, where:

[0046]

[0047]

[0048]

[0049] in Let represent the power consumption of the i-th CPU core on the physical host.

[0050]

[0051] get

[0052]

[0053] in U mem and U I / o Let i represent the utilization rate of the physical CPU core numbered i, the physical memory utilization rate, and the disk I / O throughput, respectively. N is the number of CPU cores, n is the maximum power of the polynomial (usually n=2 for best results), and h represents the exponent. The final result is:

[0054]

[0055] Steps 2-3: Using the power consumption data of the physical host sampled multiple times, along with the current physical CPU core utilization, physical memory utilization, and disk I / O throughput, as the dataset, multinomial regression is used to fit the coefficients (α). 1,0 ,α 1,1 ,α 1,2 ,…,α 1,n ,α 2,0 ,α 2,1 ,α 2,2 ,…,α 2,n ,...α N,0 ,α N,1 ,α N,2 ,…,α N,n (β0,β1,β2,…β) n (γ0,γ1,γ2,…,γ) n Substitute the values ​​of ) back into formula (13) to obtain the relationship between the power consumption of the physical host and the utilization rate of the physical CPU, the utilization rate of the physical memory and the disk I / O throughput.

[0056] Steps 2-4: The energy consumption of the physical host is calculated by integrating the power consumption over time. For the physical host, its energy consumption within the time period [0, T] is:

[0057] E=∫0 T P(t)dt#(14)

[0058] Where E is the energy consumption of the physical host, and P(t) is the real-time power consumption of the physical host at time t.

[0059] Step 3: Based on the model data from Step 2, analyze the virtual host's use of physical resources, analyze the proportion of various physical resources occupied by the virtual host in the total usage, and construct a quantitative analysis model for data center virtual host energy consumption.

[0060] Step 3-1: First, obtain the physical resource information allocated to the virtual host, including the allocated physical CPU cores and physical memory size. Then, as described in Steps 1-3, obtain the virtual host's utilization rate of each CPU core, physical memory utilization rate, and disk I / O throughput. Record these data as follows: Where j is the ID of the current virtual host.

[0061] Step 3-2: The power consumption of a virtual machine is generated by the operation of the physical CPU, physical memory, and disk I / O devices. The power consumption of a virtual machine is the sum of the power consumption of its physical CPU, physical memory, and disk I / O devices. The loads running on the physical machine include three types: virtual machines, containers, and computing tasks. The power consumption generated by the physical machine can be regarded as the sum of the power consumption generated by these three types of loads. Therefore, this sum of power consumption can be distributed to each component proportionally, that is, according to the ratio of the CPU, memory, and disk I / O throughput occupied by the virtual machine to the total CPU, memory, and disk I / O throughput occupied by all loads.

[0062]

[0063] in This represents the power consumption of the virtual machine with the ID j. This represents the power consumption generated by virtual host j on the physical CPU numbered i. This represents the power consumption of virtual host j in memory. This represents the power consumption generated by disk I / O of virtual host j; U mem-sum The acquisition is as described in step 1; P mem and P I / o It is obtained using steps 2-2 and 2-3.

[0064] Step 3-3: The energy consumption of a virtual host is calculated by integrating power consumption over time. For a virtual host, its energy consumption within the time period [0, T] is:

[0065]

[0066] in This refers to the energy consumption of the virtual host. Let t be the real-time power consumption of the virtual host.

[0067] Step 4: Containers can run on either physical hosts or virtual hosts. Based on the model data from Steps 2 and 3, analyze the proportion of the physical host's resource usage by the container to the total physical host resource usage, or the proportion of the virtual host's resource usage by the container to the total virtual host resource usage. Model these proportions separately, with the specific steps as follows:

[0068] Step 4-1: Treat the container as a special process. The data center monitoring module obtains the container's CPU usage, memory usage, and disk I / O throughput on the system it resides in through the container engine's built-in interface.

[0069] Step 4-2: For containers running on a physical host, the container's power consumption is the sum of its CPU, memory, and disk I / O power consumption. The power consumption generated by the CPU, memory, and disk I / O is calculated according to the corresponding proportions, i.e.:

[0070]

[0071] in This represents the power consumption of the container numbered r. These represent the container's CPU utilization, memory utilization, and disk I / O throughput, respectively. P mem and P I / o The results have already been obtained in steps 2-3.

