Data center resource management method and device

Abstract

The present invention discloses a data center resource management method and device, which includes: modeling at least one data center and at least one computing service to obtain a data center model for each data center and a computing service model corresponding to each computing service; based on the data center model and the computing service model, collecting computing service information and data center information; based on the data center model, the computing service model and at least one resource optimization target, determining a target optimization function for each resource optimization target; based on the computing service information, the data center information and at least one target constraint condition, iteratively solving the target optimization function to obtain solved target service deployment information; and based on the target service deployment information, performing resource allocation management on the data center. Through the technical solutions of the embodiments of the present invention, data center resource management can be achieved under the consideration of multiple target influences, the rationality of resource allocation can be improved, and the overall performance of the data center can be improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a data center resource management method and device. Background Art

[0002] With the continuous development of data center technology systems, data center resource management has become a key technology to improve data center resource utilization, reduce operating costs, and ensure service quality.

[0003] Currently, traditional data center resource management approaches primarily focus on a single objective, such as minimizing energy consumption or maximizing resource utilization. However, these approaches often overlook the interactions and constraints between multiple objectives, leading to irrational resource allocation and impacting the overall performance of the data center. Summary of the Invention

[0004] The present invention provides a data center resource management method and device to achieve data center resource management under the consideration of multiple objectives, improve the rationality of resource allocation, and improve the overall performance of the data center.

[0005] In a first aspect, an embodiment of the present invention provides a data center resource management method, including:

[0006] Modeling at least one data center and at least one computing service to obtain a data center model corresponding to each data center and a computing service model corresponding to each computing service;

[0007] Based on the data center model and the computing service model, collecting corresponding computing service information and data center information;

[0008] Determining a target optimization function corresponding to each resource optimization objective based on the data center model, the computing service model, and at least one resource optimization objective;

[0009] Iteratively solving the target optimization function based on the computing service information, the data center information, and at least one target constraint condition to obtain solved target service deployment information;

[0010] Based on the target service deployment information, resource allocation management is performed on the data center.

[0011] In a second aspect, an embodiment of the present invention further provides a data center resource management device, including:

[0012] A model building module is used to model at least one data center and at least one computing service to obtain a data center model corresponding to each data center and a computing service model corresponding to each computing service;

[0013] A data collection module, configured to collect corresponding computing service information and data center information based on the data center model and the computing service model;

[0014] An optimization function determination module, configured to determine a target optimization function corresponding to each resource optimization objective based on the data center model, the computing service model, and at least one resource optimization objective;

[0015] a deployment information determination module, configured to iteratively solve the target optimization function based on the computing service information, the data center information, and at least one target constraint condition to obtain solved target service deployment information;

[0016] A resource allocation module is used to manage resource allocation for the data center based on the target service deployment information.

[0017] The technical solution of the embodiment of the present invention, by modeling at least one data center and at least one computing service, obtains a data center model corresponding to each data center and a computing service model corresponding to each computing service, thereby facilitating understanding of the complexity of data centers and computing services and providing a basic framework for subsequent data collection and optimization solutions. Based on the data center model and the computing service model, corresponding computing service information and data center information are collected to obtain the operating status of the data center in real time, providing a basis for subsequent resource allocation. Based on the data center model, the computing service model, and at least one resource optimization objective, the target optimization function corresponding to each resource optimization objective is determined, making the optimization process more flexible and controllable. Based on the computing service information, the data center information, and at least one objective constraint, the target optimization function is iteratively solved to obtain the solved target service deployment information, which can meet resource management requirements in different scenarios and improve resource utilization efficiency. Based on the target service deployment information, resource allocation management is performed on the data center, enabling dynamic adjustment and optimal configuration of resources, thereby improving the operational efficiency and flexibility of the data center. Through systematic data center modeling, state collection, multi-objective optimization solution and resource allocation, efficient and dynamic management of data center resources is achieved, which helps to improve the operating efficiency of the data center, improve the rationality of resource allocation, improve the overall performance of the data center, and accurately match computing services, thereby improving resource utilization and the efficiency and quality of computing service provision.

[0018] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present invention, nor is it intended to limit the scope of the present invention. Other features of the present invention will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0020] Figure 1 This is a flow chart of a data center resource management method provided according to the first embodiment of the present invention;

[0021] Figure 2 This is a flow chart of a data center resource management method provided according to the second embodiment of the present invention;

[0022] Figure 3 It is a structural diagram of a data center resource management device provided according to the third embodiment of the present invention. DETAILED DESCRIPTION

[0023] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described 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 should fall within the scope of protection of the present invention.

[0024] It should be noted that the terms "target", "current", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0025] Example 1

[0026] Figure 1 The first embodiment of the present invention provides a flow chart of a data center resource management method. This embodiment is applicable to the case of resource management in a data center. Figure 1As shown, the method can be executed by a data center resource management device, which can be implemented in the form of hardware and / or software, and can be configured in an electronic device. Figure 1 As shown, the method specifically includes the following steps:

[0027] S110: Model at least one data center and at least one computing service to obtain a data center model corresponding to each data center and a computing service model corresponding to each computing service.

[0028] A data center can refer to a centralized computing facility (such as a computer room), typically containing a large number of servers, storage devices, network equipment, and related software and services for processing, storing, and distributing data. A computing service can refer to a service that utilizes the computing resources provided by a data center to perform specific tasks or provide specific functions. A data center model can refer to an abstract representation of the physical and logical structure of a data center. A computing service model can refer to an abstract representation of computing service functions and resource requirements.

