Energy-saving server operation method and system based on dynamic load adjustment

Abstract

The present invention discloses an energy-saving server operation method and system based on dynamic load regulation, which relates to the field of server technology. The energy-saving server operation method includes: generating a plurality of virtual heterogeneous resource units through a pooling method and a virtualization method for a plurality of heterogeneous resources in a single physical machine, and each heterogeneous resource in each virtual heterogeneous resource unit has a different ratio. The present invention establishes a No. 1 framework, which supports collaborative management and scheduling between multiple heterogeneous resources. The No. 1 framework dynamically adjusts the scale of the virtual heterogeneous resource unit for a single physical machine through load calculation, comparison and trigger conditions. Based on the reduced scale, the various heterogeneous resources in the single physical machine are subjected to different voltage and frequency reductions at the same time. While reducing the energy consumption of the physical machine, the voltage and frequency of each heterogeneous resource are reduced separately, thereby ensuring the effective collaborative scheduling of heterogeneous resources.

Description

Technical Field

[0001] The present invention relates to the technical field of servers, and in particular to an energy-saving server operation method and system based on dynamic load adjustment. Background Art

[0002] Energy consumption on servers accounts for the vast majority of total expenditure, and energy consumption is considered a key issue on servers.

[0003] For example, the patent publication number is "CN103970256A", and its name is "An energy-saving method and system based on memory compression and CPU dynamic frequency modulation". The above method includes compressing pages with the same content in the memory space, putting the saved space into a low-energy consumption state, and reducing the CPU frequency to the minimum frequency required to run the memory space. The above invention can well control the energy consumption of the server, effectively reduce the energy consumption caused by the CPU and memory, and improve memory utilization efficiency while reducing server energy consumption.

[0004] When achieving energy saving, the above method combines CPU dynamic frequency modulation technology with memory compression technology to reduce energy consumption during the memory compression process. However, on the server, there are not only CPU and memory resources, but also heterogeneous resources such as graphics processing units (GPUs), field programmable gate arrays (FPGAs), and disk networks. The above method fails to effectively coordinate the scheduling of heterogeneous resources such as CPUs and GPUs, resulting in some redundant heterogeneous resources being overloaded or idle, thereby causing energy consumption to continue on other heterogeneous resources. For this reason, an energy-saving server operation method and system based on dynamic load adjustment are invented. Summary of the Invention

[0005] The object of the present invention is to provide an energy-saving server operation method and system based on dynamic load adjustment to solve the problems raised in the above background technology.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an energy-saving server operation method based on dynamic load adjustment, the energy-saving server operation method comprising:

[0007] Multiple heterogeneous resources in a single physical machine are pooled and virtualized to generate multiple virtual heterogeneous resource units, with each virtual heterogeneous resource unit having a different ratio of heterogeneous resources.

[0008] Establishing a second load threshold of the physical machine and a first load threshold of the virtual heterogeneous resource unit based on the device parameters of the physical machine;

[0009] Build a framework for optimizing the energy consumption of a single physical machine and a framework for optimizing the energy consumption of a server cluster;

[0010] The first framework includes:

[0011] Obtain the load rates of several virtual heterogeneous resource units and the real-time global load rate through load calculation , establish the number one trigger condition for optimizing the energy consumption of physical machines;

[0012] The comparison method is used to compare the load rates of several virtual heterogeneous resource units with the load threshold of the virtual heterogeneous resource units, and the real-time global load rate is calculated. Compare it with the load threshold No. 2 to obtain the comparison result No. 1;

[0013] Trigger condition No. 1 determines whether to enable energy consumption optimization of the physical machine based on comparison result No. 1. When comparison result No. 1 meets the trigger condition in trigger condition No. 1, energy consumption optimization of the single physical machine is triggered. The scale of several virtual heterogeneous resource units in the physical machine is adjusted based on elastic resource scaling technology. The total heterogeneous scale of the combination of the several virtual heterogeneous resource units after adjustment is obtained. The total heterogeneous scale is compared with the total scale of the physical machine to obtain a comparison value set. Dynamic load adjustment is performed on several heterogeneous resources in the single physical machine through the comparison value set and the adjustment method.

[0014] The comparison value set includes the comparison values between the actual scale of several heterogeneous resources in the physical machine and the virtual scale of several heterogeneous resources in the total heterogeneous scale. , i represents the i-th heterogeneous resource;

[0015] The adjustment method includes: dynamically adjusting the load of a plurality of heterogeneous resource source devices, and adjusting the energy consumption comparison value between the unit energy consumption of the plurality of heterogeneous resources after modification and the unit energy consumption of the heterogeneous resources before modification. , energy consumption comparison value and contrast value Positive correlation;

[0016] The second framework includes:

[0017] Obtain the real-time global load rate of the physical machine through load calculation ;

[0018] Real-time global load factor achieved through comparison method Compare the load threshold with the second load threshold to obtain the second comparison result;

[0019] The second trigger condition determines whether to enable the energy consumption optimization of the server cluster based on the second comparison result. When the second comparison result meets the trigger condition in the second trigger condition, the energy consumption optimization of the server cluster is triggered;

[0020] Get real-time global load rate The difference between the load threshold and the second load threshold is used to establish a safety value, and the real-time global load rate is used to Divide the tasks in the physical machine into proportions to obtain a task scale ratio, obtain the scale of tasks to be processed based on the difference, the safety value, and the task scale ratio, and obtain the tasks to be processed from the virtual heterogeneous resource unit with the highest load rate;

[0021] The tasks to be processed in the physical machine are transferred between physical machines through feature selection method;

[0022] The feature selection method includes selecting from the server cluster based on the scale of the task to be processed that can receive the task to be processed and will not cause the real-time global load rate after receiving the task. For physical machines that exceed the second load threshold, pending tasks are sent to the selected physical machine;

[0023] The pending tasks are processed, and the task processing includes:

[0024] Perform feature processing on the pending tasks to obtain the coordination requirements of the pending tasks. The coordination requirements represent the task's demand for different heterogeneous computing resources. A feature vector of the task is established based on the coordination requirements. A feature vector of the virtual heterogeneous resource unit is established based on the ratio of each heterogeneous resource in the virtual heterogeneous resource unit.

[0025] The tasks to be processed are input to corresponding virtual heterogeneous resource units through a similar selection method. Dynamic load regulation includes adjusting the voltage and frequency of the heterogeneous resources.

