A parameter tuning method, device and related equipment
By processing the performance results of some parameters of the business system through algorithms such as fuzzy logic and Bayesian optimization, the problem of unreliable performance caused by incomplete parameters is solved, and suitable parameter sets are recommended to improve system performance and are applicable to a variety of business systems.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HUAWEI CLOUD COMPUTING TECHNOLOGIES CO LTD
- Filing Date
- 2020-12-30
- Publication Date
- 2026-06-02
AI Technical Summary
In the process of parameter tuning for complex business systems, it is impossible to list all the parameters that affect performance, which leads to unreliable tuning results and failure to meet performance requirements.
A parameter tuning method based on machine learning is adopted. By using fuzzy logic and Bayesian optimization algorithms, the performance results of the business system under some parameters are obtained. Fuzzification is performed to obtain the membership degree, the fuzzy tuning algorithm is executed, and finally defuzzification is performed to obtain the target recommended parameters.
In cases where parameters are incomplete, a parameter set that better meets user needs is recommended, improving the performance of the business system, reducing reliance on experts, saving time, and being applicable to different business systems.
Smart Images
Figure CN114692878B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of machine learning (ML) technology, and in particular to a parameter tuning method, apparatus and related equipment. Background Technology
[0002] Various business systems, such as database systems, content delivery networks (CDNs), and live video streaming systems, typically require specific parameter configurations to function correctly. For example, database systems need to set parameters such as "index block size" and "maximum number of parallel query threads," while CDNs require settings for parameters such as "maximum number of cached group of pictures (GOPs)" and "maximum cache duration for GOPs." Under different parameters, business systems often exhibit different performance characteristics. To achieve better performance for business systems, it is usually necessary to find a suitable set of parameters; this process of finding suitable parameters is called parameter tuning.
[0003] Because a complex business system often has many parameters, in parameter tuning scenarios, it may be impossible to list all the parameters that affect system performance. The tuning process may only adjust a subset of these parameters. In such cases of incomplete parameter settings, the final set of parameters obtained through tuning often fails to meet performance requirements. Summary of the Invention
[0004] This application provides a parameter tuning method, apparatus, and related equipment, which enables parameter tuning even when parameters are incomplete. The final set of parameters provided by the parameter tuning can better meet the user's needs and improve the performance of the business system.
[0005] Firstly, this application provides a parameter tuning method, which includes: obtaining the performance results of a business system running under certain parameters, wherein the certain parameters are a subset of all parameters of the business system; performing fuzzification processing on the performance results according to multiple fuzzy sets to obtain multiple membership degrees, wherein the multiple membership degrees correspond one-to-one with the multiple fuzzy sets; determining multiple sets of fuzzy recommended parameters based on the multiple membership degrees, wherein the multiple membership degrees correspond one-to-one with the multiple sets of fuzzy recommended parameters; and performing defuzzification processing on the multiple sets of fuzzy recommended parameters to obtain target recommended parameters, wherein the target recommended parameters are used to update the aforementioned certain parameters.
[0006] As can be seen, this embodiment first obtains the performance results of the business system running under partial parameters, then fuzzifies the performance results to obtain multiple membership degrees corresponding to multiple fuzzy sets, expressing the degree to which the performance results belong to each fuzzy set, thus solving the problem of unreliable performance results caused by incomplete parameters; then, based on multiple membership degrees, multiple sets of fuzzy recommended parameters are determined; finally, the multiple sets of fuzzy recommended parameters are defuzzified, and the target recommended parameters are obtained by combining them. This embodiment realizes parameter tuning under the condition of incomplete parameters, and the set of parameters finally recommended by the parameter tuning can better improve the performance of the business system.
[0007] Based on the first aspect, in possible embodiments, determining multiple sets of fuzzy recommendation parameters according to multiple membership degrees includes: performing a fuzzy optimization algorithm based on multiple membership degrees to obtain multiple sets of fuzzy recommendation parameters, wherein the fuzzy optimization algorithm includes one or more of Bayesian optimization, genetic algorithm, and fuzzy logic.
[0008] As can be seen, by inputting multiple membership degrees corresponding to the performance results into the fuzzy optimization algorithm, multiple sets of fuzzy recommendation parameters can be obtained. The fuzzy optimization algorithm is a combination of fuzzy logic and Bayesian optimization algorithm, genetic algorithm or other parameter tuning algorithm. This method realizes parameter tuning when parameters are incomplete, and the final recommended set of parameters can well meet the performance requirements of the business system.
[0009] Based on the first aspect, in possible embodiments, the aforementioned performance results include one or more of the following: first-screen time, average response time, and transaction success rate. It should be understood that the performance results are not limited to the above-mentioned types, and embodiments of this application may obtain different performance results depending on the type of business system and different parameter tuning requirements.