[0029] Specifically, the elements and their relationships in at least one data center are abstracted, an abstract representation of the physical and logical structure of each data center is constructed, and a data center model corresponding to each data center is obtained; at least one computing service is modeled, the computing service functions and resource requirements corresponding to each computing service are abstracted, and a computing service model corresponding to each computing service is obtained, so that the complexity of the data center can be understood through clear model construction, which can provide a basis for subsequent data collection and optimization solutions.

[0030] Exemplarily, S110 may include: modeling at least one data center to obtain a first-type multivariate mathematical expression corresponding to each data center; wherein the dependent variables in the first-type multivariate mathematical expression are the data center's operating performance parameters, and the independent variables are the data center's operating resource parameters; and modeling at least one computing service to obtain a second-type multivariate mathematical expression corresponding to each computing service; wherein the dependent variables in the second-type multivariate mathematical expression are the computing service's operating performance parameters, and the independent variables are the operating resource parameters for providing the computing service. Of course, modeling at least one data center and at least one computing service to obtain the mathematical models corresponding to each data center and each service may also take other forms, such as containing only multiple independent variables or dependent variables, or containing both independent variables and dependent variables, but with the positions of the independent variables and dependent variables different from those in the first-type and second-type multivariate mathematical expressions, or containing both independent variables and dependent variables, but with the contents of the independent variables and dependent variables different from those in the first-type and second-type multivariate mathematical expressions. It is understood that those skilled in the art may establish models from different perspectives to achieve a description of at least one data center and at least one computing service, and this patent is not limiting herein.

[0031] The first type of multivariate mathematical expression can refer to a data center model, which is used to describe the relationship between the operational performance of a data center and its operational resources. The second type of multivariate mathematical expression can refer to a computing service model, which is used to describe the relationship between the operational performance of a computing service and the required operational resources. Operational performance parameters may include any one or more of service latency, service throughput, computing power matching, network bandwidth, computer room energy consumption, and computer room resource utilization. Operational resource parameters may include any one or more of the number, model, and performance of central processing units and / or graphics processing units, and may also include the number of computing service requests.

[0032] Specifically, the operational performance parameters of each data center are defined as dependent variables (such as service throughput, computing power matching, network bandwidth, computer room energy consumption, and computer room resource utilization), and operational resource parameters are defined as independent variables (such as the number, model, and performance of CPUs and / or GPUs). Based on these independent and dependent variables, a first-class multivariate mathematical expression is constructed for each data center to describe the complex relationship between each data center's operational performance parameters and its operational resource parameters. The operational performance parameters of each computing service are defined as dependent variables (such as service latency, service throughput, computing power matching, network bandwidth, computer room energy consumption, and computer room resource utilization), and the operational resource parameters required to provide the computing service are defined as independent variables (such as the number, model, and performance of CPUs and / or GPUs). Based on these independent and dependent variables, a second-class multivariate mathematical expression is constructed for each computing service to describe the complex relationship between each computing service's operational performance parameters and its operational resource parameters. Through precise model construction, data center resources can be more rationally allocated and managed, computing services can be accurately matched, resource waste and over-provisioning can be avoided, and resource utilization, as well as the efficiency and quality of computing service provision, can be improved.

[0033] S120: Based on the data center model and the computing service model, corresponding computing service information and data center information are collected.

[0034] Computing service information refers to various data and metrics related to the current operating status of computing services, as determined based on the computing service model. Data center information refers to various data and metrics related to the current operating status of the data center as a whole and its components, as determined based on the data center model. These data and metrics are used to assess the performance, resource usage, health, and resource requirements of the data center and computing services.

[0035] Specifically, based on the data center model and computing service model, real-time collection of computing service information such as the number of requests for each computing service, service delay, service throughput, service quality evaluation and other parameters, and data center information such as network bandwidth, transmission delay within and / or between computer rooms, computer room power consumption, computing power model and quantity, computer room resource utilization, computing power matching degree and other parameters can ensure the accuracy and real-time nature of data collection, and provide a basis for subsequent reallocation of resources and optimization decisions.

[0036] Exemplarily, S120 may include: determining target information corresponding to each data center and each computing service based on the data center model and the computing service model; obtaining computing service information corresponding to the computing service and current data center information corresponding to the data center based on the target information corresponding to each data center and each computing service.

[0037] The target information may refer to key data and indicators that need to be collected according to the computing service model and the data center model.

[0038] Specifically, based on the first-class multivariate mathematical expressions in the data center model, the key performance parameters and resource usage that need to be monitored and collected for each data center (i.e., target information) are determined. Based on the second-class multivariate mathematical expressions corresponding to the computing service model, the performance indicators and resource consumption that need to be tracked for each computing service (i.e., target information) are determined. Monitoring tools and system logs are used to collect computing service information corresponding to the computing service and data center information corresponding to the data center in real time based on the target information. Real-time monitoring and analysis of resource usage in data centers and computing services accurately reflects the demand for computing services and the idle or overallocated status of data center resources, facilitating subsequent resource optimization and adjustment, improving resource utilization and the efficiency and quality of computing service provision.

[0039] S130. Determine a target optimization function corresponding to each resource optimization target based on a data center model, a computing service model, and at least one resource optimization target.

[0040] Resource optimization goals can refer to the objectives that are desired through the effective management and allocation of resources in data centers and computing services. For example, resource optimization goals can include maximizing resource utilization, minimizing power consumption, minimizing bandwidth, minimizing latency, and maximizing service throughput.

