A method and system for intelligent allocation of machine room computing power resources for machine room resource survey

Abstract

The present application relates to the technical field of data center resource management, and more particularly to a method and system for intelligent allocation of computer room computing power resources for computer room resource survey. The present application directly embeds the node real-time attenuation coefficient and the node comprehensive energy efficiency score into the heuristic function of the ant colony algorithm, realizes the real-time linkage of scheduling decision and computer room physical environment, significantly improves the response speed of the improved ant colony algorithm to physical environment mutation, and when the node real-time attenuation coefficient decreases due to aging or local overheating of the node, the heuristic factor will immediately decrease, thereby preventing the ant from selecting this path at the probability level, and the fault risk can be avoided without waiting for the slow evaporation of pheromone. This design effectively overcomes the defect of traditional algorithms that continuously dispatch orders to faulty nodes due to feedback lag, and greatly enhances the robustness and physical adaptability of the system at a very low implementation cost without changing the main structure of the ant colony algorithm.

Description

Technical Field

[0001] This invention relates to the field of data center resource management technology, and in particular to a method and system for intelligent allocation of computing resources in a data center for a comprehensive survey of data center resources. Background Technology

[0002] With the rapid development of technologies such as cloud computing, big data, and artificial intelligence, the scale and complexity of data centers are increasing daily. As the core physical carrier of a data center, the rational allocation and efficient utilization of its computing, storage, and network resources directly affect the overall performance and operating costs of the data center.

[0003] In existing data center resource management, the first step is usually to conduct a comprehensive survey of data center resources. This involves collecting information such as hardware configuration, operating status, and network connectivity of each device in the data center through methods such as manual inventory, SNMP, IPMI, or dedicated asset management software. Then, based on the static or quasi-static information obtained from this survey, administrators may use scheduling algorithms such as ant colony optimization to allocate computing tasks to the appropriate server nodes.

[0004] In existing ant colony-based computing power scheduling schemes, the state transition probability is typically determined by both pheromone concentration and a heuristic function. Current technologies often employ fixed heuristic functions, such as using only the remaining computing power of nodes or only the reciprocal of network bandwidth. This static design cannot adapt to dynamically changing loads and physical environments within the data center. Furthermore, when computing power quality deteriorates in a certain area of ​​the data center due to external factors such as air conditioning malfunctions causing localized temperature increases or power grid fluctuations, the standard ant colony algorithm relies solely on the slow evaporation of pheromones to adjust path selection, resulting in a delayed response. During this period, the ant colony algorithm will continue to incorrectly allocate tasks to that area, leading to an increased allocation failure rate.

[0005] Meanwhile, in actual data centers, when using the original state transition probability formula for task allocation, if multiple tasks consecutively select the same high-performance node within a short period of time, the load on that node will quickly become saturated. The state transition probability formula cannot predict this impending congestion before the selection action occurs, resulting in node stampede in the allocation results on a micro-timescale. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention provides a method and system for intelligent allocation of computing resources in data centers for data center resource surveys, thereby resolving the issues existing in the prior art.

[0007] This invention provides a method for intelligent allocation of computing resources in a data center for a comprehensive survey of data center resources. The method includes the following steps:

[0008] S1: Collect multi-dimensional survey data of computer room resources;

[0009] S2: Perform data preprocessing and normalization operations on the multidimensional census data;

[0010] S3: Calculate the node dynamic computing power attenuation coefficient and the node comprehensive energy efficiency score based on the normalized multidimensional census data;

[0011] S4: Calculate the heuristic function of the improved ant colony algorithm based on the node's comprehensive energy efficiency score and the node's dynamic computing power attenuation coefficient, and allocate computing power resources according to the improved ant colony algorithm;

[0012] Specifically, S4 consists of: S4.1: Obtaining the set of currently available nodes in the data center and initializing the pheromone matrix; S4.2: Calculating the heuristic function of the improved ant colony algorithm based on the node's comprehensive energy efficiency score and the node's dynamic computing power attenuation coefficient, obtaining the heuristic factor, and performing iterative search and solution construction; S4.3: Performing global pheromone updates, termination determination, and output, thereby obtaining the computing power resource allocation scheme.