[0010] Based on the first aspect, in possible embodiments, before obtaining the performance results of the business system running under certain parameters, the above method further includes: the business system displaying an optimization interface, wherein the optimization interface is used to obtain user-inputted optimization settings information, and the optimization settings information is used to set fuzzification, fuzzy optimization algorithms, and defuzzification. It can be seen that this application embodiment obtains user-inputted optimization settings information by displaying an optimization interface. Users can make corresponding settings on the optimization interface according to specific parameter optimization needs. The optimization interface provides a visual service, enhancing the user experience.
[0011] Based on the first aspect, in possible embodiments, the above-mentioned tuning settings information also includes the identifier of the business system.
[0012] As can be seen, by adding the identifier of the business system to the optimization settings information, it is easy to distinguish different business systems. Then, corresponding parameter optimization services can be provided for different business systems, and parameter combinations that meet different business needs can be recommended.
[0013] Secondly, this application provides a parameter tuning device, comprising: an acquisition module for acquiring performance results of a business system running under certain parameters, wherein the certain parameters are a subset of all parameters of the business system; a fuzzification module for fuzzifying the performance results according to multiple fuzzy sets to obtain multiple membership degrees, wherein the multiple membership degrees correspond one-to-one with the multiple fuzzy sets; a determination module for determining multiple sets of fuzzy recommended parameters according to the multiple membership degrees, wherein the multiple membership degrees correspond one-to-one with the multiple sets of fuzzy recommended parameters; and a defuzzification module for defuzzifying the multiple sets of fuzzy recommended parameters to obtain target recommended parameters, wherein the target recommended parameters are used to update certain parameters.
[0014] Each module of the aforementioned device is specifically used to implement the method described in the first aspect.
[0015] Thirdly, this application provides a computing device including a processor and a memory, which can be interconnected via a bus or integrated together. The processor executes code stored in the memory to implement the method described in the first aspect.
[0016] Fourthly, this application provides a computer-readable storage medium including a program or instructions that, when executed on a computer device, cause the computer device to perform the method described in the first aspect.
[0017] Fifthly, this application provides a computer program product; the computer program product includes program instructions that, when executed by a computing device, perform the methods described in any embodiment of the first aspect. The computer program product may be a software installation package, which can be downloaded and executed on a controller to implement the methods described in any embodiment of the first aspect.
[0018] It is understandable that any of the above aspects can be implemented together with any other aspect or multiple aspects, or can be implemented independently.
[0019] As can be seen, this embodiment of the application obtains the performance results of the business system running under partial parameters, then fuzzifies the performance results to obtain multiple membership degrees corresponding to multiple fuzzy sets, each membership degree expressing the degree to which the performance result belongs to each fuzzy set; then, the above multiple membership degrees are input into a fuzzy tuning algorithm for execution to obtain multiple sets of fuzzy recommended parameters; finally, the multiple sets of fuzzy recommended parameters are defuzzified to obtain the target recommended parameters. This embodiment of the application realizes parameter tuning under incomplete parameter conditions, solving the problem of unreliable system performance results caused by incomplete parameters. The set of parameters finally recommended by parameter tuning can better meet the user's needs and improve the performance of the business system. This parameter tuning method based on machine learning can reduce the dependence on product experts, saving manpower and shortening the time for determining suitable parameters for the business system. Moreover, by adding the identifier of the business system to the tuning settings information, this method can easily distinguish different business systems and can provide parameter tuning services for different business systems, with a wide range of applications. Attached Figure Description
[0020] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below.
[0021] Figure 1 This is a schematic diagram of a CDN system provided in an embodiment of this application;
[0022] Figure 2 This is a schematic diagram of a parameter tuning system provided in an embodiment of this application;
[0023] Figure 3 This is a schematic diagram of another parameter tuning system provided in the embodiments of this application;
[0024] Figure 4 This is a schematic diagram of an optimization interface provided in an embodiment of this application;
[0025] Figure 5 This is a flowchart illustrating a parameter tuning method provided in an embodiment of this application;
[0026] Figure 6 This is a flowchart illustrating another parameter tuning method provided in an embodiment of this application;
[0027] Figure 7 This is a schematic diagram of the structure of a parameter tuning device provided in an embodiment of this application;
[0028] Figure 8 This is a schematic diagram of the structure of a computing device provided in an embodiment of this application. Detailed Implementation
[0029] To facilitate understanding of the technical solutions of the embodiments of this application, some terms involved in this application will be explained below.
[0030] Machine learning refers to the method of using digital computers or computing nodes controlled by digital computers to simulate human learning, discover complex patterns from known data, and use these patterns to predict and determine future events and unknown situations. Depending on the learning method, machine learning can be divided into several types, including supervised learning, unsupervised learning, and reinforcement learning. As a means of realizing artificial intelligence (AI), machine learning has a wide range of applications, including data processing, risk prediction, and facial recognition.