[0026] Furthermore, the similarity selection method includes a direct pairing method and a combined pairing method;

[0027] The direct pairing method includes:

[0028] Calculate the similarity between the feature vector of the task to be processed and the matching feature vectors of several virtual heterogeneous resource units in the physical machine to obtain a similarity degree value, select the feature of the virtual heterogeneous resource unit with the highest similarity degree value as the input of the task to be processed, establish a degree threshold, and input the tasks whose similarity degree values are all lower than the degree threshold through the combined pairing method;

[0029] The combination pairing method includes:

[0030] Intelligently pair and combine pending tasks whose similarity values are all below a threshold to obtain the combined task and its feature vector. Calculate the similarity between the feature vector of the combined task and the feature vector of several virtual heterogeneous resource units in the physical machine to obtain a similarity value set. Select the feature of the virtual heterogeneous resource unit with the highest similarity value as the input of the combined task.

[0031] A time threshold is established. If the tasks in the combination matching method cannot be combined when the time threshold is exceeded, the degree threshold is ignored and the virtual heterogeneous resource unit feature with the highest similarity value is directly selected as the input of the combination task.

[0032] Furthermore, through load calculation including normalization processing and load calculation;

[0033] The normalization process includes establishing a normalization formula: ,The minimum load rate and the maximum load rate are the theoretical minimum load rate and the theoretical maximum load rate of each heterogeneous resource, respectively;

[0034] The load of each heterogeneous resource is calculated using a normalization formula to obtain a normalized load rate of each heterogeneous resource.

[0035] Furthermore, elastic scaling technology includes dynamically scaling the scale of each heterogeneous resource in the virtual heterogeneous resource unit, while maintaining the heterogeneous resource ratio unchanged during the scaling process;

[0036] The comparison value set includes comparison values between the actual scale of each heterogeneous resource in the physical machine and the virtual scale of each heterogeneous resource in the total heterogeneous scale;

[0037] Energy consumption comparison value and contrast value Positive correlation, establish contrast coefficient , .

[0038] Furthermore, the load calculation includes:

[0039] Real-time global load rate ;

[0040] ,N represents the number of groups in the heterogeneous resource grouping set, represents the overall proportion of the i-th heterogeneous resource in the physical machine, represents the normalized load rate of the i-th heterogeneous computing resource, Indicates the real-time global load rate based on heterogeneous resource load calculation;

[0041] The overall proportion is replaced by the energy consumption proportion. The energy consumption proportion represents the energy consumption weight of several heterogeneous resources in the physical machine, and the sum of the energy consumption proportions must be 1.

[0042] Furthermore, the second trigger condition includes:

[0043] Real-time global load rate The load value is continuously greater than or equal to the maximum value of the second load threshold for at least three minutes.

[0044] Furthermore,

[0045] The load calculation includes:

[0046] Establishing a formula for the load rate of virtual heterogeneous resource units ;

[0047] , represents the load rate of the jth virtual heterogeneous resource unit, the normalized processor load , normalized memory load , normalized disk load , normalized network bandwidth utilization ;

[0048] 、 、 and Indicates the energy consumption proportion of virtual heterogeneous resource units, and the sum of the energy consumption proportions must be 1;

[0049] , Y represents the number of virtual heterogeneous resource units in the physical machine, It represents the ratio of the size of the Y-th virtual heterogeneous resource unit to the total size of the physical machine. Indicates the real-time global load rate calculated based on the load of virtual heterogeneous resource units.

[0050] Furthermore, the first trigger condition includes:

[0051] Condition 1: The virtual heterogeneous resource unit (VHRU) remains below the VHRU threshold for at least three minutes.

[0052] Condition 2: Real-time global load rate The load is continuously lower than the minimum value of the second load threshold for at least three minutes;

[0053] Conditions 1 and 2 must be met at the same time.

[0054] An energy-saving server operation system based on dynamic load adjustment adopts the above-mentioned energy-saving server operation method based on dynamic load adjustment. Furthermore, the energy-saving server operation system includes:

[0055] Resource allocation module: Multiple heterogeneous resources in a single physical machine are pooled and virtualized to generate multiple virtual heterogeneous resource units. The ratio of heterogeneous resources in each virtual heterogeneous resource unit is different.

[0056] Threshold establishment module: establishes the second load threshold of the physical machine and the first load threshold of the virtual heterogeneous resource unit based on the device parameters of the physical machine;

[0057] Energy consumption optimization module: The energy consumption optimization module includes the first optimization module and the second optimization module;

[0058] Optimization module No. 1: Obtains the load rates of several virtual heterogeneous resource units and the real-time global load rate through load calculation , establish the number one trigger condition for optimizing the energy consumption of physical machines;

[0059] The comparison method is used to compare the load rates of several virtual heterogeneous resource units with the load threshold of the virtual heterogeneous resource units, and the real-time global load rate is calculated. Compare it with the load threshold No. 2 to obtain the comparison result No. 1;

[0060] Trigger condition No. 1 determines whether to enable energy consumption optimization of the physical machine based on comparison result No. 1. When comparison result No. 1 meets the trigger condition in trigger condition No. 1, energy consumption optimization of the single physical machine is triggered. The scale of several virtual heterogeneous resource units in the physical machine is reduced based on elastic resource scaling technology. The total heterogeneous scale of the combination of the several virtual heterogeneous resource units after reduction is obtained. The total heterogeneous scale is compared with the total scale of the physical machine to obtain a comparison value set. Dynamic load adjustment is performed on several heterogeneous resource source devices in the single physical machine through the comparison value set and the adjustment method.

[0061] Optimization module 2: Obtain the real-time global load rate of the physical machine through load calculation ;

[0062] Real-time global load factor achieved through comparison method Compare the load threshold with the second load threshold to obtain the second comparison result;

[0063] The second trigger condition determines whether to enable server cluster energy consumption optimization based on the second comparison result. When the second comparison result meets the trigger condition in the second trigger condition, the server cluster energy consumption optimization is triggered:

[0064] Get real-time global load rate The difference between the load threshold and the second load threshold is used to establish a safety value, and the real-time global load rate is used to Divide the tasks in the physical machine into proportions to obtain a task scale ratio, obtain the scale of tasks to be processed based on the difference, the safety value, and the task scale ratio, and obtain the tasks to be processed from the virtual heterogeneous resource unit with the highest load rate;

[0065] The pending tasks in the physical machine are transferred between physical machines through the feature selection method.