[0013] Preferably, in step S4.2, the expression for the heuristic function of the improved ant colony algorithm is:

[0014] ;

[0015] In the formula, Choose the heuristic function value for node j for task i. The overall energy efficiency score for node j. Let ξ be the dynamic computing power attenuation coefficient of node j, and ξ be the adjustment coefficient.

[0016] Preferably, for each task i to be assigned, the following sub-steps are performed:

[0017] The heuristic function of the improved ant colony algorithm is calculated based on the node's comprehensive energy efficiency score and the node's dynamic computing power attenuation coefficient, and the heuristic factor is obtained.

[0018] Using the improved state transition probability formula, calculate the probability that task i selects each node j in the set;

[0019] The improved state transition probability formula is as follows:

[0020] ;

[0021] Where q is a random number uniformly distributed in the interval [0,1]. For pseudo-random thresholds, The current set of allowed nodes In the middle, making The node that takes the maximum value. For pheromone concentration, Here, α is the heuristic factor, β is the pheromone importance factor, and β is the heuristic function importance factor.

[0022] Then, for task i, iterate through all optional nodes j and calculate... Find the value of the node with the largest value and denote it as . Then, generate a random number q, if Then task i is directly assigned to node i. ,like Then calculate the probability of all nodes and select the target node by roulette wheel selection; finally, determine the target node for task i. ;

[0023] Bind task i to the target node. Update the target node The real-time status is calculated, and the corresponding computing power and memory resources are deducted, and the CPU utilization of the target node is estimated and updated. Mark the remaining available capacity of the node in the current iteration;

[0024] After all m tasks have completed node selection, the current iteration count is [number missing]. .

[0025] Preferably, in S4.3, the global pheromone update specifically involves:

[0026] Perform a global evaporation operation on the pheromone matrix to reduce the weight of historical paths and increase the ability to explore new paths;

[0027] To evaluate the quality of the allocation scheme generated in this iteration, a global cost function, Cost, is defined. If the Cost of this iteration is less than the historical best cost... Then perform the following operations: update the historical best allocation scheme, and increase the pheromone concentration for all paths (i,j) that were successfully selected in this iteration.

[0028] Preferably, the termination determination and output are performed as follows:

[0029] Determine if the current iteration number NC has reached the preset maximum iteration number. If yes, end the algorithm process; otherwise, start a new round of search.

[0030] Preferably, in step S4.1, preset ant colony algorithm running parameters are read, including: the number of ants m, the maximum number of iterations. The parameters include pheromone importance factor α, heuristic function importance factor β, pheromone volatile coefficient ρ, and adjustment coefficient ξ. Simultaneously, the comprehensive energy efficiency score of all nodes in all data centers is obtained. The dynamic computing power attenuation coefficient δ of the nodes is calculated, and then nodes in maintenance or faulty state are excluded to construct the set of allowed nodes. .

[0031] Preferably, in step S3, the formula for calculating the node dynamic computing power attenuation coefficient δ is:

[0032] ;

[0033] In the formula, α is the aging sensitivity coefficient. β represents the service life of the node equipment, and β is the temperature drift sensitivity coefficient. This refers to the microenvironmental temperature of the cabinet where the node is located. The reference temperature for the equipment design.

[0034] Preferably, boundary processing conditions are set to constrain the value of the dynamic computing power attenuation coefficient of the node, specifically: when At this time, the thermal resistance attenuation term will theoretically be less than or equal to 0. Therefore, the dynamic computing power attenuation coefficient δ of the node is forcibly set to 0.05. At this time, theoretically, the dynamic computing power attenuation coefficient δ of the node is greater than 1. Therefore, the dynamic computing power attenuation coefficient δ of the node is forcibly set to 1.