[0072] Step 4-3: For containers running on virtual machines, the container's power consumption is the sum of its CPU, memory, and disk I / O power consumption. The power consumption generated by CPU, memory, and disk I / O is calculated according to the corresponding proportions, i.e.:

[0073]

[0074] Among them This represents the power consumption of the container numbered y. P vmem and P vI / o These represent the CPU power consumption, memory power consumption, and disk I / O power consumption of the virtual host where the container resides, respectively, which are obtained through step 3-2.

[0075] Step 4-4: The energy consumption of the container is calculated by integrating the power consumption over time. For the container, its energy consumption in the time period [0, T] is:

[0076] E docker =∫0 T P docker (t)dt#(19)

[0077] Where E docker It is the energy consumption of the container, P docker (t) represents the real-time power consumption of the container at time t. This calculation is based on the power consumption of the Docker container and does not consider the location of the container. or Either is acceptable, with P docker express.

[0078] Step 5: The computing task runs on a physical host, virtual host, or container; Based on the model data from Steps 2, 3, and 4, if the computing task is located on a physical host or virtual host, analyze the usage of various resources of the physical host or virtual host used by the computing task. If the computing task runs in a container, the overall power consumption of the container is regarded as the power consumption of the computing task running in the container, and a quantitative analysis model of data center computing task energy consumption is constructed.

[0079] Step 5-1: Classify computing tasks according to their running environment into three categories: running on physical hosts, running on virtual hosts, and running in containers. Calculate the power consumption of each process separately, treating the power consumption of a computing task as the sum of the power consumption of all its processes.

[0080] Step 5-2: When a process runs directly on the physical host, the power consumption of a single process is represented by the sum of its power consumption on each CPU and memory, as well as the power consumption generated by its disk I / O operations. The model is as follows:

[0081]

[0082] in This represents the power consumption of process s. Let represent the CPU utilization of process s on CPU number i, the memory utilization of process s, and the disk I / O throughput of process s, respectively, 1 <= s <= K, where K is the total number of processes on the physical host;

[0083] Step 5-2: When a process runs in a virtual host, the power consumption of a single process is represented by the sum of its power consumption on each CPU and memory, as well as the power consumption generated by its disk I / O operations. The model is as follows:

[0084]

[0085] in P vmem P vI / o These represent the CPU power consumption, memory power consumption, and disk I / O power consumption of the virtual host, respectively, which are obtained through step 3-2.

[0086] Step 5-3: When a process runs in a container, it typically either exclusively occupies the container or shares it with several other processes closely related to the same computing task. In this case, the power consumption of the container can be considered as the power consumption of the process or process group. Therefore:

[0087]

[0088] Step 5-4: Consider the power consumption of the computation task as the sum of the power consumption of its individual processes, that is:

[0089]

[0090] Where total represents the total number of processes that make up the computing task. P represents the power consumption of process k. job This represents the total power consumption of the computation task.

[0091] Step 5-5: The energy consumption of the computation task is calculated by integrating the power consumption over time. For the computation task, its energy consumption within the time period [0, T] is:

[0092] E job =∫0 T P job (t)dt#(24)

[0093] Where E job It is the energy consumption of the computing task, P job (t) represents the real-time power consumption of the computation task at time t.

[0094] The "#" in the above formula represents the separation between the formula itself and the formula label.

[0095] The energy consumption quantification analysis model for virtual hosts, containers, and computing tasks mentioned above is built upon the energy consumption quantification analysis model for physical hosts. This invention studies the dynamic correlation between different levels, linking them together. The data center multi-dimensional energy consumption quantification analysis model, composed of the aforementioned quantification analysis models, reflects the energy consumption of the data center from multiple levels and dimensions. It is used to quantify the energy consumption of the data center from multiple dimensions, analyzing both the overall energy consumption of the data center and the energy consumption of physical hosts, virtual hosts, containers, and computing tasks in detail, thus achieving coarse-grained to fine-grained energy consumption analysis of the data center.