[0031] Fuzzy logic refers to the process of imitating the human brain's judgment and reasoning regarding uncertain concepts. For descriptive systems with unknown or uncertain models, and for controlled objects with strong nonlinearity and large time lag, it applies fuzzy sets and fuzzy rules for reasoning, expressing transitional boundaries or qualitative knowledge and experience, and solving rule-based fuzzy information problems that are difficult to handle using conventional methods. Fuzzy sets are sets used to express fuzzy concepts, also known as fuzzy subsets. Fuzzy logic excels at expressing qualitative knowledge and experience with unclear boundaries. It uses the concept of membership functions to distinguish fuzzy sets, handle fuzzy relationships, and simulate the human brain's rule-based reasoning. Generally, fuzzy logic first converts the input values into membership degrees to various fuzzy sets through fuzzification, then performs reasoning and calculation according to the set fuzzy rules, and finally converts them into output values through defuzzification.
[0032] Content Delivery Network (CDN): A CDN is an intelligent virtual network built on top of existing networks. Key technologies for CDN primarily include content storage and distribution. CDNs rely on edge nodes (servers) deployed in various locations. Through the load balancing, content distribution, and scheduling modules of a central platform, they distribute content from the origin server to the edge nodes closest to the user. Edge nodes, also known as CDN nodes or cache nodes, are a concept introduced to address the complexity of the network structure. They refer to network nodes with fewer intermediate links to the final user, offering better responsiveness and connection speeds. Figure 1 This is a schematic diagram of a CDN system, using only four CDN nodes as an example. It should be understood that in actual applications, this system can include more CDN nodes and servers with other functions; this application does not limit this. For example, by pre-caching the content from a video provider's origin server to a CDN system, terminal devices can directly access the required video content from the nearest available source, effectively distributing traffic. These terminal devices can be smartphones, cell phones, smartwatches, smart tablets, personal digital assistants (PDAs), laptops, etc. Figure 1 Terminal device 1 can directly obtain video content from the nearest CDN node 1 without accessing the origin server, which reduces the pressure on the origin server, reduces network congestion, improves user access speed and hit rate, reduces latency, and enhances user experience.
[0033] Bayesian optimization (BO) is a model-based sequential optimization (i.e., successive evaluation) method that can obtain an approximate optimal solution with minimal evaluation cost. Its optimization process utilizes Bayes' theorem. The Bayesian optimization framework mainly consists of two core components: a probabilistic surrogate model and an acquisition function. The probabilistic surrogate model is used to proxy the evaluation of the costly unknown objective function. Starting from a prior hypothesis, it iteratively increases the amount of information and corrects the prior, resulting in a more accurate surrogate model. The acquisition function is constructed based on the posterior probability distribution. By maximizing the acquisition function, the next most "potential" evaluation point is selected for evaluation, avoiding unnecessary sampling and improving search efficiency.
[0034] Genetic algorithm (GA) refers to a computational model that simulates the biological evolutionary process based on natural selection and genetic mechanisms in Darwin's theory of evolution. It is a method for searching for optimal solutions by simulating the natural evolutionary process. This algorithm uses mathematical methods and computer simulation to transform the problem-solving process into processes similar to the crossover and mutation of chromosomes and genes in biological evolution.
[0035] With the development of internet technology, various business systems are generating an increasing number of parameters. For example, in the HBase database system, users are required to define hundreds of parameters. When multiple parameters need to be considered, manually debugging each parameter individually is cumbersome, and exploring the parameter space requires a large number of system executions. Therefore, manually setting the most suitable set of parameters is clearly no longer applicable. Consequently, machine learning-based parameter tuning techniques are currently commonly used to adaptively find a suitable set of parameters for the business system. Figure 2 This is a schematic diagram of a commonly used parameter tuning system 200, such as... Figure 2As shown, the parameter tuning system 200 includes a business system 210 and a tuning platform 220. The business system 210 includes a running module 211 and a performance monitoring module 212. The running module 211 runs a given set of parameters, and the performance monitoring module 212 monitors the execution of the running module 211, provides the performance results corresponding to these parameters, and then sends the performance results to the tuning platform 220. The tuning platform 220 includes a quantization module 221, a tuning algorithm module 222, and a parameter recommendation module 223. The quantization module 221 quantizes the performance results sent by the business system 210, for example, quantizing 1.01 to 1 to obtain the quantized performance result. The optimization algorithm module 222 executes a parameter optimization algorithm based on the quantized performance result. Each iteration of the parameter optimization algorithm yields a set of recommended parameters, which are then sent to the parameter recommendation module 223. The parameter optimization algorithm can be Bayesian optimization, genetic algorithm, etc., as mentioned above. These algorithms search for the optimal solution of the recommended parameters through multiple iterations. The parameter recommendation module 223 sends the recommended parameters to the business system 210. It should be understood that before the parameter optimization algorithm starts its first iteration, the parameter recommendation module 223 first provides a set of initial parameters to the business system 210. The performance result obtained by the business system 210 under these initial parameters is given to the quantization module 221 for quantization. Then, the quantization module 221 sends the obtained quantized performance result to the optimization algorithm module 222 before starting the first iteration. Each iteration of the parameter optimization algorithm outputs a set of recommended parameters, which are then sent to the business system 211 by the parameter recommendation module 223. Business system 210 and optimization platform 220 have undergone Figure 2 Through multiple rounds of interaction, the optimization platform 220 ultimately provides the business system 210 with a set of optimal recommended parameters.