[0035] Preferably, the comprehensive energy efficiency score The calculation formula is:

[0036] ;

[0037] In the formula, The comprehensive energy efficiency score is given. The larger the value, the more suitable the node is for receiving new tasks. The unit is a dimensionless composite score value. The nominal computing power of a node, This refers to the CPU utilization of the target node.

[0038] According to another aspect of the present invention, a smart allocation system for computing resources in a data center for a survey of data center resources is provided. The system employs the aforementioned smart allocation method for computing resources in a data center for a survey of data center resources. The system includes:

[0039] The data acquisition module is used to collect multi-dimensional survey data of computer room resources;

[0040] The data preprocessing module is used to perform data preprocessing and normalization operations on the multidimensional census data;

[0041] The node dynamic determination module is used to calculate the node dynamic computing power attenuation coefficient and the node comprehensive energy efficiency score based on the normalized multidimensional census data.

[0042] The computing power allocation module is used to calculate the heuristic function of the improved ant colony algorithm based on the node's comprehensive energy efficiency score and the node's dynamic computing power attenuation coefficient, and to allocate computing power resources according to the improved ant colony algorithm.

[0043] The embodiments of the present invention have the following technical effects:

[0044] This invention directly embeds the real-time decay coefficient of nodes and the comprehensive energy efficiency score of nodes into the heuristic function of the ant colony algorithm, realizing real-time linkage between scheduling decisions and the physical environment of the data center. This significantly improves the response speed of the improved ant colony algorithm to sudden changes in the physical environment. When the real-time decay coefficient of a node decreases due to aging or local overheating, the heuristic factor will immediately decrease accordingly, thereby preventing ants from choosing that path at the probabilistic level. It avoids the risk of failure without waiting for the pheromone to slowly evaporate. This design effectively overcomes the defect of traditional algorithms that continuously dispatch orders to faulty nodes due to feedback lag. Without changing the main architecture of the ant colony algorithm, it greatly enhances the robustness and physical adaptability of the system with extremely low implementation cost.

[0045] This invention establishes a balance between exploration and utilization in ant colony optimization by introducing a pseudo-random proportional rule, significantly improving the performance and stability of computing resource allocation. It utilizes a threshold q0 to dynamically switch algorithm behavior: when the random number q ≤ q0, a deterministic strategy is adopted to directly select the current optimal node. This mechanism effectively avoids the problem of traditional roulette wheel algorithms incorrectly allocating tasks to inferior nodes in early iterations due to excessive randomness, significantly shortening the algorithm's convergence time and reducing resource overhead from ineffective scheduling. Simultaneously, this scheme retains the probabilistic selection mechanism when q > q0, ensuring the algorithm does not get stuck in a local optimum loop and maintaining global optimization capabilities. By frequently selecting the optimal node, it avoids multiple concurrent tasks simultaneously occupying the same high-load node due to random fluctuations, thereby significantly reducing the task allocation failure rate and improving the overall response speed and robustness of data center computing resource scheduling. Attached Figure Description

[0046] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0047] Figure 1 This is a flowchart of an intelligent allocation method for data center computing resources for data center resource survey, provided by an embodiment of the present invention;

[0048] Figure 2This is a flowchart of the allocation of computing resources based on the improved ant colony algorithm provided in an embodiment of the present invention. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0050] Example 1, Figure 1 A flowchart of an intelligent allocation method for computing resources in a data center, used for a comprehensive survey of data center resources, is shown. Figure 1 As shown, a method for intelligent allocation of computing resources in a data center for data center resource surveys includes the following steps:

[0051] S1: Collect multi-dimensional survey data of computer room resources;

[0052] In this step, data is collected from the smallest computing unit within the data center, namely the physical server node, and its immediate physical environment. Through an out-of-band management system and sensor network, multidimensional survey data is collected for each physical server node within the data center. This multidimensional survey data includes: the node's nominal computing power. Service life of node equipment Current CPU utilization Current memory utilization and the microenvironmental temperature of the rack where the node is located. .