[0066] Compared with the prior art, the present invention has the following beneficial effects:

[0067] The energy-saving server operation method and system based on dynamic load adjustment, through the establishment of the No. 1 framework, the No. 1 framework supports the coordinated management and scheduling of multiple heterogeneous resources. The No. 1 framework dynamically adjusts the scale of the virtual heterogeneous resource unit for a single physical machine through load calculation, comparison and trigger conditions. Based on the reduced scale, the No. 1 framework simultaneously performs different voltage and frequency reductions on each heterogeneous resource in a single physical machine, reducing the energy consumption of the physical machine while reducing the voltage and frequency of each heterogeneous resource, thereby ensuring the effective coordinated scheduling of heterogeneous resources. At the same time, the No. 2 framework targets the server cluster. Based on the comparison between the real-time global load rate and the load threshold, when the specific No. 2 trigger condition is met, the No. 1 framework maintains the physical machine in a normal state through task migration and resource reallocation, avoids being in a high energy consumption state, and optimizes the energy consumption of the entire cluster.

[0068] Through feature selection and similarity selection methods, tasks are reasonably allocated to the most suitable virtual heterogeneous resource units, reducing resource fragmentation, avoiding mismatches between tasks and resources, and improving task processing efficiency. This method reduces the voltage and frequency of the physical machine when subsequently optimizing energy consumption for a single physical machine. At the same time, the reduction of resource fragments reduces idle resources, ensuring that the physical machine after voltage and frequency reduction is at lower energy consumption and can also perform tasks in the physical machine. At the same time, the heterogeneous resource ratio remains unchanged during the expansion and contraction process to ensure the stability and performance of the system. The scale of the virtual heterogeneous resource unit is dynamically adjusted according to the load conditions, which can not only meet the task requirements under high load, but also avoid resource waste under low load. The optimization of the No. 1 framework can be guaranteed by feature selection and similarity selection methods. Task allocation based on feature selection and similarity selection methods results in less resource fragmentation in heterogeneous resources after voltage and frequency reduction.

[0069] By establishing a load rate formula for virtual heterogeneous resource units and a real-time global load rate calculation method, unified management and scheduling of different heterogeneous resources are achieved. The system can intelligently allocate tasks to the most appropriate resource unit based on the feature vector of the task and the matching feature vector of the virtual heterogeneous resource unit, using similarity calculation. This intelligent scheduling method improves the accuracy of task allocation and optimizes the task processing flow of the entire server system. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 A schematic diagram of the operating method of the present invention;

[0071] Figure 2 Schematic diagram of the feature selection method of the present invention;

[0072] Figure 3 This is a schematic diagram of the first frame of the present invention;

[0073] Figure 4 Schematic diagram of the operating system of the present invention. DETAILED DESCRIPTION

[0074] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0075] like Figures 1-4 As shown, the present invention provides a technical solution: an energy-saving server operation method based on dynamic load adjustment, the energy-saving server operation method comprising:

[0076] Multiple heterogeneous resources in a single physical machine are pooled and virtualized to generate multiple virtual heterogeneous resource units, with each virtual heterogeneous resource unit having a different ratio of heterogeneous resources.

[0077] Establishing a second load threshold of the physical machine and a first load threshold of the virtual heterogeneous resource unit based on the device parameters of the physical machine;

[0078] Build a framework for optimizing the energy consumption of a single physical machine and a framework for optimizing the energy consumption of a server cluster;

[0079] Framework No. 1 includes:

[0080] Obtain the load rates of several virtual heterogeneous resource units and the real-time global load rate through load calculation , establish the number one trigger condition for optimizing the energy consumption of physical machines;

[0081] The comparison method is used to compare the load rates of several virtual heterogeneous resource units with the load threshold of the virtual heterogeneous resource units, and the real-time global load rate is calculated. Compare it with the load threshold No. 2 to obtain the comparison result No. 1;

[0082] Trigger condition No. 1 determines whether to enable energy consumption optimization of the physical machine based on comparison result No. 1. When comparison result No. 1 meets the trigger condition in trigger condition No. 1, energy consumption optimization of the single physical machine is triggered. The scale of several virtual heterogeneous resource units in the physical machine is adjusted based on elastic resource scaling technology. The total heterogeneous scale of the combination of the several virtual heterogeneous resource units after adjustment is obtained. The total heterogeneous scale is compared with the total scale of the physical machine to obtain a comparison value set. Dynamic load adjustment is performed on several heterogeneous resources in the single physical machine through the comparison value set and the adjustment method.

[0083] The comparison value set includes the comparison values between the actual scale of several heterogeneous resources in the physical machine and the virtual scale of several heterogeneous resources in the total heterogeneous scale. , i represents the i-th heterogeneous resource;

[0084] The adjustment method includes: dynamically adjusting the load of several heterogeneous resource source devices, and comparing the energy consumption of the unit energy consumption of the several heterogeneous resources after modification with the energy consumption of the unit energy consumption of the heterogeneous resources before modification. , energy consumption comparison value and contrast value Positive correlation;

[0085] Framework II includes:

[0086] Obtain the real-time global load rate of the physical machine through load calculation ;

[0087] Real-time global load factor achieved through comparison method Compare the load threshold with the second load threshold to obtain the second comparison result;

[0088] The second trigger condition determines whether to enable the energy consumption optimization of the server cluster based on the second comparison result. When the second comparison result meets the trigger condition in the second trigger condition, the energy consumption optimization of the server cluster is triggered;

[0089] Get real-time global load rate The difference between the load threshold and the second load threshold is used to establish a safety value, and the real-time global load rate is used to Divide the tasks in the physical machine into proportions to obtain a task scale ratio, obtain the scale of tasks to be processed based on the difference, the safety value, and the task scale ratio, and obtain the tasks to be processed from the virtual heterogeneous resource unit with the highest load rate;

[0090] The tasks to be processed in the physical machine are transferred between physical machines through feature selection method;

[0091] The feature selection method includes selecting servers from the server cluster based on the size of the tasks to be processed that can receive the tasks to be processed and will not cause the real-time global load rate after receiving them. For physical machines that exceed the second load threshold, pending tasks are sent to the selected physical machine;

[0092] The pending tasks are processed, which includes:

[0093] Perform feature processing on the pending tasks to obtain the coordination requirements of the pending tasks. The coordination requirements represent the task's demand for different heterogeneous computing resources. A feature vector of the task is established based on the coordination requirements. A feature vector of the virtual heterogeneous resource unit is established based on the ratio of each heterogeneous resource in the virtual heterogeneous resource unit.