[0036] Because a complex business system 210 typically has many parameters, in the aforementioned parameter tuning scenario, it's sometimes impossible to list all the parameters affecting the performance of the business system 210. Therefore, during the parameter tuning process, the tuning platform 220 can only adjust a subset of the parameters of the business system 210, resulting in an incomplete adjustment of all parameters. However, Figure 2The machine learning-based parameter tuning technique requires adjusting all parameters affecting the performance of business system 210 to execute the parameter tuning process effectively and ultimately obtain a better set of parameters. This is because, with incomplete parameters, the performance changes of business system 210 may not be determined solely by the parameters adjusted during tuning; other parameters not listed may also influence the results. In other words, incomplete parameters can lead to unreliable system performance results, meaning the final set of parameters obtained through tuning may not meet the performance requirements of business system 210. Therefore, how to better perform parameter tuning when parameters are incomplete, and find a suitable set of parameters for business system 210 to improve its performance, has become an urgent problem to be solved.
[0037] To address the aforementioned problems, this application discloses a parameter tuning method: such as... Figure 3 As shown, based on the tuning settings information obtained by the business system 210, the tuning platform 310 is controlled to iteratively execute parameter tuning based on fuzzy logic, ultimately obtaining a suitable set of parameters. This method enables parameter tuning even with incomplete parameters, solving the problem of unreliable performance results caused by incomplete parameters. The final set of parameters provided by the tuning better meets user needs and improves the performance of the business system 210. This parameter tuning method based on machine learning reduces reliance on product experts, saving manpower and shortening the time to determine suitable parameters. Moreover, different business systems 210 can call the tuning platform 310 for parameter tuning separately, making it more targeted, simple, convenient, fast, and efficient.
[0038] Figure 3 This is a schematic diagram of the parameter tuning system 300 involved in the embodiments of this application, as shown below. Figure 3 As shown, the parameter tuning system 300 includes a business system 210 and a tuning platform 320. The business system 210 and the tuning platform 320 are connected via a network, which can be a wired network, a wireless network, or a combination of both.
[0039] Business system 210, also known as a production system, is used to provide corresponding business services to users. Business system 210 can run under a given set of parameters and provide final performance results. For example, business system 210 can be a video live streaming business system, providing video live streaming services to users; it can be a database system, providing data access services to users; it can be a CDN system, providing acceleration services for scenarios such as downloading, on-demand, and live streaming; or it can be a cloud service platform, providing various intelligent cloud services to users. This application does not specifically limit the type of business system 210. Business system 210 can be deployed on physical servers, such as x86 servers, ARM servers, etc.; it can also be deployed on virtual machines (VMs) implemented using network functions virtualization (NFV) technology on general-purpose physical servers. A virtual machine refers to a complete computer system simulated by software, possessing complete hardware system functions and running in a completely isolated environment, such as a virtual machine in a cloud data center. This application does not specifically limit this type. It should be understood that the business system 210 can be deployed on a single physical server or a single virtual machine, or on a computer cluster; this application does not impose any specific limitations. The internal unit modules of the business system 210 can be divided in various ways; each module can be a software module, a hardware module, or a combination of both; this application does not impose any limitations on this. Figure 3 As shown on the left, the business system 210 may include a running module 211 and a performance monitoring module 212. The running module 211 is used to run a given set of parameters, and the performance monitoring module 212 is used to monitor the execution of the running module 211, collect various performance index data, and provide corresponding performance results to help users understand the problems existing in the operation process and analyze the performance bottlenecks.
[0040] The tuning platform 320 provides parameter tuning services for the business system 210, ultimately providing a set of parameters that meet the performance requirements of the business system 210. It should be understood that the tuning platform 320 can provide a single parameter tuning service for a specific business system 210, or it can be called by multiple business systems 210 to perform parameter tuning for different systems, providing a set of recommended parameters to meet different business needs. The tuning platform 320 can be a hardware device, such as a physical server, or software, such as a virtual machine deployed on a physical server; this application does not impose specific limitations. The internal unit modules of the tuning platform 320 can be divided in various ways; each module can be a software module, a hardware module, or a combination of both; this application does not impose limitations on this. Figure 3As shown on the right, the tuning platform 320 may include a parameter recommendation module 321, a fuzzification module 322, a fuzzy tuning algorithm module 323, and a defuzzification module 324. The parameter recommendation module 321 is used to send a set of recommended parameters to the business system 210; the fuzzification module 322 is used to fuzzify the performance results sent by the business system 210, obtaining multiple membership degrees of multiple fuzzy sets corresponding to the performance results, wherein multiple membership degrees correspond one-to-one with multiple fuzzy sets; the fuzzy tuning algorithm module 323 is used to execute the fuzzy tuning algorithm to obtain multiple sets of fuzzy recommended parameters, wherein the fuzzy tuning algorithm is an algorithm obtained by combining fuzzy logic with some traditional parameter tuning algorithms, including Bayesian optimization, genetic algorithms, etc., as described above. Of course, other parameter tuning algorithms can also be combined with fuzzy logic for the parameter tuning method in this embodiment; the defuzzification module 324 is used to defuzzify the multiple sets of fuzzy recommended parameters and synthesize them to obtain a set of recommended parameters.