[0041] Specifically, based on the resource optimization goals (such as minimizing energy consumption, maximizing resource utilization, etc.), combined with the data center model and computing service model, a target optimization function corresponding to each resource optimization goal is constructed, so that the optimization direction can be clarified, resource allocation can be made more reasonable and efficient, and different business needs and management goals can be flexibly responded to.

[0042] Exemplarily, S130 may include: obtaining at least one resource optimization target corresponding to at least one data center and at least one computing service; taking each resource optimization target as a dependent variable, and taking the deployment scheme of at least one computing service in at least one data center as an independent variable, and constructing a target optimization function corresponding to each resource optimization target based on the data center model and the computing service model.

[0043] The target optimization function may refer to a function used to describe and quantify how the deployment scheme of computing services in a data center affects the resource optimization target.

[0044] Specifically, at least one resource optimization target corresponding to at least one data center and at least one computing service is obtained, and the key factors affecting each resource optimization target, that is, the deployment scheme of each computing service in each data center, are analyzed. Each resource optimization target is used as a dependent variable (that is, the target value to be optimized), and the deployment scheme of the computing service in the data center is used as an independent variable (that is, an adjustable parameter). Based on the data center model and the computing service model, combined with mathematical and statistical methods, a target optimization function that can reflect the relationship between the resource optimization target and the deployment scheme is constructed. This function is used to quantitatively describe the changes in resource optimization targets under different deployment schemes. By optimizing the deployment scheme of computing services in the data center, resources can be allocated and utilized more reasonably, resource idleness and waste can be reduced, and resource utilization as well as the efficiency and quality of computing service provision can be improved.

[0045] S140: Based on the computing service information, the data center information, and at least one target constraint condition, iteratively solve the target optimization function to obtain the solved target service deployment information.

[0046] Target constraints can refer to a set of restrictions that must be followed in achieving resource optimization goals. These constraints limit the solution space for resource optimization goals, improving the feasibility and effectiveness of resource allocation and optimization results in actual operations. Target service deployment information can refer to specific information about how computing services are deployed in the data center.

[0047] In one embodiment, before iteratively solving the target optimization function, the method further includes obtaining parameter values and / or parameter value functions of different computing services under deployment scenarios in different data centers through stress testing, simulated deployment, actual deployment, etc. Function values of the target optimization functions of different computing services under deployment scenarios in different data centers are then determined based on the parameter values and / or parameter value functions.

[0048] Specifically, based on the computing service information, data center information, and target constraints, an appropriate solution and / or optimization algorithm (such as a genetic algorithm, ant colony algorithm, linear programming, etc.) is used to iteratively solve the target optimization function. During the solution process, the resource allocation plan is continuously adjusted until all constraints are met and an optimal or near-optimal solution is achieved. The solved target service deployment information is then obtained, thereby finding an optimal or near-optimal resource allocation plan, improving resource utilization, reducing energy consumption and costs, and enhancing the efficiency and quality of computing service provision.

[0049] S150: Perform resource allocation management on the data center based on the target service deployment information.

[0050] Specifically, based on the target service deployment information obtained after the solution, specific resource scheduling and configuration operations are implemented to adjust the resource allocation plan for the data center to provide computing services, realize dynamic resource allocation management of the data center, improve the overall performance and service quality of the data center, and at the same time improve the efficiency and quality of computing service provision.

[0051] Exemplarily, S150 may include: allocating, based on the target service deployment information, at least one resource in the corresponding at least one data center to each computing service in the at least one computing service to meet computing service requirements in a next time period.

[0052] Computing service requirements may refer to specific requirements and expectations for computing resources, storage resources, network resources, and other related resources and service capabilities corresponding to a data center.

[0053] Specifically, based on the target service deployment information, the specific computing service requirements for each computing service in the next time period are clarified. Based on the computing service requirements and the data center's resource availability, the corresponding data center resources are allocated to each computing service to meet the computing service requirements in the next time period. Through refined resource allocation management, we can ensure that computing services receive sufficient resources when needed, ensuring the efficiency and quality of computing service provision while avoiding over-allocation and waste of resources.

[0054] It's important to note that after completing resource allocation management for a data center, the next round of model abstractions (data center model and computing service model) can be revised based on the management results. This revision can be performed periodically, for example, every half hour, or in real time. Regular or real-time model revisions can more accurately reflect current data center resource usage and changing business needs. This helps fine-tune resource allocation strategies, reduce resource waste, and improve overall resource utilization efficiency, as well as the efficiency and quality of computing service provision.

[0055] The technical solution of the embodiment of the present invention, by modeling at least one data center and at least one computing service, obtains a data center model corresponding to each data center and a computing service model corresponding to each computing service. This helps understand the complexity of data centers and computing services and provides a basic framework for subsequent data collection and optimization solutions. Based on the data center model and computing service model, corresponding computing service information and data center information are collected to obtain the operating status of the data center in real time, providing a basis for subsequent resource allocation. Based on the data center model, computing service model, and at least one resource optimization goal, the target optimization function corresponding to each resource optimization goal is determined, making the optimization process more flexible and controllable. Based on the computing service information, data center information, and at least one target constraint, the target optimization function is iteratively solved to obtain the solved target service deployment information, which can meet resource management requirements in different scenarios, improve resource utilization efficiency, and the efficiency and quality of computing service provision. Based on the target service deployment information, resource allocation management is performed on the data center, enabling dynamic adjustment and optimal configuration of resources, thereby improving the operational efficiency and flexibility of the data center. Through systematic data center modeling, state collection, multi-objective optimization solution and resource allocation, efficient and dynamic management of data center resources is achieved, which helps to improve the operating efficiency of the data center, improve the rationality of resource allocation, improve the overall performance of the data center, and at the same time improve the efficiency and quality of computing service provision.