[0094] The tasks to be processed are input to corresponding virtual heterogeneous resource units through a similar selection method. Dynamic load regulation includes adjusting the voltage and frequency of the heterogeneous resources.

[0095] Similarity selection methods include direct matching method and combined matching method;

[0096] Direct pairing methods include:

[0097] Calculate the similarity between the feature vector of the task to be processed and the matching feature vectors of several virtual heterogeneous resource units in the physical machine to obtain a similarity degree value, select the feature of the virtual heterogeneous resource unit with the highest similarity degree value as the input of the task to be processed, establish a degree threshold, and input the tasks whose similarity degree values are all lower than the degree threshold through the combined pairing method;

[0098] Combined pairing methods include:

[0099] Intelligently pair and combine pending tasks whose similarity values are all below a threshold to obtain the combined task and its feature vector. Calculate the similarity between the feature vector of the combined task and the feature vector of several virtual heterogeneous resource units in the physical machine to obtain a similarity value set. Select the feature of the virtual heterogeneous resource unit with the highest similarity value as the input of the combined task.

[0100] A time threshold is established. If the tasks in the combination matching method cannot be combined when the time threshold is exceeded, the degree threshold is ignored and the virtual heterogeneous resource unit feature with the highest similarity value is directly selected as the input of the combination task.

[0101] The load calculation includes normalization processing and load calculation;

[0102] Normalization processing includes establishing a normalization formula: ,The minimum load rate and the maximum load rate are the theoretical minimum load rate and the theoretical maximum load rate of each heterogeneous resource, respectively;

[0103] The load of each heterogeneous resource is calculated using a normalization formula to obtain a normalized load rate of each heterogeneous resource.

[0104] Elastic scaling technology involves dynamically scaling the scale of each heterogeneous resource in a virtual heterogeneous resource unit, while maintaining the same heterogeneous resource ratio during the scaling process.

[0105] The comparison value set includes comparison values between the actual scale of each heterogeneous resource in the physical machine and the virtual scale of each heterogeneous resource in the total heterogeneous scale;

[0106] Energy consumption comparison value and contrast value Positive correlation, establish contrast coefficient , .

[0107] Load calculations include:

[0108] Real-time global load rate ;

[0109] ,N represents the number of groups in the heterogeneous resource grouping set, represents the overall proportion of the i-th heterogeneous resource in the physical machine, represents the normalized load rate of the i-th heterogeneous computing resource, Indicates the real-time global load rate based on heterogeneous resource load calculation;

[0110] The overall proportion is replaced by the energy consumption proportion. The energy consumption proportion represents the energy consumption weight of several heterogeneous resources in the physical machine, and the sum of the energy consumption proportions must be 1.

[0111] Trigger conditions No. 2 include:

[0112] Real-time global load rate The load value is continuously greater than or equal to the maximum value of the second load threshold for at least three minutes.

[0113] Load calculations include:

[0114] Establishing a formula for the load rate of virtual heterogeneous resource units ;

[0115] , represents the load rate of the jth virtual heterogeneous resource unit, the normalized processor load , normalized memory load , normalized disk load , normalized network bandwidth utilization ;

[0116] 、 、 and Indicates the energy consumption proportion of virtual heterogeneous resource units, and the sum of the energy consumption proportions must be 1;

[0117] , Y represents the number of virtual heterogeneous resource units in the physical machine, It represents the ratio of the size of the Y-th virtual heterogeneous resource unit to the total size of the physical machine. Indicates the real-time global load rate calculated based on the load of virtual heterogeneous resource units.

[0118] Trigger conditions No. 1 include:

[0119] Condition 1: The virtual heterogeneous resource unit (VHRU) remains below the VHRU threshold for at least three minutes.

[0120] Condition 2: Real-time global load rate The load is continuously lower than the minimum value of the second load threshold for at least three minutes;

[0121] Conditions 1 and 2 must be met at the same time.

[0122] An energy-saving server operation system based on dynamic load adjustment adopts the above-mentioned energy-saving server operation method based on dynamic load adjustment. The energy-saving server operation system includes:

[0123] Resource allocation module: Multiple heterogeneous resources in a single physical machine are pooled and virtualized to generate multiple virtual heterogeneous resource units. The ratio of heterogeneous resources in each virtual heterogeneous resource unit is different.

[0124] Threshold establishment module: establishes the second load threshold of the physical machine and the first load threshold of the virtual heterogeneous resource unit based on the device parameters of the physical machine;

[0125] Energy consumption optimization module: The energy consumption optimization module includes the first optimization module and the second optimization module;

[0126] Optimization module No. 1: Obtains the load rates of several virtual heterogeneous resource units and the real-time global load rate through load calculation , establish the number one trigger condition for optimizing the energy consumption of physical machines;

[0127] The comparison method is used to compare the load rates of several virtual heterogeneous resource units with the load threshold of the virtual heterogeneous resource units, and the real-time global load rate is calculated. Compare it with the load threshold No. 2 to obtain the comparison result No. 1;

[0128] Trigger condition No. 1 determines whether to enable energy consumption optimization of the physical machine based on comparison result No. 1. When comparison result No. 1 meets the trigger condition in trigger condition No. 1, energy consumption optimization of the single physical machine is triggered. The scale of several virtual heterogeneous resource units in the physical machine is reduced based on elastic resource scaling technology. The total heterogeneous scale of the combination of the several virtual heterogeneous resource units after reduction is obtained. The total heterogeneous scale is compared with the total scale of the physical machine to obtain a comparison value set. Dynamic load adjustment is performed on several heterogeneous resource source devices in the single physical machine through the comparison value set and the adjustment method.

[0129] Optimization module 2: Obtain the real-time global load rate of the physical machine through load calculation ;

[0130] Real-time global load factor achieved through comparison method Compare the load threshold with the second load threshold to obtain the second comparison result;

[0131] The second trigger condition determines whether to enable server cluster energy consumption optimization based on the second comparison result. When the second comparison result meets the trigger condition in the second trigger condition, the server cluster energy consumption optimization is triggered:

[0132] Get real-time global load rate The difference between the load threshold and the second load threshold is used to establish a safety value, and the real-time global load rate is used to Divide the tasks in the physical machine into proportions to obtain a task scale ratio, obtain the scale of tasks to be processed based on the difference, the safety value, and the task scale ratio, and obtain the tasks to be processed from the virtual heterogeneous resource unit with the highest load rate;

[0133] The pending tasks in the physical machine are transferred between physical machines through the feature selection method.