[0041] It is worth noting that, Figure 3 This is merely a schematic diagram of a parameter tuning system 300 provided in an embodiment of this application. The positional relationships between the systems and modules shown in the diagram do not constitute any limitation. For example, Figure 3 The business system 210 and the optimization platform 320 in the paper are described using independent devices or server clusters as examples. In specific implementations, the business system 210 and the optimization platform 320 can also be deployed on the same server or server cluster. This application does not make any specific limitations.
[0042] After introducing the parameter tuning system 300 involved in this application, the following will describe in detail the process of controlling the tuning platform 320 to iteratively execute parameter tuning based on fuzzy logic, based on the tuning setting information obtained by the business system 210, and finally obtaining a set of suitable parameters.
[0043] To execute the parameter tuning method provided in this application embodiment, the business system 210 first needs to obtain tuning setting information. This information is used to configure the execution methods for fuzzification, fuzzy tuning algorithms, and defuzzification. The fuzzy tuning algorithms include one or more of Bayesian optimization, genetic algorithms, and fuzzy logic. Other parameter tuning algorithms may also be included for use in the parameter tuning method of this application embodiment. In one possible embodiment, the business system 210 will first display a tuning interface, where technicians can configure the entire parameter tuning process and input the corresponding tuning setting information. Figure 4 This is a schematic diagram of an optimization interface 400 provided in an embodiment of this application, as shown below. Figure 4As shown, the tuning interface 400 includes a parameter setting bar 410, a fuzzy tuning setting bar 420, and a membership function graph 430. The parameter setting bar 410 is used to set the number, name, and corresponding value range of the parameters adjusted during parameter tuning. The fuzzy tuning setting bar 420 is used for fuzzification and defuzzification settings: fuzzification settings select the fuzzification method for the input performance results, including selecting the number, name, value range, and membership function type of the fuzzy sets; defuzzification settings select the defuzzification method for the multiple sets of fuzzy recommendation parameters obtained by executing the fuzzy tuning algorithm. Defuzzification methods include weighted average decision method, maximum mean method, center method, etc., which are not limited in this application. How to fuzzify the performance results and defuzzify multiple sets of fuzzy recommendation parameters will not be detailed here; please refer to the relevant content below. It should be noted that... Figure 4 This is merely an illustrative diagram of an optimization interface 400. The optimization interface 400 may include more or fewer status bars, controls, etc., and the text within the optimization interface 400 may also change accordingly. This application does not impose specific limitations on the content, style, or user operation methods of the optimization interface 400. This application also does not limit the number or type of performance indicators, and the business system 210 may display multiple performance results in real time as needed.
[0044] For example, we need to adjust the parameters on the CDN system to achieve near-instantaneous first-screen loading time on the user side, thus improving the user experience. First-screen loading time here refers to the time from when the user clicks to play to when the image loads. For instance, if the first-screen loading time is 0.1 seconds, the user will hardly notice any system lag; if it's 1.0 second, the user will clearly notice the delay, but their thinking will still be continuous within that time; if it's 10 seconds, the user will lose interest in waiting, resulting in a very poor user experience. Currently, the business system displays as follows... Figure 4 The tuning interface 400 shown allows technicians to input tuning settings. Assume the CDN system has two adjustable parameters: x1 and x2, which are some of the CDN system's parameters. Parameter x1 is the number of GOPs cached, ranging from 1 to 5; parameter x2 is the GOP cache duration, ranging from 1 to 10 seconds. After setting these parameters, the following can be obtained: Figure 4The result is shown in parameter setting column 410. Then, in fuzzy tuning setting column 420, the technicians set the number of fuzzy sets for the first screen time to three, naming them "Excellent," "Medium," and "Poor," representing the three levels of the first screen time. The value ranges for each fuzzy set are [0,2], [0.8,3], and [2, +∞], respectively, in seconds. It can be seen that the value ranges of the three fuzzy sets overlap to express the transitional boundaries between the three levels, providing a basis for fuzzifying the performance results. Next, the membership function type corresponding to each fuzzy set is selected. Figure 4 The top left corner shows three commonly used membership function types: triangular, trapezoidal, and bell-shaped. Other forms are also possible, but this application does not limit their application. Technicians can choose the appropriate type based on parameter tuning requirements. After completing the above settings, you can obtain... Figure 4 The membership function diagram is shown in Figure 430. Technicians also need to set the defuzzification method in the fuzzy optimization settings section 420; here, the weighted average decision method is selected. It should be understood that... Figure 4 The optimization interface 400 is only for the purpose of understanding the embodiments of this application. The optimization interface 400 may also have other forms of expression and user operation methods. This application does not make specific limitations.