[0053] Furthermore, the nominal computing power of the node The sampling frequency is triggered by a change, and the node device's service life is also specified. The data collection frequency is once a day, and the current CPU utilization rate... and the current memory utilization rate The sampling frequency is 5 seconds / time, and the micro-environment temperature of the cabinet where the node is located is... The sampling frequency is 10 seconds per time.

[0054] S2: Perform data preprocessing and normalization operations on the multidimensional census data;

[0055] The data preprocessing operations include outlier removal, timestamp alignment, and data encapsulation. Outlier removal utilizes the Raida criterion to remove outliers generated by transient jumps in the sensor data. Due to the different acquisition frequencies from different sources, all data are aligned to a uniform 5-second time window using linear interpolation to ensure consistency in subsequent formula calculations. All data from the same node are encapsulated into a single JSON object.

[0056] The multidimensional census data are mapped to the standard interval [0,1] using the minimum-maximum linear normalization method.

[0057] S3: Calculate the node dynamic computing power attenuation coefficient and the node comprehensive energy efficiency score based on the normalized multidimensional census data;

[0058] In this step, the node dynamic computing power attenuation coefficient is a dimensionless scalar with a value range of (0,1]. When the node dynamic computing power attenuation coefficient δ is 1, it means that the node is in an ideal state and its nominal computing power is usable without loss. When the node dynamic computing power attenuation coefficient approaches 0, it means that the node has lost its effective computing power output capability due to severe aging or extreme thermal environment.

[0059] Specifically, the formula for calculating the node dynamic computing power attenuation coefficient δ is as follows:

[0060] ;

[0061] In the formula, α is the aging sensitivity coefficient, which in this embodiment has a typical value of 0.15. Where β represents the service life of the node equipment, and β is the temperature drift sensitivity coefficient, which in this embodiment has a typical value of 0.02. This refers to the microenvironmental temperature of the cabinet where the node is located. The reference temperature designed for the device, with a default value of 25℃.

[0062] Furthermore, to prevent extreme environments from causing the node's dynamic computing power attenuation coefficient to become negative or zero, this embodiment sets boundary processing conditions to constrain the value of the node's dynamic computing power attenuation coefficient. Specifically, when... At this time, the thermal resistance attenuation term will theoretically be less than or equal to 0. Therefore, the dynamic computing power attenuation coefficient δ of the node is forcibly set to 0.05. At this time, theoretically, the dynamic computing power attenuation coefficient δ of the node is greater than 1. At this time, the dynamic computing power attenuation coefficient δ of the node is forcibly set to 1, that is, it is considered that when the temperature is lower than the reference temperature, heat dissipation has no benefit to the performance of the computer room.

[0063] In this step, the node comprehensive performance score is set to accurately reflect the relative capacity and available capacity of each physical node to undertake new tasks at the current moment, serving as a direct basis for subsequent task scheduling decisions; wherein, the node comprehensive performance score The calculation formula is:

[0064] ;

[0065] In the formula, The overall performance score of a node is given. The higher the score, the more suitable the node is for receiving new tasks. The unit is a dimensionless composite score. The nominal computing power of a node, This represents the current CPU utilization rate.