[0134] The server runs on a physical machine, which provides the server with necessary hardware resources, such as the computing power of each processor, memory storage space, hard disk data storage, etc., so that the server can perform various tasks and services. The energy-saving server operation method is actually to enable the physical machine that hosts the server to save energy during operation. By dynamically adjusting the load of the physical machine and the load of each heterogeneous resource on the physical machine, the physical machine can be energy-efficient. At the same time, a server cluster built by multiple physical machines will also adopt this method to save energy. On a single physical machine, there will be multiple processors. Therefore, during the operation of the physical machine, if a single physical machine is in a low-load state, the load of the processor on the physical machine will be reduced, so that some processors are in low-power mode, so that the remaining processors can be in a high-efficiency area and can adapt to the physical machine. When a single physical machine is in an overloaded state, the processor will also be in an overloaded state, and this will trigger When task migration is initiated, the input of tasks to the physical machine will be stopped, and some tasks in the physical machine will be migrated to other physical machines in low load or normal state. At the same time, it will ensure that other physical machines receiving tasks will not be in a high load state after receiving the tasks. Dynamic Voltage Frequency Scaling (DVFS) dynamically adjusts the processor frequency and voltage of the physical machine according to the load of the virtual machine, reduces the processor frequency and voltage at low load, reduces dynamic power consumption, and increases the frequency and voltage at high load to meet performance requirements and achieve energy saving. Dynamic voltage frequency scaling is used to adjust the processor voltage to achieve energy saving. The heterogeneous resources on a single physical machine are obtained by combining multiple identical devices. Therefore, when a single physical machine is de-voltaged and de-frequencyed, and can meet the needs of processing the remaining tasks, it will be reflected on multiple identical devices. Some of the identical devices are in low consumption or even sleep state, and the remaining identical devices will be in normal mode. The identical devices in normal mode will process the remaining tasks.

[0135] The device parameters of a physical machine are obtained by combining the operating parameters of various heterogeneous resources. Heterogeneous resources include processing resources, memory resources, storage resources, and network resources, and represent the physical devices that provide these resources. Pooling is a technology that integrates distributed computing, storage, or network resources into a unified resource pool to improve resource utilization, enhance flexibility, and reduce energy consumption. Pooling physically pools the CPU, GPU, FPGA, and other heterogeneous resources in a physical server to build a unified heterogeneous resource pool that includes logical operation units, parallel computing units, and hardware acceleration units. This breaks the physical isolation between resources and enables centralized resource management and scheduling. This allows resources to be dynamically allocated to different tasks or applications based on demand, improving resource sharing and utilization while enhancing system elasticity and flexibility. Virtualization technology divides a single physical server into multiple virtual machines, sharing resources such as processors and memory, and avoiding resource idleness during low load periods. Virtualization technology uses an abstraction layer to logically divide the hardware resources (such as CPU, memory, storage, and network) of a single physical server into multiple independent and isolated virtual machines. Each virtual machine can run different operating systems and applications, achieving resource sharing and efficient utilization.

[0136] Intelligent pairing combination refers to the associated combination of pending tasks whose similarity values are all lower than the degree threshold. Since the feature vector and task amount of the pending task are clear, and the matching feature vector of the virtual heterogeneous resource unit is known, during the calculation, through intelligent pairing, those pending tasks that can be combined are analyzed, so that the similarity between the obtained combined task and the matching feature vector is maximized, and calculation is performed on this basis. In the combination pairing method, if the task is not input into the virtual heterogeneous resource unit within the set time, the degree threshold limit can be directly ignored, and the task can be input according to the characteristics of the virtual heterogeneous resource unit with the highest similarity value. This design is mainly to ensure that the task can be carried out in the absence of adaptation and to avoid task shelving. This situation occurs less frequently, and the task amount is small relative to the capacity of the virtual heterogeneous resource unit, so the impact on the fragment rate of the virtual heterogeneous resource unit is also small.

[0137] By aligning tasks and resources through similarity matching, the fragmentation rate in virtual heterogeneous resource units can be reduced. By setting up several virtual heterogeneous resource units in a single server, several virtual heterogeneous resources can process multiple tasks at the same time, while ensuring that tasks on the same physical machine avoid mutual interference due to the limitations of virtual heterogeneous resources. The system can more accurately assign tasks to virtual heterogeneous resource units that match their resource requirements, which reduces resource fragmentation caused by mismatch between tasks and resources, thereby reducing the fragmentation rate. Through similarity calculation, the system can more reasonably allocate resources to ensure that each task can run on the most suitable resources and reduce the residual resources in virtual heterogeneous resource units. This optimized allocation reduces resource overload and idleness and improves resource utilization. Therefore, when the processor in a low-load state is subsequently reduced in voltage and frequency, the system can release excess resources, so that the processor is in a low-consumption state, reducing processor energy consumption, thereby reducing overall energy consumption, and a single physical After the heterogeneous resources in the physical machine are de-voltaged and de-frequencyed, the tasks originally in a single physical machine and in the virtual heterogeneous resource unit can still continue to operate normally. Through feature selection and similarity selection methods, the tasks are reasonably allocated to the most suitable virtual heterogeneous resource unit, reducing resource fragmentation, avoiding the mismatch between tasks and resources, and improving task processing efficiency. When this method subsequently optimizes energy consumption for a single physical machine, the physical machine is de-voltaged and de-frequencyed. At the same time, the reduction of resource fragments also reduces idle resources, ensuring that the physical machine after de-voltaged and de-frequencyed is in a state of lower energy consumption and can also be used for tasks in the physical machine. At the same time, the heterogeneous resource ratio remains unchanged during the expansion and contraction process to ensure the stability and performance of the system. The scale of the virtual heterogeneous resource unit is dynamically adjusted according to the load conditions, which can not only meet the task requirements under high load but also avoid resource waste under low load. The optimization of the No. 1 framework can be guaranteed through feature selection and similarity selection methods. Task allocation based on feature selection and similarity selection methods results in less resource fragmentation in heterogeneous resources after de-voltaged and de-frequencyed.

[0138] The load rates of different heterogeneous resources are normalized to make them comparable and additive. Different heterogeneous resources (such as CPU, GPU, FPGA, etc.) have different working characteristics and load ranges. Normalization eliminates the impact of dimensional differences by unifying these data of different dimensions or ranges to the same scale (such as between 0 and 1), allowing direct comparison and analysis of the load conditions of different resources. By establishing a load rate formula for virtual heterogeneous resource units and a real-time global load rate calculation method, unified management and scheduling of different heterogeneous resources are achieved. The system can intelligently allocate tasks to the most appropriate resource unit based on the feature vector of the task and the matching feature vector of the virtual heterogeneous resource unit, using similarity calculation. This intelligent scheduling method improves the accuracy of task allocation and optimizes the task processing flow of the entire server system.