[0045] After obtaining the aforementioned tuning settings information, business system 210 sends it to tuning platform 320. Based on the tuning settings information, tuning platform 320 begins iteratively executing parameter tuning based on fuzzy logic. Parameter tuning based on fuzzy logic mainly consists of three steps: fuzzification, execution of the fuzzy tuning algorithm, and defuzzification. These are described below:
[0046] (1) Blurring
[0047] To perform fuzzy logic-based parameter tuning, the tuning platform 320 requires not only the aforementioned tuning settings information but also the performance results of the business system 210 running under a certain set of parameters during each iteration. To obtain these performance results, after receiving the tuning settings information, the tuning platform 320 first sends an initial set of recommended parameters to the business system 210. Specifically, after the business system 210 sends the acquired tuning settings information to the tuning platform 320, the tuning platform 320 sends an initial set of recommended parameters to the business system 210 based on this information. The business system 210's running module 211 then runs under these recommended parameters. The performance monitoring module 212 monitors the running status of the running module 211 in real time and then sends the performance results corresponding to these parameters to the fuzzification module 322 of the tuning platform 320. The fuzzification module 322 fuzzifies the performance results according to multiple fuzzy sets, obtaining multiple membership degrees, each corresponding one-to-one with a fuzzy set.
[0048] Continuing the example above, the CDN system runs under a set of recommended parameters, giving a first-screen time of 1.1 seconds. Inputting 1.1 seconds into the fuzzification module 322, fuzzifying 1.1 seconds according to multiple fuzzy sets yields multiple membership degrees. These membership degrees correspond one-to-one with the multiple fuzzy sets. The membership degree represents the degree to which the precise value of the first-screen time belongs to each fuzzy set. The membership degree ranges from 0 to 1, where 0 indicates no membership in the fuzzy set, 1 indicates full membership, and values between 0 and 1 indicate partial membership in one fuzzy set and partial membership in another. Observation... Figure 4 As shown in Figure 430, the membership function of the first screen time of 1.1s has a membership degree of 0.8 in the "Excellent" fuzzy set, 0.7 in the "Medium" fuzzy set, and 0 in the "Poor" fuzzy set. It is understandable that the performance results of business system 210 are unreliable when parameters are incomplete, because the performance results are affected not only by the adjustable parameters (x1, x2) but also by other parameters not listed. By fuzzifying the performance results using the above method, a certain degree of uncertainty is introduced, expressing the transitional boundaries between the various fuzzy sets. Then, by inputting multiple membership degrees into the fuzzy tuning algorithm, the parameter tuning process can be executed effectively.
[0049] (2) Execute the fuzzy optimization algorithm
[0050] After fuzzification, the performance results yield multiple membership degrees corresponding to multiple fuzzy sets. These membership degrees are then input into the fuzzy optimization algorithm module 323, which executes the fuzzy optimization algorithm to obtain multiple sets of fuzzy recommendation parameters corresponding to the multiple membership degrees.
[0051] Continuing from the previous example, based on the three membership degrees of 0.8, 0.7, and 0, the fuzzy optimization algorithm module 323 executes the fuzzy optimization algorithm. Assuming a fuzzy optimization algorithm combining Bayesian optimization and fuzzy logic is used, since the relationship between parameter combinations and performance results is uncertain, the Bayesian optimization algorithm first initializes a surrogate probability model. This surrogate model is used to fit the unknown objective function; here, a Gaussian process (GP) is used as the surrogate probability model by default. Since the surrogate probability model may not be accurate, BO selects the next most "potential" evaluation point for evaluation by maximizing the acquisition function. An expected improvement (EI) strategy can be used, and this embodiment does not specifically limit the acquisition function. Specifically, multiple membership degrees can be multiplied into the acquisition function as weights. Corresponding to the three fuzzy sets of "excellent," "medium," and "poor," a set of fuzzy recommended parameters (x1, x2) can be obtained. Here, it is assumed that the three recommended parameter sets are (3, 2.5), (4, 10), and (4, 5), respectively.
[0052] (3) Deblurring
[0053] In order to provide only one set of recommended parameters to the business system 210, the optimization platform 320 also needs to deblur the multiple sets of fuzzy recommended parameters. The deblurring work is performed by the deblurring module 324. The deblurring module 324 deblurs the multiple sets of fuzzy recommended parameters, synthesizes them into one set of recommended parameters, and then provides this set of recommended parameters to the parameter recommendation module 321, which then sends this set of recommended parameters to the business system 210.
[0054] Continuing with the previous example, the optimization platform 320 deblurred multiple sets of fuzzy performance results according to the deblurring method set by the technical staff. For example... Figure 4 As shown in the fuzzy tuning settings section 420, the defuzzification method selected is the weighted average decision method. Here, the membership degree of the performance result in the "Excellent", "Medium", and "Poor" fuzzy sets is directly used as the respective weights. Of course, other weight factors can also be specified, which is not limited in this application. As shown in equations (i) and (ii), the three sets of fuzzy recommended parameters (3,2.5), (4,10) and (4,5) are defuzzified to obtain a set of recommended parameters (3.47, 7.33). Then, the defuzzification module 324 gives this set of recommended parameters to the parameter recommendation module 321, which sends (3.47, 7.33) to the business system 210.