[0056] It's important to note that data centers often consist of multiple computer rooms (data centers), which may be located across different regions to meet the access needs of users from all over the world. Some of these locations are designed to improve the local user experience, some to provide backup services, and some to support cluster capacity expansion. Regardless of the service being deployed, when deploying across multiple computer rooms, it's crucial to fully consider potential issues to avoid mistakes.

[0057] Taking bandwidth as an example, connecting two or more data centers in different locations often requires a wide area network (WAN). This requires leasing bandwidth from a carrier and establishing a VPN tunnel between the two data centers. The bandwidth required depends on the actual needs of the data centers and the bandwidth available from the carrier. In principle, traffic forwarding within a data center should be avoided across data centers to reduce the burden of cross-data center traffic and the cost of leasing carrier network bandwidth. When deploying services across data centers, the first consideration is whether the bandwidth at the data center entrances and exits is sufficient. If it is not, the service will be vetoed and cannot be built or deployed.

[0058] Taking security as an example, within the data center, various software and hardware security measures can be deployed to ensure data security. However, when data is exchanged between computer rooms, it is no longer under the control of the computer room. Therefore, the security of the data center computer room is less effective when transmitted over the wide area network (WAN). While some data may be encapsulated by adding an outer IP header, the inner message can still be intercepted and decrypted. Others may be encrypted by security devices in the sending and receiving computer rooms, but these encryption algorithms are not unbreakable. Once cracking methods are mastered, data can still be stolen during transmission. Wide area network protocols inherently have security vulnerabilities. If exploited, data could be leaked, potentially resulting in fatal losses. Therefore, computing data or intermediate process data should be transmitted between computer rooms as much as possible. Data containing personal information or commercial secrets should be avoided from being repeatedly transferred across computer rooms.

[0059] 3. Taking latency as an example, data transmission latency is also a key metric. Intra-data center latency is typically very low, in the microsecond range. However, inter-data center data transmission is affected by inter-data center network latency, resulting in higher latency, typically in the millisecond range. The data transmission speed and latency within an internet data center depend on multiple factors, including network bandwidth, network topology, and the performance of the equipment within the data center. Generally speaking, data transmission speeds within a data center can be very fast, often reaching gigabit or even 10Gbps speeds, with internal network communication latency typically around 300µs (0.3ms). However, inter-data center data transmission is limited by inter-data center network bandwidth and is generally slower than intra-data center transmission speeds. Inter-data center communication latency can be as high as 50,000µs (50ms). Inter-data center latency must be fully considered during service deployment.

[0060] Example 2

[0061] Figure 2 This is a flow chart of a data center resource management method provided in Example 2 of the present invention. This example, based on the previous examples, optimizes the step of "iteratively solving the target optimization function based on computing service information, data center information, and at least one target constraint to obtain solved target resource deployment information." Explanations of terms that are identical or corresponding to those in the previous examples are omitted here.

[0062] See also Figure 2 Another data center resource management method provided in this embodiment specifically includes the following steps:

[0063] S210: Model at least one data center and at least one computing service to obtain a data center model corresponding to each data center and a computing service model corresponding to each computing service.

[0064] S220: Based on the data center model and the computing service model, collect corresponding computing service information and data center information.

[0065] S230: Based on the data center model, the computing service model, and at least one resource optimization goal, determine a target optimization function corresponding to each resource optimization goal.

[0066] S240: Obtain at least one target constraint condition.

[0067] Specifically, target constraints refer to a set of restrictions that must be followed in achieving resource optimization goals. These restrictions limit the solution space for resource optimization goals and improve the feasibility and effectiveness of the optimization results in actual operations.

[0068] It should be noted that the objective constraints can be in the same dimension as the objective optimization function. For example, the power consumption of each data center must not exceed a preset value, while a sub-constraint objective is to minimize the power consumption of the data center (computer room) (the power consumption varies when the computer room is lightly loaded and heavily loaded); the inter-computer room network bandwidth must be limited to less than 100Gb / s, while a sub-constraint objective is to minimize the inter-computer room network bandwidth. In this case, the power consumption not exceeding the constraint value and the bandwidth not exceeding 100G are strong constraints, while the sub-constraint objectives can be weak constraints.

[0069] Exemplarily, S240 may further include: acquiring a constraint level corresponding to at least one target constraint condition.

[0070] Specifically, each objective constraint has a corresponding constraint level, which is used to indicate the importance and priority of the condition in the solution and optimization process.

[0071] S250: Based on the computing service information, the data center information, and at least one target constraint condition, preliminarily solve the target optimization function to obtain at least one piece of original resource deployment information after the preliminarily solution.

[0072] The original resource deployment information may refer to one or more resource allocation schemes obtained after preliminarily solving the target optimization function, which are used to describe how to allocate computing resources and other related resources to different computing services.

[0073] Specifically, based on computing service information, data center information, and target constraints, a preliminary solution to the target optimization function is performed through methods such as linear programming to obtain at least one original resource deployment information, providing a basis for subsequent optimization iterations.