[0096] The above-described method of this invention can be implemented based on a computer program. Therefore, this invention provides a computer-readable storage medium storing a computer program. When executed, the computer program implements the method for constructing a multi-dimensional energy consumption quantitative analysis model for data centers, thereby obtaining the multi-dimensional energy consumption quantitative analysis model for data centers. This invention also provides an electronic device, including a memory and a processor. The memory stores a computer program, which, when executed by the processor, implements the method for constructing a multi-dimensional energy consumption quantitative analysis model for data centers, thereby obtaining the multi-dimensional energy consumption quantitative analysis model for data centers, used for quantitative analysis of data center energy consumption.

Claims

1. A method for constructing a multidimensional energy consumption quantitative analysis model for data centers, characterized by: Data center energy consumption is divided into four levels: physical hosts, virtual hosts, containers, and computing tasks. The relationships between these levels are analyzed, and a hierarchical and multi-dimensional energy consumption quantitative analysis model for the data center is established. This results in a data center energy consumption model ranging from coarse-grained to fine-grained, including the following steps: Step 1: The data center monitoring module collects system monitoring information, including the resource utilization and real-time power consumption of physical hosts, as well as the physical resource allocation and load of virtual hosts, containers, and computing tasks. Step 2: Build a data center physical host energy consumption quantitative analysis model based on a single physical host: obtain the power consumption data of the physical host, as well as the physical CPU utilization, physical memory utilization and disk I / O throughput information at this time, perform multiple sampling to obtain a monitoring dataset, use the dataset to build a power consumption model, and use the power consumption model to calculate and analyze the energy consumption of the physical host in real time. Step 3: Based on the model data from Step 2, analyze the usage of physical resources by virtual hosts, analyze the proportion of various physical resources occupied by virtual hosts in the total usage, and construct a quantitative analysis model for the energy consumption of data center virtual hosts. Step 4: The container runs on a physical host or a virtual host. Depending on the location of the container, based on the model data from Steps 2 and 3, analyze the proportion of the usage of various resources of the physical host used by the container in the total usage of various resources of the physical host, or the proportion of the usage of various resources of the virtual host used by the container in the total usage of various resources of the virtual host, and build a data center container energy consumption quantitative analysis model. Step 5: The computing task runs on a physical host, virtual host, or container; Based on the model data from Steps 2, 3, and 4, if the computing task is located on a physical host or virtual host, analyze the usage of various resources of the physical host or virtual host used by the computing task; if the computing task runs in a container, the overall power consumption of the container is regarded as the power consumption of the computing task running in that container; construct a quantitative analysis model for the energy consumption of data center computing tasks. Step 6: The constructed quantitative analysis model forms a multi-dimensional energy consumption quantitative analysis model for data centers, which is used to perform multi-level and multi-dimensional quantitative analysis of data center energy consumption.

2. The method for constructing a multidimensional energy consumption quantification analysis model for data centers according to claim 1, characterized in that the data... The central monitoring module includes the following configuration: 1-1: Sample the CPU core utilization (U) of the current physical host using the API provided by the Linux operating system. cpui Memory usage U mem and disk I / O throughput U I / o , where i is the CPU core number. Suppose that the current physical host has N CPU cores, and the value of i ranges from 1 to N; 1-2: Using the API provided by libvirt, the raw information on virtual host CPU and memory usage, as well as disk I / O throughput, is obtained. The data center monitoring module preprocesses this data to calculate the utilization rate of each CPU core, memory utilization rate, and disk I / O throughput of each virtual host, denoted as follows: Where i is the CPU core number and j is the virtual host number; 1-3: The system hosting the Docker container includes physical hosts and virtual hosts. The Docker API is used to obtain the CPU, memory usage, and disk I / O throughput of the Docker container on the system. The data center monitoring module preprocesses this data to determine the Docker container's utilization per CPU core, memory utilization, and disk I / O throughput. These data are denoted as U... cpui-docker U mem-docker U I / o-docker , where i is the CPU core number; 1-4: The systems where each process resides include physical hosts and virtual hosts. The memory usage and disk I / O throughput of each process on its respective system are obtained using the APIs provided by Linux, denoted as U. mem-process U I / o-process And obtain the CPU cores where the process resides and their utilization rate, denoted as U. cpui-process , where i is the CPU core number; 1-5: From the perspective of the physical host, obtain the utilization rate of CPU core numbered i, let... Regarding memory usage, For disk I / O throughput, let Where V represents the number of virtual hosts on the physical host, D represents the number of containers on the physical host, K represents the number of processes on the physical host, r represents the container number on the physical host, and s represents the process number on the physical host. 1-6: From the perspective of the virtual host, obtain the utilization rate of CPU core numbered i, let: Regarding memory utilization, let: Regarding disk I / O utilization, let: D v K represents the number of containers on the current virtual host. v This indicates the number of processes currently running on the virtual host, where y is the container number on the virtual host and q is the process number on the virtual host. 1-7: Obtain real-time power consumption data of physical hosts through data center environment monitoring systems or power meters.