[0055]
[0056]
[0057] It should be noted that steps (1) to (3) above are only one iteration of parameter tuning in this embodiment of the application. It should be understood that after the parameter recommendation module 321 sends a new set of recommended parameters to the running module 211 of the business system 210, the running module 211 will replace the original set of recommended parameters with the new set of recommended parameters and run under the new set of recommended parameters. The performance monitoring module 212 will continue to monitor the running status of the running module 211 and obtain the performance results corresponding to the new set of recommended parameters. Then the performance monitoring module 212 sends the performance results to the tuning platform 320. The tuning platform 320 will continue to repeat the operation of steps (1) to (3) based on the performance results corresponding to the new set of recommended parameters, and after obtaining another set of recommended parameters, it will send them to the business system 210 again. That is to say, the tuning platform 320 needs to perform the above-mentioned interaction process with the business system 210 in each iteration of parameter tuning based on fuzzy logic. After multiple rounds of interaction between the business system 210 and the tuning platform 320, the entire parameter tuning process will end after the termination condition is met. In a possible embodiment, upon reaching the termination condition, the tuning platform 320 will use the recommended parameters obtained from the last iteration as the optimal parameters, and then send this set of optimal parameters along with tuning termination information to the business system 210, thus ending the entire parameter tuning process. After receiving the optimal parameters, the business system 210 will terminate its calls to the tuning platform 320 and save the optimal parameters.
[0058] Figure 5 This is a flowchart illustrating a parameter tuning method provided in an embodiment of this application. The method includes the following steps:
[0059] S501: Obtain the performance results of business system 210 under certain parameters.
[0060] The aforementioned partial parameters are a subset of all parameters in business system 210. In cases where it's impossible to list all parameters, these partial parameters represent the adjustable parameters of business system 210 during parameter tuning. The entire parameter tuning process involves finding suitable values for these partial parameters to improve the performance of business system 210. For more information on business system 210, please refer to [link to relevant documentation]. Figure 3 The above descriptions will not be repeated here.
[0061] In one possible embodiment, performance results include one or more of the following: first-screen time, average response time, and transaction success rate. This is an example of performance results; the number and types of performance results provided by the business system 210 are determined by the type of business system 210 and the actual tuning requirements, and are not limited in this application.
[0062] In one possible embodiment, the business system 210 displays an optimization interface, which is used to obtain optimization settings information input by the user. These settings are used to configure fuzzification, fuzzy optimization algorithms, and defuzzification execution methods. It should be understood that the "user" here refers to relevant technical personnel of the business system 210. For details regarding the optimization interface and optimization settings information, please refer to [link to relevant documentation]. Figure 3 And as described above. In a possible embodiment, the tuning interface can also display the performance results of the business system 210 in real time, facilitating the monitoring of performance changes during the parameter tuning process.
[0063] In one possible embodiment, the aforementioned optimization settings information also includes the identifier of the business system 210. It is understood that when the business system 210 first accesses the optimization platform 320, the optimization platform 320 assigns a business system identifier to the business system 210. When the business system 210 accesses the optimization platform 320 again, the optimization settings information will include the identifier of the business system 210, facilitating the optimization platform 320's identification of different business systems for separate parameter optimization.
[0064] S502: Perform fuzzification on the performance results based on multiple fuzzy sets to obtain multiple membership degrees.
[0065] In this context, the aforementioned membership degrees and multiple fuzzy sets correspond one-to-one. The fuzzification method has already been introduced above and will not be repeated here.
[0066] S503: Determine multiple sets of fuzzy recommendation parameters based on multiple membership degrees.
[0067] The multiple membership degrees and the multiple sets of fuzzy recommendation parameters correspond one-to-one.
[0068] In one possible embodiment, based on the aforementioned multiple membership degrees, the optimization platform 320 executes a fuzzy optimization algorithm to obtain multiple sets of fuzzy recommendation parameters. The fuzzy optimization algorithm includes one or more of Bayesian optimization, genetic algorithms, and fuzzy logic. Please refer to the relevant description above for further details on fuzzy optimization algorithms; they will not be repeated here.
[0069] S504: Defuzzify multiple sets of fuzzy recommendation parameters to obtain the target recommendation parameters.
[0070] The target recommended parameters are used to update the aforementioned partial parameters. That is, steps S501-S504 yield a set of target recommended parameters, which are then used to replace the original values of the aforementioned partial parameters in the business system 210. The defuzzing method can be found in the relevant description above and will not be repeated here.