[0074] For example, based on the current computing service requirements and the corresponding data center information of each data center, each service is determined to be deployed in which data center, and even when it is accurate to certain computing resources in a certain data center, the value of each independent variable and the value of each objective optimization function are used to obtain at least one piece of original resource deployment information. For example, at a certain moment, there are four computing requirements, 1234, and four data centers, ABCD. If the original resource deployment information is deployed according to the 1A, 2B, 3C, and 4D schemes, then the values of variables such as power consumption, latency, and bandwidth requirements, and the values of each objective function can be determined; if the original resource deployment information is deployed according to the 1B, 2A, 3D, and 4C schemes, then the values of variables such as power consumption, latency, and bandwidth requirements, and the values of each objective function can also be determined. The solution to each objective function is a potential original solution to a multi-computer room service deployment scheme based on multi-objective constraint programming, that is, potential original resource deployment information. The original solution, that is, original resource deployment information, is then determined by determining whether at least one objective constraint condition is satisfied.

[0075] Exemplarily, S250 may include: performing a preliminary solution to the target optimization function based on computing service information, data center information, at least one target constraint condition and the constraint level corresponding to the target constraint condition, and obtaining at least one original resource deployment information after the preliminary solution.

[0076] Specifically, based on computing service information, data center information, and at least one objective constraint, a preliminary solution is performed for each objective optimization function using methods such as greedy, linear programming, enumeration, and exhaustive search. The resulting raw resource deployment information is then obtained. By considering the objective constraints and their levels, resource deployment plans can be more accurately aligned with actual needs and priorities, improving the rationality of resource deployment.

[0077] Exemplarily, based on computing service information, data center information, at least one target constraint and the constraint level corresponding to the target constraint, a target optimization function is preliminarily solved to obtain at least one original resource deployment information after the preliminary solution, and also includes: if there is no original resource deployment information after the preliminary solution of the target optimization function, or the amount of original resource deployment information is lower than a preset value, then at least one constraint level corresponding to the target constraint is downgraded; based on computing service information, data center information, at least one target constraint and the adjusted target constraint corresponding to the target constraint, the target optimization function is re-preliminarily solved to obtain at least one original resource deployment information after the preliminary solution.

[0078] Specifically, if there is no original resource deployment information, or the amount of original resource deployment information is less than a preset value (e.g., the minimum solution set size requirement is not met), the current constraints are judged to be too strict. The target constraints corresponding to the target optimization function for which no original resource deployment information exists are selected, and the corresponding constraint levels are downgraded. This usually means relaxing the scope of these constraints. Based on the computing service information, data center information, at least one target constraint, and the adjusted target constraint corresponding to the target constraint, the target optimization function is re-preliminarily solved to obtain at least one original resource deployment information after the preliminary solution. The new solution result is checked to confirm whether the original resource deployment information that meets the requirements is obtained. If the requirements are still not met after the re-solution, the constraint level can be further adjusted and the above process can be repeated. Through iterative adjustment and solution, the optimal solution set that meets all conditions is gradually approached. By downgrading the constraint level, the difficulty of solving the problem can be reduced and the success rate of finding a feasible solution can be increased.

[0079] For example, if the original resource deployment information does not exist, the target constraint conditions are adjusted and downgraded, and an attempt is made to re-solve. At this time, the target constraint conditions can be downgraded through preset adjustment rules. This usually occurs when the target optimization function cannot obtain any solution. For example, during the first solution, a possible strong constraint is that all computing service demand 1 is deployed in data center A, which is a strong constraint. At this time, according to the collected data, demand 1 is greater than the service capacity of computer room A. At this time, under this strong constraint condition, any deployment plan cannot meet the target constraint conditions, that is, the target optimization function has no solution. At this time, the strong constraint must be reduced to a weak constraint through constraint adjustment, that is, demand 1 is preferentially deployed in data center A, and demands that exceed the service capacity can be provided across data centers. At this time, there may be an optimal solution.

[0080] S260 : Optimize and iterate at least one original resource deployment information corresponding to the target optimization function based on a multi-objective optimization algorithm to obtain target resource deployment information after optimization and iteration.

[0081] A multi-objective optimization algorithm refers to an algorithm that processes multiple optimization objectives simultaneously. The goal of such an algorithm is to find the best trade-off solution between multiple conflicting or interrelated objectives so that all objectives can be satisfied to some extent, rather than optimizing only one of the objectives.

[0082] Specifically, because resource deployment problems often involve multiple optimization objectives (such as minimizing cost, maximizing performance, and maximizing resource utilization), a multi-objective optimization algorithm is needed to find the optimal solution. The multi-objective optimization algorithm employed may include any one or more heuristic algorithms, such as genetic algorithms, particle swarm optimization algorithms, ant colony algorithms, and greedy algorithms. In one embodiment, when multiple multi-objective optimization algorithms are employed, weights can be assigned to the optimal solutions obtained by each multi-objective optimization algorithm. The computing service deployment schemes of the multiple optimal solutions are then multiplied by the corresponding weights to obtain the final solution. The original resource deployment information obtained through the initial solution is iteratively optimized using the multi-objective optimization algorithm, and the resource allocation scheme is continuously adjusted to approach the optimal solution. During the iterative process, the algorithm gradually improves the resource deployment effect based on the feedback from the objective optimization function and the constraints of the objective constraints, obtaining the target resource deployment information after the optimization iterations. Through optimization iterations, resources can be more rationally allocated and utilized, reducing resource waste and improving overall resource utilization efficiency and the efficiency and quality of computing service provision.

[0083] S270: Based on the target service deployment information, perform resource allocation management on the data center.