[0066] S4: Calculate the heuristic function of the improved ant colony algorithm based on the node's comprehensive energy efficiency score and the node's dynamic computing power attenuation coefficient, and allocate computing power resources according to the improved ant colony algorithm;

[0067] In existing ant colony-based computing power scheduling schemes, the state transition probability is typically determined by both pheromone concentration and a heuristic function. Current technologies often employ fixed heuristic functions, such as using only the remaining computing power of nodes or only the reciprocal of network bandwidth. This static design cannot adapt to dynamically changing loads and physical environments within the data center. Furthermore, when computing power quality deteriorates in a certain area of ​​the data center due to external factors such as air conditioning malfunctions causing localized temperature increases or power grid fluctuations, the standard ant colony algorithm relies solely on the slow evaporation of pheromones to adjust path selection, resulting in a delayed response. During this period, the ant colony algorithm will continue to incorrectly allocate tasks to that area, leading to an increased allocation failure rate.

[0068] This invention defines the heuristic function as a function of the node's overall performance score and the decay coefficient, ensuring that the heuristic function is updated in real time with changes in the physical environment in each iteration; specifically, as follows... Figure 2 As shown, S4 specifically includes:

[0069] S4.1: Obtain the set of currently available nodes in the computer room and initialize the pheromone matrix;

[0070] In this step, the preset ant colony algorithm running parameters are read, including: the number of ants m, and the maximum number of iterations. The parameters include pheromone importance factor α, heuristic function importance factor β, pheromone volatile coefficient ρ, and adjustment coefficient ξ. Simultaneously, the comprehensive energy efficiency score of all nodes in all data centers is obtained. The dynamic computing power attenuation coefficient δ of the nodes is calculated, and then nodes in maintenance or faulty state are excluded to construct the set of allowed nodes. .

[0071] The initialization of the pheromone matrix specifically involves: initializing the pheromone concentration matrix. To prevent the algorithm from stalling in the early stages, the initial pheromone concentration for all paths is set to the same small constant. .

[0072] S4.2: Calculate the heuristic function of the improved ant colony algorithm based on the node's comprehensive energy efficiency score and the node's dynamic computing power attenuation coefficient, obtain the heuristic factor, and perform iterative search and solution construction;

[0073] The expression for the heuristic function of the improved ant colony algorithm is as follows:

[0074] ;

[0075] In the formula, Choose the heuristic function value for node j for task i. The overall energy efficiency score for node j. ξ is the dynamic computing power attenuation coefficient of node j, and ξ is the adjustment coefficient used to control the correction strength of the attenuation coefficient on the heuristic function.

[0076] When the ambient temperature of node j is normal and the equipment is not aging, the dynamic computing power attenuation coefficient of node j is... The heuristic function value of task i for selecting node j approaches 1. The positive amplification is achieved, and the path is preferentially selected. When node j experiences a dynamic computing power attenuation due to overheating or aging, the dynamic computing power attenuation coefficient of node j is... When the value decreases significantly, task i selects the heuristic function value of node j. The selection bias of a node will be significantly reduced, even if its overall energy efficiency score remains numerically competitive. The improved ant colony algorithm will also significantly reduce the probability of selecting that node.

[0077] For each task i to be assigned, perform the following sub-steps:

[0078] The heuristic function of the improved ant colony algorithm is calculated based on the node's comprehensive energy efficiency score and the node's dynamic computing power attenuation coefficient, and the heuristic factor is obtained.

[0079] Using the improved state transition probability formula, calculate the probability that task i selects each node j in the set;

[0080] In existing technologies, the formula for the original state transition probability is usually in the following form:

[0081] ;

[0082] In the formula, For pheromone concentration, Let α be the heuristic factor, β be the pheromone importance factor, and β be the heuristic function importance factor. This formula assumes that the selection of each node is independent. However, in actual data centers, when multiple tasks consecutively select the same high-performance node within a short period, the node's load will quickly become saturated. The state transition probability formula cannot predict this impending congestion before the selection action occurs, leading to a stampede in the allocation results on a micro-timescale. Therefore, this embodiment uses a pseudo-random proportional rule to reconstruct the state transition probability formula, no longer relying solely on probability roulette, but instead setting a random threshold. Each time a node is selected, a uniformly distributed random number q is generated. If not, then no probability selection is performed; instead, the optimal node in the current state is directly selected. If the original probability formula is still used for selection, then the improved state transition probability formula is as follows:

[0083] ;

[0084] Where q is a random number uniformly distributed in the interval [0,1]. For pseudo-random thresholds, The current set of allowed nodes In the middle, making The node that takes the maximum value. For pheromone concentration, α is the heuristic factor, β is the pheromone importance factor, and β is the heuristic function importance factor.