[0139] Processor usage has a load range, including low-load, high-efficiency, and overload zones. The corresponding processors are in low-load, normal, and overload states. For example, in the low-load zone (0%-30%), power consumption increases slowly with load, resulting in low energy efficiency. In the high-efficiency zone (30%-70%), power consumption and load have an approximately linear relationship, resulting in optimal energy efficiency. In the overload zone (70%), power consumption increases sharply, but performance gains diminish at a marginal rate (for example, CPU frequency is reduced due to thermal constraints). This ensures that physical machines and physical resources are in the high-efficiency zone, ensuring energy conservation and processing efficiency. For physical machines in the overload zone, their load is reduced, and some of their tasks are transferred to other physical machines. When triggering task migration, virtual heterogeneous resource units with the same heterogeneous resource ratio on other servers are prioritized. However, if other servers do not have virtual heterogeneous resource units with the same heterogeneous resource ratio, virtual heterogeneous resource units on other physical machines are reselected using a similar selection method. Note that other physical machines cannot be highly loaded after receiving tasks. Similarly, the load range within a physical machine also has similar characteristics.

[0140] Virtual heterogeneous resource units with the same heterogeneous resource ratio across multiple servers are interconnected through a cross-node method to form a global resource pool. When task migration is triggered, virtual heterogeneous resource units with the same heterogeneous resource ratio in other servers will be selected through the global resource pool. However, if other servers do not have virtual heterogeneous resource units with the same heterogeneous resource ratio, other virtual heterogeneous resource units will be reselected through a similar selection method.

[0141] The real-time global load rate refers to the overall load rate of each heterogeneous resource in a single physical machine. Since a physical machine is composed of multiple heterogeneous resources, the real-time global load rate can be obtained by calculation. It represents the overall proportion of the i-th heterogeneous resource in the physical machine. The overall proportion can be calculated by the energy consumption of the heterogeneous resources under normal circumstances. Different heterogeneous resources (such as CPU, GPU, FPGA, etc.) have different energy consumption performance when processing tasks. The energy consumption of these resources is not only closely related to their load rate (that is, the ratio of the actual processed task volume to its maximum processing capacity), but by measuring the energy consumption of heterogeneous resources, its load rate can be estimated, and then the overall proportion of the resource in the physical machine can be calculated. Moreover, the energy consumption proportion can be used to determine which group has the highest energy consumption. Since its energy consumption is the highest, its impact on the real-time global load rate is also the greatest. Therefore, in this application, the energy consumption proportion is used instead of the overall proportion.

[0142] Establish task load regulation for physical machines. For example, when the real-time global load rate is higher than 75% for three consecutive minutes, it indicates that the current task load of the server is heavy and the physical machine can no longer meet the task requirements. At this time, the virtual unit scale is expanded in proportion to the growth rate of the task requirements. For example, if the task requirements increase by 20%, the virtual unit scale is also increased by 20% accordingly. This can allocate more computing resources to the tasks, improve the system's processing capabilities, and ensure that tasks can be processed in a timely and efficient manner. When the real-time global load rate is lower than 30% for five consecutive minutes, it indicates that the current task load of the server is light and some virtual units have low resource utilization, resulting in resource waste. These low-utilization virtual units are shut down, and the tasks in them are migrated to other virtual units with higher utilization, and the power supply of the merged redundant physical nodes is turned off. When transferring tasks, it is necessary to ensure that the transferred tasks are also selectively transferred according to similar selection methods and coordination requirements.

[0143] By identifying the degree of coordination required, a virtual heterogeneous resource unit is selected so that the degree of collaborative computing of the heterogeneous resources in the virtual heterogeneous resource unit is close to the identified degree of coordination required. This enables the virtual heterogeneous resource unit to fully utilize the various heterogeneous resources in the virtual heterogeneous resource unit when processing the corresponding task, avoiding the situation where a single heterogeneous resource has an excessive utilization rate while other heterogeneous resources have an excessive utilization rate.

[0144] At the same time, since the processor will handle multiple tasks, multiple tasks can be combined with each other to input corresponding virtual heterogeneous resource units. After the two tasks are combined, the original feature vector is changed, so that the matching feature vector of the virtual heterogeneous resource unit can be adapted. The two feature vectors are calculated by cosine similarity to obtain a similarity value. If the similarity value of a single task is lower than the degree threshold, it means that the similarity between the task and the virtual heterogeneous resource unit is low. Therefore, it is not recommended to directly input it into the virtual heterogeneous resource unit. At this time, it will be combined with other single tasks or multiple tasks with low similarity through a combination matching method to obtain a combined task with high similarity. The combined task will be input according to the direct matching method. At the same time, if a single task with low similarity fails to be combined within the set time, such as within two seconds, it will be input regardless of the degree threshold to avoid the task being idle for too long.

[0145] The load rate of the virtual heterogeneous resource unit is compared with the load threshold of the virtual heterogeneous resource unit. The total energy supply scale of the virtual heterogeneous resource unit is dynamically expanded or reduced based on the comparison results. The heterogeneous resource ratio is maintained during the expansion and reduction process. When the load rate of all virtual heterogeneous resource units is higher than 75% for 3 consecutive minutes, and the global load rate is When the load rate is less than the maximum value of the second load threshold, the virtual unit scale will be expanded proportionally. When the load rate of the virtual heterogeneous resource unit is lower than 25% for 5 consecutive minutes, the virtual unit scale will be reduced proportionally. When several virtual heterogeneous resource units in a single physical machine are reduced in scale simultaneously, the voltage and frequency reduction of the single physical machine can be triggered at this time, so that the single physical machine can achieve energy saving while being able to process tasks.

[0146] The actual scale of different heterogeneous resources can be regarded as the maximum operating load of heterogeneous resources under normal conditions. The total heterogeneous scale refers to the minimum load at which several virtual heterogeneous resource units can process tasks and maintain efficiency after reduction. By comparing the two, the source devices of heterogeneous resources can be dynamically load-adjusted so that after adjustment, the physical machine can meet the minimum load required to process tasks and maintain efficiency.