[0084] The technical solution of the embodiment of the present invention can ensure that the optimization process is always carried out in the direction of meeting actual needs by obtaining at least one target constraint. Based on computing service information, data center information and at least one target constraint, the target optimization function is preliminarily solved, and at least one original resource deployment information after the preliminary solution is obtained. This can verify the rationality of the optimization function and the effectiveness of the solution algorithm, reduce the risk in the subsequent optimization iteration process, and provide a basis for subsequent multi-objective optimization iterations. Based on the multi-objective optimization algorithm, at least one original resource deployment information corresponding to the target optimization function is optimized and iterated, and the target resource deployment information after the optimization iteration is obtained. This can find a balance point between multiple optimization objectives, thereby allocating resources more reasonably and improving resource utilization. By clarifying multiple target constraints, preliminary solutions and iterations of the multi-objective optimization algorithm, the data center resource deployment problem can be effectively solved, efficient and reasonable allocation of data center resources is achieved, the overall performance and resource utilization of the data center are improved, and the efficiency and quality of computing service provision are improved.

[0085] For example, the aforementioned greedy algorithm (also known as a greedy algorithm) always chooses the best solution at the moment when solving a problem. In other words, it doesn't consider the overall optimal solution, and the algorithm only obtains a local optimal solution. Greedy algorithms don't always yield the optimal solution for all problems; the key lies in the choice of greedy strategy.

[0086] The basic idea of a greedy algorithm is to proceed step by step from a certain initial solution to the problem. Each step ensures that a local optimal solution is achieved, based on a specific optimization metric. Each step considers only one piece of data, which must satisfy the local optimization criteria. If the next piece of data, combined with the partial optimal solution, is no longer a feasible solution, it is not added to the partial solution. The algorithm terminates until all data are exhausted or no more data can be added.

[0087] The greedy algorithm generally proceeds as follows:

[0088] ① Establish a mathematical model to describe the problem.

[0089] ② Divide the problem to be solved into several sub-problems.

[0090] ③Solve each sub-problem and obtain the local optimal solution of the sub-problem.

[0091] ④Synthesize the local optimal solution of the sub-problem into a solution to the original problem.

[0092] Greedy algorithms are a simpler and faster design technique for finding optimal solutions to certain problems. Greedy algorithms are characterized by a step-by-step approach, often selecting the optimal solution based on a specific optimization metric based on the current situation, without considering all possible overall scenarios. This eliminates the time-consuming exhaustive search for the optimal solution. Greedy algorithms use a top-down, iterative approach to make successive greedy choices. Each greedy choice simplifies the problem into smaller subproblems, and through each greedy step, an optimal solution is obtained. While each step ensures a local optimal solution, the resulting global solution may not always be optimal, so greedy algorithms avoid backtracking.

[0093] Linear programming (LP) is an important branch of operations research that has been studied early, developed rapidly, is widely used, and has relatively mature methods. It is a mathematical method that assists people in scientific management and is a mathematical theory and method for studying the extreme value problem of linear objective functions under linear constraints.

[0094] Linear programming is an important branch of operations research, widely used in military operations, economic analysis, business management, and engineering technology. It provides a scientific basis for making optimal decisions based on the rational use of limited human, material, and financial resources.

[0095] Introduction to Linear Programming

[0096] (1) List the constraints and objective function

[0097] (2) Draw the feasible region represented by the constraints

[0098] (3) Find the optimal solution and optimal value of the objective function within the feasible region.

[0099] Example 3

[0100] Figure 3 This is a structural diagram of a data center resource management device provided by the third embodiment of the present invention. Figure 3 As shown, the device includes: a model building module 310, a data collection module 320, an optimization function determination module 330, a deployment information determination module 340 and a resource allocation module 350.

[0101] The model building module 310 is used to model at least one data center and at least one computing service to obtain a data center model corresponding to each data center and a computing service model corresponding to each computing service;

[0102] A data collection module 320 is configured to collect corresponding computing service information and data center information based on the data center model and the computing service model;

[0103] An optimization function determination module 330 is configured to determine a target optimization function corresponding to each resource optimization objective based on the data center model, the computing service model, and at least one resource optimization objective;

[0104] A deployment information determination module 340 is configured to iteratively solve the target optimization function based on the computing service information, the data center information, and at least one target constraint condition to obtain solved target service deployment information;

[0105] The resource allocation module 350 is configured to manage resource allocation for the data center based on the target service deployment information.

[0106] The technical solution of this embodiment, by modeling at least one data center and at least one computing service, obtains a data center model corresponding to each data center and a computing service model corresponding to each computing service, thereby helping to understand the complexity of data centers and computing services and providing a basic framework for subsequent data collection and optimization solutions. Based on the data center model and the computing service model, corresponding computing service information and data center information are collected to obtain the operating status of the data center in real time, providing a basis for subsequent resource allocation. Based on the data center model, the computing service model, and at least one resource optimization goal, the target optimization function corresponding to each resource optimization goal is determined, making the optimization process more flexible and controllable. Based on the computing service information, the data center information, and at least one target constraint, the target optimization function is iteratively solved to obtain the solved target service deployment information, which can meet the resource management requirements in different scenarios and improve resource utilization efficiency. Based on the target service deployment information, resource allocation management is performed on the data center, dynamic adjustment and optimal configuration of resources are achieved, and the operating efficiency and flexibility of the data center are improved. Through systematic data center modeling, state collection, multi-objective optimization solution and resource allocation, efficient and dynamic management of data center resources is achieved, which helps to improve the operating efficiency of the data center, improve the rationality of resource allocation, and improve the overall performance of the data center.

[0107] Optionally, the model building module 310 is specifically used to: model at least one data center to obtain a first type of multivariate mathematical expression corresponding to each data center; wherein the dependent variable in the first type of multivariate mathematical expression is the operating performance parameter of the data center, and the independent variable is the operating resource parameter of the data center; model at least one computing service to obtain a second type of multivariate mathematical expression corresponding to each computing service; wherein the dependent variable in the second type of multivariate mathematical expression is the operating performance parameter of the computing service, and the independent variable is the operating resource parameter of providing the computing service.