3. The method for constructing a multidimensional energy consumption quantification analysis model for data centers according to claim 1, characterized in that: The specific steps for establishing a quantitative analysis model for energy consumption of physical hosts are as follows: Step 2-1: Obtain the power consumption data of the physical host, as well as the physical CPU utilization, physical memory utilization and disk I / O throughput information at this time through the data center monitoring module. Perform multiple samplings to obtain the monitoring dataset, and use the dataset to build a power consumption model. Step 2-2: Consider the power consumption of the physical host as the sum of the power consumption of the physical CPU, the power consumption of the physical memory, and the power consumption generated by disk I / O operations. The model is as follows: P=P cpu +P mem +P I / o #(7) Where P represents the power consumption of the physical host, P cpu P represents the power consumption of the physical CPU. mem P represents the power consumption of physical memory. I / O This represents the power consumption generated by disk I / O, where: in Let represent the power consumption of the i-th CPU core on the physical host. get in U mem and U i / o Let i represent the utilization rate of the physical CPU core numbered i, the physical memory utilization rate, and the disk I / O throughput, respectively. Let N be the number of CPU cores, n be the maximum power of the polynomial, and h represent the power. The final result is: Steps 2-3: Using the power consumption data of the physical host sampled multiple times, along with the current physical CPU core utilization, physical memory utilization, and disk I / O throughput, as the dataset, multinomial regression is used to fit the coefficients (α). 1,0 ,α 1,1 ,α 1,2 ,…,α 1,n ,α 2,0 ,α 2,1 ,α 2,2 ,…,α 2,n ,...α N,0 ,α N,1 ,α N,2 ,…,α N,n (β0,β1,β2,…β) n (γ0,γ1,γ2,…,γ) n Substitute the values ​​of ) back into formula (13) to obtain the relationship between the power consumption of the physical host and the utilization rate of the physical CPU, the utilization rate of the physical memory and the disk I / O throughput. Steps 2-4: The energy consumption of the physical host is calculated by integrating the power consumption over time. For the physical host, its energy consumption within the time period [0, T] is: E=∫0 T P(t)dt#(14) Where E is the energy consumption of the physical host, and P(t) is the real-time power consumption of the physical host at time t.

4. The method for constructing a multidimensional energy consumption quantitative analysis model for data centers according to claim 1, characterized in that: The specific steps for establishing an energy consumption quantification analysis model for virtual hosts are as follows: Step 3-1: First, obtain the physical resource information allocated to the virtual host, including the allocated physical CPU cores and physical memory size. Then, obtain the virtual host's utilization rate of each CPU core, physical memory utilization rate, and disk I / O throughput. Record these data as follows: Where j is the ID of the current virtual host; Step 3-2: The power consumption of a virtual host is generated by the operation of the physical CPU, physical memory, and disk I / O devices. The power consumption of a virtual host is the sum of the power consumption of its physical CPU, physical memory, and disk I / O. The power consumption generated by the physical CPU, physical memory, and disk I / O is calculated according to the corresponding proportions, that is: in This represents the power consumption of the virtual machine with the ID j. This represents the power consumption generated by virtual host j on the physical CPU numbered i. This represents the power consumption of virtual host j in memory. This represents the power consumption generated by disk I / O of virtual host j; Step 3-3: The energy consumption of a virtual host is calculated by integrating power consumption over time. For a virtual host, its energy consumption within the time period [0, T] is: in This refers to the energy consumption of the virtual host. Let t be the real-time power consumption of the virtual host.