[0147] Contrast coefficient Between 0.8-1.2, by comparing the values Directly perform dynamic load adjustment on heterogeneous resources, comparing coefficients There will be changes, and the unit energy consumption of several heterogeneous resources after modification needs to be able to meet the processing of tasks in the physical machine. The total heterogeneous scale is compared with the total scale of the physical machine to obtain a comparison value set. The comparison value set is a comparison of the total scale of each heterogeneous resource in the heterogeneous total scale and the total scale of each heterogeneous resource in the physical machine to obtain a comparison value set.

[0148] Example 1: CPU, GPU, memory three resource scenarios, resource type and parameter settings, CPU normalized load rate L1 = 0.7, energy consumption ratio A1 = 0.5, GPU normalized load rate L2 = 0.6, energy consumption ratio A2 = 0.3, memory normalized load rate L3 = 0.5, energy consumption ratio A3 = 0.2, perform global load rate Calculation, =(0.5×0.7)+(0.3×0.6)+(0.2×0.5)=0.35+0.18+0.10=0.63, global load factor is 0.63, the global load rate Compare it with the second load threshold to determine the load situation.

[0149] Example 2: Parameter settings in virtual heterogeneous resource units, resource allocation weights: α=0.4 (CPU), β=0.3 (memory), γ=0.2 (disk), θ=0.1 (network), normalized load rate: processor load =0.8 (CPU load rate is high), memory load =0.6, disk load =0.4, network bandwidth utilization =0.5, calculation process =(0.4×0.8)+(0.3×0.6)+(0.2×0.4)+(0.1×0.5)=0.32+0.18+0.08+0.05=0.63, the load data in the virtual heterogeneous resource unit is calculated, and combined with the load data in the virtual heterogeneous resource unit and the ratio of the scale of the virtual heterogeneous resource unit to the total scale of the physical machine, the impact of the load data in the virtual heterogeneous resource unit on the total load of the physical machine can be obtained.

[0150] The present invention establishes a No. 1 framework, which supports collaborative management and scheduling among multiple heterogeneous resources. The No. 1 framework dynamically adjusts the scale of virtual heterogeneous resource units for a single physical machine through load calculation, comparison and trigger conditions. Based on the reduced scale, the No. 1 framework simultaneously performs different voltage and frequency reductions on each heterogeneous resource in a single physical machine. While reducing the energy consumption of the physical machine, the No. 1 framework reduces the voltage and frequency of each heterogeneous resource separately, thereby ensuring effective collaborative scheduling of heterogeneous resources. At the same time, the No. 2 framework targets a server cluster. Based on the comparison between the real-time global load rate and the load threshold, when a specific No. 2 trigger condition is met, the No. 1 framework maintains the physical machine in a normal state through task migration and resource reallocation, avoids being in a high-energy consumption state, and optimizes the energy consumption of the entire cluster.

[0151] Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is limited by the accompanying embodiments and their equivalents.

Claims

1. An energy-saving server operation method based on dynamic load adjustment, characterized in that: The energy-saving server operation method includes: Multiple heterogeneous resources in a single physical machine are pooled and virtualized to generate multiple virtual heterogeneous resource units, with each virtual heterogeneous resource unit having a different ratio of heterogeneous resources. Establishing a second load threshold of the physical machine and a first load threshold of the virtual heterogeneous resource unit based on the device parameters of the physical machine; Build a framework for optimizing the energy consumption of a single physical machine and a framework for optimizing the energy consumption of a server cluster; The first framework includes: Obtain the load rates of several virtual heterogeneous resource units and the real-time global load rate through load calculation , establish the number one trigger condition for optimizing the energy consumption of physical machines; The comparison method is used to compare the load rates of several virtual heterogeneous resource units with the load threshold of the virtual heterogeneous resource units, and the real-time global load rate is calculated. Compare it with the load threshold No. 2 to obtain the comparison result No. 1; Trigger condition No. 1 determines whether to enable energy consumption optimization of the physical machine based on comparison result No.

1. When comparison result No. 1 meets the trigger condition in trigger condition No. 1, energy consumption optimization of the single physical machine is triggered. The scale of several virtual heterogeneous resource units in the physical machine is adjusted based on elastic resource scaling technology. The total heterogeneous scale of the combination of the several virtual heterogeneous resource units after adjustment is obtained. The total heterogeneous scale is compared with the total scale of the physical machine to obtain a comparison value set. Dynamic load adjustment is performed on several heterogeneous resources in the single physical machine through the comparison value set and the adjustment method. The comparison value set includes the comparison values between the actual scale of several heterogeneous resources in the physical machine and the virtual scale of several heterogeneous resources in the total heterogeneous scale. , i represents the i-th heterogeneous resource; The adjustment method includes: dynamically adjusting the load of a plurality of heterogeneous resource source devices, and adjusting the energy consumption comparison value between the unit energy consumption of the plurality of heterogeneous resources after modification and the unit energy consumption of the heterogeneous resources before modification. , energy consumption comparison value and contrast value Positive correlation; The second framework includes: Obtain the real-time global load rate of the physical machine through load calculation ; Real-time global load factor achieved through comparison method Compare the load threshold with the second load threshold to obtain the second comparison result; The second trigger condition determines whether to enable the energy consumption optimization of the server cluster based on the second comparison result. When the second comparison result meets the trigger condition in the second trigger condition, the energy consumption optimization of the server cluster is triggered; Get real-time global load rate The difference between the load threshold and the second load threshold is used to establish a safety value, and the real-time global load rate is used to Divide the tasks in the physical machine into proportions to obtain a task scale ratio, obtain the scale of tasks to be processed based on the difference, the safety value, and the task scale ratio, and obtain the tasks to be processed from the virtual heterogeneous resource unit with the highest load rate; The tasks to be processed in the physical machine are transferred between physical machines through feature selection method; The feature selection method includes selecting from the server cluster based on the scale of the task to be processed that can receive the task to be processed and will not cause the real-time global load rate after receiving the task. For physical machines that exceed the second load threshold, pending tasks are sent to the selected physical machine; The pending tasks are processed, and the task processing includes: Perform feature processing on the pending tasks to obtain the coordination requirements of the pending tasks. The coordination requirements represent the task's demand for different heterogeneous computing resources. A feature vector of the task is established based on the coordination requirements. A feature vector of the virtual heterogeneous resource unit is established based on the ratio of each heterogeneous resource in the virtual heterogeneous resource unit. The tasks to be processed are input to corresponding virtual heterogeneous resource units through a similar selection method. Dynamic load regulation includes adjusting the voltage and frequency of the heterogeneous resources.