[0108] Optionally, the operating performance parameters include: any one or more of service delay, service throughput, computing power matching, network bandwidth, computer room energy consumption and computer room resource utilization; the operating resource parameters include: any one or more of the number, model and performance of the central processing unit and / or graphics processing unit.

[0109] Optionally, the data acquisition module 320 is specifically used to: determine the target information corresponding to each data center and each computing service based on the data center model and the computing service model; and obtain the computing service information corresponding to the computing service and the current data center information corresponding to the data center based on the target information corresponding to each data center and each computing service.

[0110] Optionally, the optimization function determination module 330 is specifically used to: obtain at least one resource optimization target corresponding to the at least one data center and the at least one computing service; take each resource optimization target as a dependent variable, and take the deployment scheme of the at least one computing service in the at least one data center as an independent variable, and construct a target optimization function corresponding to each resource optimization target based on the data center model and the computing service model.

[0111] Optionally, the deployment information determination module 340 includes:

[0112] A constraint condition acquisition unit, configured to acquire at least one target constraint condition;

[0113] an original information acquisition unit, configured to perform a preliminary solution to the target optimization function based on the computing service information, the data center information, and at least one target constraint condition, and obtain at least one original resource deployment information after the preliminary solution;

[0114] The target information acquisition unit is used to optimize and iterate at least one original resource deployment information corresponding to the target optimization function based on a multi-objective optimization algorithm to obtain target resource deployment information after optimization and iteration.

[0115] Optionally, the constraint condition acquisition unit is specifically used to: acquire a constraint level corresponding to at least one of the target constraint conditions.

[0116] Optionally, the original information acquisition unit is specifically used to: perform a preliminary solution to the target optimization function based on the computing service information, the data center information, at least one target constraint condition and the constraint level corresponding to the target constraint condition, and obtain at least one original resource deployment information after the preliminary solution.

[0117] Optionally, the original information acquisition unit is specifically used to: if there is no original resource deployment information after the preliminary solution of the target optimization function, or the amount of the original resource deployment information is lower than a preset value, then downgrade at least one constraint level corresponding to the target constraint condition; based on the computing service information, the data center information, at least one target constraint condition and the adjusted target constraint condition corresponding to the target constraint condition, re-preliminarily solve the target optimization function to obtain at least one original resource deployment information after the preliminary solution.

[0118] Optionally, the resource allocation module 350 is specifically configured to allocate at least one resource in at least one corresponding data center to each computing service in at least one computing service based on the target service deployment information to meet the computing service demand in the next time period.

[0119] The data center resource management device provided in the embodiment of the present invention can execute the data center resource management method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0120] Those skilled in the art will appreciate that the modules or steps of the present invention described above can be implemented using a general-purpose computing device. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Alternatively, they can be implemented using program code executable by a computer device, which can then be stored in a storage device and executed by the computing device. Alternatively, they can be fabricated into separate integrated circuit modules, or multiple modules or steps can be fabricated into a single integrated circuit module. Thus, the present invention is not limited to any specific combination of hardware and software.

[0121] Note that the above are only preferred embodiments of the present invention and the technical principles employed. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and that various obvious changes, readjustments, and substitutions can be made by those skilled in the art without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments and may include many other equivalent embodiments without departing from the concept of the present invention. The scope of the present invention is determined by the scope of the appended claims.

Claims

1. A data center resource management method, characterized in that: include: Modeling at least one data center and at least one computing service to obtain a data center model corresponding to each data center and a computing service model corresponding to each computing service, where the computing service is a service that performs a specific task based on computing resources provided by the data center; Based on the data center model and the computing service model, collecting corresponding computing service information and data center information; Determining a target optimization function corresponding to each resource optimization objective based on the data center model, the computing service model, and at least one resource optimization objective; Iteratively solving the target optimization function based on the computing service information, the data center information, and at least one target constraint condition to obtain solved target service deployment information; Performing resource allocation management on the data center based on the target service deployment information; The step of modeling at least one data center and at least one computing service to obtain a data center model corresponding to each data center and a computing service model corresponding to each computing service includes: Modeling at least one data center to obtain a first-type multivariate mathematical expression corresponding to each data center; wherein the dependent variable in the first-type multivariate mathematical expression is an operating performance parameter of the data center, and the independent variable is an operating resource parameter of the data center; Modeling at least one computing service to obtain a second-type multivariate mathematical expression corresponding to each computing service; wherein the dependent variable in the second-type multivariate mathematical expression is an operating performance parameter of the computing service, and the independent variable is an operating resource parameter for providing the computing service; The determining, based on the data center model, the computing service model, and at least one resource optimization goal, a target optimization function corresponding to each resource optimization goal includes: Obtain at least one resource optimization goal corresponding to the at least one data center and the at least one computing service; Taking each resource optimization objective as a dependent variable and the deployment scheme of the at least one computing service in the at least one data center as an independent variable, constructing a target optimization function corresponding to each resource optimization objective based on the data center model and the computing service model; Before iteratively solving the objective optimization function, the method further includes obtaining various parameter values and / or parameter value functions of different computing services under deployment schemes in different data centers through stress testing, simulated deployment, and actual deployment, so as to determine the function values of various objective optimization functions of different computing services under deployment schemes in different data centers based on the various parameter values and / or parameter value functions; wherein the algorithm adopted for iteratively solving the objective optimization function is a multi-objective optimization algorithm, and the multi-objective optimization algorithm includes multiple heuristic algorithms, and weights are set for each optimal solution obtained by the optimization of the multi-objective optimization algorithm, and then the computing service deployment schemes of the multiple optimal solutions are multiplied by the corresponding weights to obtain a final solution; The performing resource allocation management on the data center based on the target service deployment information includes: Based on the target service deployment information, for each computing service in the at least one computing service, at least one resource in the corresponding at least one data center is allocated to meet the computing service demand in the next time period, wherein the computing service demand refers to the specific requirements and expectations for the computing resources, storage resources, network resources and service capabilities corresponding to the data center.