[0085] Then, for task i, iterate through all optional nodes j and calculate... Find the value of the node with the largest value and denote it as . Then, generate a random number q, if Then task i is directly assigned to node i. ,like Then calculate the probability of all nodes and select the target node by roulette wheel selection; finally, determine the target node for task i. ;

[0086] Bind task i to the target node. Update the target node The real-time status is calculated, and the corresponding computing power and memory resources are deducted, and the CPU utilization of the target node is estimated and updated. This marks the remaining available capacity of the node in the current iteration.

[0087] After all m tasks have completed node selection, the current iteration count is [number missing]. .

[0088] This embodiment establishes a balance between exploration and utilization in the ant colony algorithm by introducing a pseudo-random ratio rule, significantly improving the performance and stability of computing resource allocation. It utilizes a threshold q0 to dynamically switch algorithm behavior: when the random number q ≤ q0, a deterministic strategy is adopted to directly select the current optimal node. This mechanism effectively avoids the problem of traditional roulette wheel algorithms incorrectly allocating tasks to inferior nodes in early iterations due to excessive randomness, significantly shortening the algorithm's convergence time and reducing resource overhead from ineffective scheduling. Simultaneously, this scheme retains the probabilistic selection mechanism when q > q0, ensuring the algorithm does not get stuck in a local optimum loop and maintaining global optimization capabilities. By frequently selecting the optimal node, it avoids multiple concurrent tasks simultaneously occupying the same high-load node due to random fluctuations, thereby significantly reducing the task allocation failure rate and improving the overall response speed and robustness of the data center's computing power scheduling.

[0089] S4.3: Perform a global pheromone update, and then determine and output the termination condition;

[0090] The specific steps for performing a global pheromone update are as follows:

[0091] Perform a global evaporation operation on the pheromone matrix to reduce the weight of historical paths and increase the ability to explore new paths;

[0092] To evaluate the quality of the allocation scheme generated in this iteration, a global cost function, Cost, is defined. If the Cost of this iteration is less than the historical best cost... Then perform the following operations: update the historical best allocation scheme, and increase the pheromone concentration for all paths (i,j) that were successfully selected in this iteration.

[0093] The termination determination and output are as follows:

[0094] Determine if the current iteration number NC has reached the preset maximum iteration number. If yes, end the algorithm process; otherwise, start a new round of search.

[0095] After the algorithm terminates, it outputs the historical best allocation scheme, which is a list of mapping relationships between each task and its target physical node. This mapping list is then sent to the underlying orchestration engine to execute the actual virtual machine creation or container startup operation.

[0096] This embodiment directly embeds the real-time decay coefficient of nodes and the comprehensive energy efficiency score of nodes into the heuristic function of the ant colony algorithm, realizing real-time linkage between scheduling decisions and the physical environment of the data center. This significantly improves the response speed of the improved ant colony algorithm to sudden changes in the physical environment. When the real-time decay coefficient of a node decreases due to aging or local overheating, the heuristic factor will immediately decrease accordingly, thereby preventing ants from choosing that path at the probability level. This avoids the risk of failure without waiting for the pheromone to slowly evaporate. This design effectively overcomes the defect of traditional algorithms that continuously dispatch orders to faulty nodes due to feedback lag. Without changing the main architecture of the ant colony algorithm, it greatly enhances the robustness and physical adaptability of the system with extremely low implementation cost.