5. The method for constructing a multidimensional energy consumption quantification analysis model for data centers according to claim 1, characterized in that: To build a quantitative analysis model for container energy consumption, whether the container runs on a physical host or a virtual host, the specific steps are as follows: Step 4-1: Treat containers as processes. The data center monitoring module obtains the container's CPU usage, memory usage, and disk I / O throughput on its host system through the container engine's built-in interface. Step 4-2: For containers running on a physical host, the container's power consumption is the sum of its CPU, memory, and disk I / O power consumption. The power consumption generated by the CPU, memory, and disk I / O is calculated according to their respective proportions of the physical host's CPU, memory, and disk I / O consumption, i.e.: in This represents the power consumption of container numbered r, where 1 <= r <= D, and D is the number of containers on the physical host. These represent the container's CPU utilization, memory utilization, and disk I / O throughput, respectively. Step 4-3: For containers running on virtual machines, the power consumption of a container is the sum of its CPU, memory, and disk I / O power consumption. The power consumption generated by CPU, memory, and disk I / O is calculated according to their respective proportions of the virtual machine's CPU, memory, and disk I / O usage, i.e.: in This represents the power consumption of container numbered y, where 1 <= y <= Dv, and Dv is the number of containers on the virtual host. P vmem and P vI / o These represent the CPU power consumption, memory power consumption, and disk I / O power consumption of the virtual host where the container resides, respectively. Step 4-4: The energy consumption of the container is calculated by integrating the power consumption over time. For the container, its energy consumption in the time period [0, T] is: E docker =∫0 T P docker (t)dt#(19) Where E docker It is the energy consumption of the container, P docker (t) represents the real-time power consumption of the container at time t.

6. The method for constructing a multidimensional energy consumption quantification analysis model for data centers according to claim 1, characterized in that: The specific steps for establishing an energy consumption quantification analysis model for computational tasks are as follows: Step 5-1: Classify computing tasks according to their running environment into three categories: running on physical hosts, running on virtual hosts, and running in containers. Calculate the power consumption of each process separately, treating the power consumption of a computing task as the sum of the power consumption of all its processes. Step 5-2: When a process runs directly on the physical host, the power consumption of a single process is represented by the sum of its power consumption on each CPU and memory, as well as the power consumption generated by its disk I / O operations. The model is as follows: in This represents the power consumption of process s. These represent the CPU utilization rate of process s on CPU number i, the memory utilization rate of process s, and the disk I / O throughput of process s, respectively. Step 5-2: When a process runs in a virtual host, the power consumption of a single process is represented by the sum of its power consumption on each CPU and memory, as well as the power consumption generated by its disk I / O operations. The model is as follows: in P vmem P vi / o These represent the CPU power consumption, memory power consumption, and disk I / O power consumption of the virtual host, respectively. Step 5-3: When a process runs in a container, the power consumption of the container is considered as the power consumption of the process or process group. At this point: Step 5-4: Consider the power consumption of the computational task as the sum of the power consumption of its individual processes, that is: Where total represents the total number of processes that make up the computing task. P represents the power consumption of process k. job This represents the total power consumption of the computational task. Step 5-5: The energy consumption of the computation task is calculated by integrating the power consumption over time. For the computation task, its energy consumption within the time period [0, T] is: E job =∫0 T P job (t)dt#(24) Where E job It is the energy consumption of the computing task, P job (t) represents the real-time power consumption of the computation task at time t.

7. A computer-readable storage medium, characterized in that... The system contains a computer program that, when executed, implements the method for constructing a multidimensional energy consumption quantitative analysis model for a data center as described in any one of claims 1-6, thereby constructing the multidimensional energy consumption quantitative analysis model for a data center.

8. An electronic device, characterized in that: The system includes a memory and a processor. The memory stores a computer program, which, when executed by the processor, implements the method for constructing a multidimensional energy consumption quantitative analysis model for data centers as described in any one of claims 1-6, thereby obtaining the multidimensional energy consumption quantitative analysis model for data centers, which is used to quantitatively analyze the energy consumption of data centers.

Images