2. The method according to claim 1, characterized in that The operating performance parameters include: any one or more of service delay, service throughput, computing power matching, network bandwidth, computer room energy consumption and computer room resource utilization; the operating resource parameters include: any one or more of the number, model and performance of the central processing unit and / or graphics processing unit.

3. The method according to claim 1, characterized in that The collecting corresponding computing service information and data center information based on the data center model and the computing service model includes: Determining target information corresponding to each data center and each computing service based on the data center model and the computing service model; Based on the target information corresponding to each data center and each computing service, computing service information corresponding to the computing service and current data center information corresponding to the data center are obtained.

4. The method according to claim 1, wherein The iteratively solving the target optimization function based on the computing service information, the data center information, and at least one target constraint condition to obtain solved target resource deployment information includes: Get at least one target constraint; Based on the computing service information, the data center information, and at least one target constraint condition, preliminarily solving the target optimization function to obtain at least one piece of original resource deployment information after the preliminarily solution; At least one original resource deployment information corresponding to the target optimization function is optimized and iterated based on a multi-objective optimization algorithm to obtain target resource deployment information after optimization and iteration.

5. The method according to claim 4, characterized in that The obtaining of at least one target constraint condition further includes: Obtaining a constraint level corresponding to at least one of the target constraint conditions; The preliminarily solving the target optimization function based on the computing service information, the data center information, and the at least one target constraint condition to obtain at least one piece of original resource deployment information after the preliminarily solution includes: Based on the computing service information, the data center information, at least one target constraint condition and the constraint level corresponding to the target constraint condition, the target optimization function is preliminarily solved to obtain at least one original resource deployment information after the preliminarily solution.

6. The method according to claim 5, characterized in that The performing a preliminary solution on the target optimization function based on the computing service information, the data center information, the at least one target constraint condition, and the constraint level corresponding to the target constraint condition to obtain at least one piece of original resource deployment information after the preliminary solution further includes: If no original resource deployment information exists after the target optimization function is initially solved, or the amount of the original resource deployment information is lower than a preset value, downgrading at least one constraint level corresponding to the target constraint condition; Based on the computing service information, the data center information, at least one target constraint and the adjusted target constraint corresponding to the target constraint, the target optimization function is re-preliminarily solved to obtain at least one original resource deployment information after preliminary solution.

7. A data center resource management device, characterized in that: include: a model building module, configured to model at least one data center and at least one computing service, to obtain a data center model corresponding to each data center and a computing service model corresponding to each computing service, wherein the computing service refers to a service that performs a specific task based on the computing resources provided by the data center; A data collection module, configured to collect corresponding computing service information and data center information based on the data center model and the computing service model; An optimization function determination module, configured to determine a target optimization function corresponding to each resource optimization objective based on the data center model, the computing service model, and at least one resource optimization objective; a deployment information determination module, configured to iteratively solve the target optimization function based on the computing service information, the data center information, and at least one target constraint condition to obtain solved target service deployment information; A resource allocation module, configured to manage resource allocation for the data center based on the target service deployment information; The model building module is specifically configured to: model at least one data center to obtain a first-type multivariate mathematical expression corresponding to each data center; wherein the dependent variable in the first-type multivariate mathematical expression is the operating performance parameter of the data center, and the independent variable is the operating resource parameter of the data center; model at least one computing service to obtain a second-type multivariate mathematical expression corresponding to each computing service; wherein the dependent variable in the second-type multivariate mathematical expression is the operating performance parameter of the computing service, and the independent variable is the operating resource parameter of the computing service; The optimization function determination module is specifically configured to: obtain at least one resource optimization objective corresponding to the at least one data center and the at least one computing service; use each resource optimization objective as a dependent variable and the deployment scheme of the at least one computing service in the at least one data center as an independent variable, and construct a target optimization function corresponding to each resource optimization objective based on the data center model and the computing service model; A parameter value acquisition module is specifically used to: before iteratively solving the target optimization function, obtain various parameter values and / or parameter value functions of different computing services under the deployment schemes of different data centers through stress testing, simulated deployment, actual deployment, etc., so as to determine the function values of various target optimization functions of different computing services under the deployment schemes of different data centers based on the various parameter values and / or parameter value functions; wherein the algorithm adopted for iteratively solving the target optimization function is a multi-objective optimization algorithm, and the multi-objective optimization algorithm includes multiple heuristic algorithms, and weights are set for the optimal solutions obtained by each multi-objective optimization algorithm respectively, and then the computing service deployment schemes of the multiple optimal solutions are multiplied by the corresponding weights to obtain the final solution; The resource allocation module is specifically used to: allocate at least one resource in at least one corresponding data center to each computing service in at least one computing service based on the target service deployment information to meet the computing service demand in the next time period, wherein the computing service demand refers to the specific requirements and expectations for the computing resources, storage resources, network resources and service capabilities corresponding to the data center.

Images