2. The energy-saving server operation method based on dynamic load adjustment according to claim 1, characterized in that: The similarity selection method includes a direct pairing method and a combined pairing method; The direct pairing method includes: Calculate the similarity between the feature vector of the task to be processed and the matching feature vectors of several virtual heterogeneous resource units in the physical machine to obtain a similarity degree value, select the feature of the virtual heterogeneous resource unit with the highest similarity degree value as the input of the task to be processed, establish a degree threshold, and input the tasks whose similarity degree values are all lower than the degree threshold through the combined pairing method; The combination pairing method includes: Intelligently pair and combine pending tasks whose similarity values are all below a threshold to obtain the combined task and its feature vector. Calculate the similarity between the feature vector of the combined task and the feature vector of several virtual heterogeneous resource units in the physical machine to obtain a similarity value set. Select the feature of the virtual heterogeneous resource unit with the highest similarity value as the input of the combined task. A time threshold is established. If the tasks in the combination matching method cannot be combined when the time threshold is exceeded, the degree threshold is ignored and the virtual heterogeneous resource unit feature with the highest similarity value is directly selected as the input of the combination task.

3. The energy-saving server operation method based on dynamic load adjustment according to claim 1, characterized in that: The load calculation includes normalization processing and load calculation; The normalization process includes establishing a normalization formula: ,The minimum load rate and the maximum load rate are the theoretical minimum load rate and the theoretical maximum load rate of each heterogeneous resource, respectively; The load of each heterogeneous resource is calculated using a normalization formula to obtain a normalized load rate of each heterogeneous resource.

4. The energy-saving server operation method based on dynamic load adjustment according to claim 1, characterized in that: Elastic scaling technology involves dynamically scaling the scale of each heterogeneous resource in a virtual heterogeneous resource unit, while maintaining the same heterogeneous resource ratio during the scaling process. The comparison value set includes comparison values between the actual scale of each heterogeneous resource in the physical machine and the virtual scale of each heterogeneous resource in the total heterogeneous scale; Energy consumption comparison value and contrast value Positive correlation, establish contrast coefficient , .

5. The energy-saving server operation method based on dynamic load adjustment according to claim 3, characterized in that: The load calculation includes: Real-time global load rate ; ,N represents the number of groups in the heterogeneous resource grouping set, represents the overall proportion of the i-th heterogeneous resource in the physical machine, represents the normalized load rate of the i-th heterogeneous computing resource, Indicates the real-time global load rate based on heterogeneous resource load calculation; The overall proportion is replaced by the energy consumption proportion. The energy consumption proportion represents the energy consumption weight of several heterogeneous resources in the physical machine, and the sum of the energy consumption proportions must be 1.

6. The energy-saving server operation method based on dynamic load adjustment according to claim 1, characterized in that: The second trigger condition includes: Real-time global load rate The load value is continuously greater than or equal to the maximum value of the second load threshold for at least three minutes.

7. The energy-saving server operation method based on dynamic load adjustment according to claim 5, characterized in that: The load calculation includes: Establishing a formula for the load rate of virtual heterogeneous resource units ; , represents the load rate of the jth virtual heterogeneous resource unit, the normalized processor load , normalized memory load , normalized disk load , normalized network bandwidth utilization ; 、 、 and Indicates the energy consumption proportion of virtual heterogeneous resource units, and the sum of the energy consumption proportions must be 1; , Y represents the number of virtual heterogeneous resource units in the physical machine, It represents the ratio of the size of the Y-th virtual heterogeneous resource unit to the total size of the physical machine. Indicates the real-time global load rate calculated based on the load of virtual heterogeneous resource units.

8. The energy-saving server operation method based on dynamic load adjustment according to claim 1, characterized in that: The first trigger condition includes: Condition 1: The virtual heterogeneous resource unit (VHRU) remains below the VHRU threshold for at least three minutes. Condition 2: Real-time global load rate The load is continuously lower than the minimum value of the second load threshold for at least three minutes; Conditions 1 and 2 must be met at the same time.

9. An energy-saving server operation system based on dynamic load adjustment, adopting the energy-saving server operation method based on dynamic load adjustment according to any one of claims 1 to 8, characterized in that: The energy-saving server operation system includes: Resource allocation module: Multiple heterogeneous resources in a single physical machine are pooled and virtualized to generate multiple virtual heterogeneous resource units. The ratio of heterogeneous resources in each virtual heterogeneous resource unit is different. Threshold establishment module: establishes the second load threshold of the physical machine and the first load threshold of the virtual heterogeneous resource unit based on the device parameters of the physical machine; Energy consumption optimization module: The energy consumption optimization module includes the first optimization module and the second optimization module; Optimization module No. 1: Obtains the load rates of several virtual heterogeneous resource units and the real-time global load rate through load calculation , establish the number one trigger condition for optimizing the energy consumption of physical machines; The comparison method is used to compare the load rates of several virtual heterogeneous resource units with the load threshold of the virtual heterogeneous resource units, and the real-time global load rate is calculated. Compare it with the load threshold No. 2 to obtain the comparison result No. 1; Trigger condition No. 1 determines whether to enable energy consumption optimization of the physical machine based on comparison result No.

1. When comparison result No. 1 meets the trigger condition in trigger condition No. 1, energy consumption optimization of the single physical machine is triggered. The scale of several virtual heterogeneous resource units in the physical machine is reduced based on elastic resource scaling technology. The total heterogeneous scale of the combination of the several virtual heterogeneous resource units after reduction is obtained. The total heterogeneous scale is compared with the total scale of the physical machine to obtain a comparison value set. Dynamic load adjustment is performed on several heterogeneous resource source devices in the single physical machine through the comparison value set and the adjustment method. Optimization module 2: Obtain the real-time global load rate of the physical machine through load calculation ; Real-time global load factor achieved through comparison method Compare the load threshold with the second load threshold to obtain the second comparison result; The second trigger condition determines whether to enable server cluster energy consumption optimization based on the second comparison result. When the second comparison result meets the trigger condition in the second trigger condition, the server cluster energy consumption optimization is triggered: Get real-time global load rate The difference between the load threshold and the second load threshold is used to establish a safety value, and the real-time global load rate is used to Divide the tasks in the physical machine into proportions to obtain a task scale ratio, obtain the scale of tasks to be processed based on the difference, the safety value, and the task scale ratio, and obtain the tasks to be processed from the virtual heterogeneous resource unit with the highest load rate; The pending tasks in the physical machine are transferred between physical machines through the feature selection method.

Images