[0097] Example 2: This invention also provides an intelligent allocation system for computing resources in a data center for a comprehensive survey of data center resources. The system employs an intelligent allocation method for computing resources in a data center for a comprehensive survey of data center resources as described in Example 1. The system includes:

[0098] The data acquisition module is used to collect multi-dimensional survey data of computer room resources;

[0099] The data preprocessing module is used to perform data preprocessing and normalization operations on the multidimensional census data;

[0100] The node dynamic determination module is used to calculate the node dynamic computing power attenuation coefficient and the node comprehensive energy efficiency score based on the normalized multidimensional census data.

[0101] The computing power allocation module is used to calculate the heuristic function of the improved ant colony algorithm based on the node's comprehensive energy efficiency score and the node's dynamic computing power attenuation coefficient, and to allocate computing power resources according to the improved ant colony algorithm.

[0102] Example 3: The present invention also provides an electronic device, including one or more processors and a memory.

[0103] A processor can be a central processing unit (CPU) or other form of processing unit with data processing and / or instruction execution capabilities, and can control other components in an electronic device to perform desired functions.

[0104] The memory may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor may execute the program instructions to implement the intelligent allocation method for data center computing resources for data center resource surveys described in any embodiment of this application above, and / or other desired functions. Various contents such as initial external parameters and thresholds may also be stored in the computer-readable storage medium.

[0105] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.

Claims

1. A method for intelligent allocation of computing resources in a data center for a comprehensive survey of data center resources, characterized in that, The method includes the following steps: S1: Collect multi-dimensional survey data of computer room resources; S2: Perform data preprocessing and normalization operations on the multidimensional census data; S3: Calculate the node dynamic computing power attenuation coefficient and the node comprehensive energy efficiency score based on the normalized multidimensional census data; S4: Calculate the heuristic function of the improved ant colony algorithm based on the node's comprehensive energy efficiency score and the node's dynamic computing power attenuation coefficient, and allocate computing power resources according to the improved ant colony algorithm; Specifically, S4 consists of: S4.1: Obtaining the set of currently available nodes in the data center and initializing the pheromone matrix; S4.2: Calculating the heuristic function of the improved ant colony algorithm based on the node's comprehensive energy efficiency score and the node's dynamic computing power attenuation coefficient, obtaining the heuristic factor, and performing iterative search and solution construction; S4.3: Performing global pheromone updates, termination determination, and output, thereby obtaining the computing power resource allocation scheme.

2. The intelligent allocation method for computing resources in a data center for a comprehensive survey of data center resources according to claim 1, characterized in that, In S4.2, the expression for the heuristic function of the improved ant colony algorithm is: ； In the formula, Choose the heuristic function value for node j for task i. The overall energy efficiency score for node j. Let ξ be the dynamic computing power attenuation coefficient of node j, and ξ be the adjustment coefficient.

3. The intelligent allocation method for computing power resources in a data center for a comprehensive survey of data center resources, as described in claim 2, is characterized in that... For each task i to be assigned, perform the following sub-steps: The heuristic function of the improved ant colony algorithm is calculated based on the node's comprehensive energy efficiency score and the node's dynamic computing power attenuation coefficient, and the heuristic factor is obtained. Using the improved state transition probability formula, calculate the probability that task i selects each node j in the set; The improved state transition probability formula is as follows: ； Where q is a random number uniformly distributed in the interval [0,1]. For pseudo-random thresholds, The current set of allowed nodes In the middle, making The node that takes the maximum value. For pheromone concentration, Here, α is the heuristic factor, β is the pheromone importance factor, and β is the heuristic function importance factor. Then, for task i, iterate through all optional nodes j and calculate... Find the value of the node with the largest value and denote it as . Then, generate a random number q, if Then task i is directly assigned to node i. ,like Then calculate the probability of all nodes and select the target node by roulette wheel selection; finally, determine the target node for task i. ; Bind task i to the target node. Update the target node The real-time status is calculated, and the corresponding computing power and memory resources are deducted, and the CPU utilization of the target node is estimated and updated. Mark the remaining available capacity of the node in the current iteration; After all m tasks have completed node selection, the current iteration count is [number missing]. .

4. The intelligent allocation method for computing resources in a data center for a comprehensive survey of data center resources, as described in claim 3, is characterized in that... In S4.3, the global pheromone update is specifically performed as follows: Perform a global evaporation operation on the pheromone matrix to reduce the weight of historical paths and increase the ability to explore new paths; To evaluate the quality of the allocation scheme generated in this iteration, a global cost function, Cost, is defined. If the Cost of this iteration is less than the historical best cost... Then perform the following operations: update the historical best allocation scheme, and increase the pheromone concentration for all paths (i,j) that were successfully selected in this iteration.

5. The intelligent allocation method for computing resources in a data center for a comprehensive survey of data center resources according to claim 4, characterized in that, The termination determination and output are as follows: Determine if the current iteration number NC has reached the preset maximum iteration number. If yes, end the algorithm process; otherwise, start a new round of search.

6. The intelligent allocation method for computing resources in a data center for a comprehensive survey of data center resources according to claim 1, characterized in that, In step S4.1, preset ant colony algorithm operating parameters are read, including: the number of ants m, the maximum number of iterations. The parameters include pheromone importance factor α, heuristic function importance factor β, pheromone volatile coefficient ρ, and adjustment coefficient ξ. Simultaneously, the comprehensive energy efficiency score of all nodes in all data centers is obtained. The dynamic computing power attenuation coefficient δ of the nodes is calculated, and then nodes in maintenance or faulty state are excluded to construct the set of allowed nodes. .

7. The intelligent allocation method for computing resources in a data center for a comprehensive survey of data center resources according to claim 6, characterized in that, In S3, the formula for calculating the node dynamic computing power attenuation coefficient δ is as follows: ； In the formula, α is the aging sensitivity coefficient. β represents the service life of the node equipment, and β is the temperature drift sensitivity coefficient. This refers to the microenvironmental temperature of the cabinet where the node is located. The reference temperature for the equipment design.

8. The intelligent allocation method for computing resources in a data center for a comprehensive survey of data center resources, as described in claim 7, is characterized in that... Boundary processing conditions are set to constrain the value of the dynamic computing power attenuation coefficient of the node, specifically: when At this time, the thermal resistance attenuation term will theoretically be less than or equal to 0. Therefore, the dynamic computing power attenuation coefficient δ of the node is forcibly set to 0.

05. At this time, theoretically, the dynamic computing power attenuation coefficient δ of the node is greater than 1. Therefore, the dynamic computing power attenuation coefficient δ of the node is forcibly set to 1.

9. A method for intelligent allocation of computing resources in a data center for data center resource surveying, as described in claim 8, characterized in that, The comprehensive energy efficiency score The calculation formula is: ； In the formula, The comprehensive energy efficiency score is given. The larger the value, the more suitable the node is for receiving new tasks. The unit is a dimensionless composite score value. The nominal computing power of a node, This refers to the CPU utilization of the target node.

10. A smart allocation system for computing resources in a data center for a comprehensive survey of data center resources, characterized in that: The system employs the intelligent allocation method for data center computing resources for data center resource survey as described in any one of claims 1-9. The system comprises: The data acquisition module is used to collect multi-dimensional survey data of computer room resources; The data preprocessing module is used to perform data preprocessing and normalization operations on the multidimensional census data; The node dynamic determination module is used to calculate the node dynamic computing power attenuation coefficient and the node comprehensive energy efficiency score based on the normalized multidimensional census data. The computing power allocation module is used to calculate the heuristic function of the improved ant colony algorithm based on the node's comprehensive energy efficiency score and the node's dynamic computing power attenuation coefficient, and to allocate computing power resources according to the improved ant colony algorithm.

Images