[0071] As can be seen, this embodiment of the application obtains the performance results of the business system 210 running under some parameters, then fuzzifies the performance results to obtain multiple membership degrees of multiple fuzzy sets corresponding to the performance results. Each membership degree expresses the degree to which the performance result belongs to each fuzzy set. Then, the above multiple membership degrees are input into the fuzzy tuning algorithm for execution to obtain multiple sets of fuzzy recommended parameters. Finally, the multiple sets of fuzzy recommended parameters are defuzzified to obtain the target recommended parameters. This embodiment of the application realizes parameter tuning under the condition of incomplete parameters, solves the problem of unreliable system performance results caused by incomplete parameters, and the set of parameters finally recommended by parameter tuning can better meet the user's needs and improve the performance of the business system 210. This parameter tuning method based on machine learning can reduce the dependence on product experts, save manpower and shorten the time to determine suitable parameters for the business system 210. Moreover, by adding the identifier of the business system 210 to the tuning setting information, this method can easily distinguish different business systems 210 and can provide parameter tuning services for different business systems 210 respectively, with a wide range of applications.
[0072] See Figure 6 , Figure 6 This is a flowchart illustrating another parameter tuning method provided in this application embodiment, which may include the following steps:
[0073] S601: Business system 210 obtains optimization setting information and sends the optimization setting information to optimization platform 320.
[0074] The tuning settings are used to configure the execution of fuzzification, fuzzy tuning algorithms, and defuzzification.
[0075] In one possible embodiment, before the business system 210 obtains the tuning settings information, the business system 210 displays a tuning interface 300. The tuning interface 300 is used to obtain the tuning settings information input by the user and also to display performance results. For details regarding the tuning interface 300 and the tuning settings information, please refer to the relevant description above.
[0076] In one possible embodiment, the tuning settings information also includes the identifier of the business system. This identifier distinguishes different business systems 210, enabling the tuning platform 320 to perform parameter tuning for each business system 210 to meet different business requirements.
[0077] S602: Based on the optimization settings information, the optimization platform 320 sends a set of initial recommended parameters to the business system 210.
[0078] Understandably, during the first iteration of parameter tuning, the initial recommended parameters sent by the tuning platform 320 to the business system 210 can be randomly generated based on the tuning settings, or they can be retrieved from a historical recommended parameter table stored by the tuning platform 320 for the business system 210. The recommended parameters are a part of all parameters in the business system 210, representing the adjustable parameters during the parameter tuning process.
[0079] S603: The business system 210 initializes the recommended parameters, obtains the performance results, and sends the performance results to the tuning platform 320.
[0080] Specifically, the business system 210 receives the initial recommended parameters from the tuning platform 320, runs under the initial recommended parameters, monitors performance indicators, provides the performance results corresponding to the initial recommended parameters, and then sends the performance results to the tuning platform 320.
[0081] S604: Based on the performance results, the optimization platform 320 executes a fuzzy optimization algorithm to obtain a set of recommended parameters, and then sends the recommended parameters to the business system 210.
[0082] In one possible embodiment, the tuning platform 320 first fuzzifies the performance results based on multiple fuzzy sets, obtaining multiple membership degrees corresponding to each of the multiple fuzzy sets, wherein each membership degree corresponds one-to-one with a fuzzy set; then, based on the aforementioned multiple membership degrees, a fuzzy tuning algorithm is executed to obtain multiple sets of fuzzy recommendation parameters; finally, the multiple sets of fuzzy recommendation parameters are defuzzified, and a single set of recommendation parameters is obtained. For details on fuzzification, the fuzzy tuning algorithm, and defuzzification, please refer to the relevant descriptions above, which will not be repeated here.
[0083] S605: The business system 210 runs the recommended parameters, updates the performance results, and sends the updated performance results to the tuning platform 320.
[0084] S606: The optimization platform 320 determines whether the termination condition has been met. If the termination condition is met, step S607 is executed; otherwise, step S604 is returned. It is understandable that the optimization platform 320 and the business system 210 continuously perform... Figure 3 The interaction shown will continue until a termination condition is met, at which point the entire parameter tuning process will end. The termination condition can be that the performance result reaches a set value, or it can be that the maximum number of iterations for parameter tuning has been reached; this application does not impose any limitations on this.
[0085] S607: The optimization platform 320 sends the optimization completion information and the optimal parameters to the business system 210.
[0086] Specifically, when the termination condition is met in S606, the optimization platform 320 uses the recommended parameters obtained from the last iteration as the optimal parameters, and then sends the optimal parameters and optimization completion information to the business system 210. In one possible embodiment, after receiving the optimal parameters sent by the optimization platform 320, the business system 210 updates the values of some parameters of the business system 210 with the optimal parameters and ends the call to the optimization platform 320.
[0087] As can be seen, this embodiment of the application, based on the tuning settings information obtained by the business system 210, controls the tuning platform 320 to iteratively execute parameter tuning based on fuzzy logic: the tuning platform 320 fuzzifies the performance results sent by the business system 210, obtaining multiple membership degrees corresponding to multiple fuzzy sets, and then inputs these multiple membership degrees into the fuzzy tuning algorithm for calculation, obtaining multiple sets of fuzzy recommended parameters; finally, through defuzzification and synthesis, a set of recommended parameters is obtained and sent to the business system 210. After multiple interactions between the tuning platform 320 and the business system 210, a suitable set of parameters is finally obtained and set as the optimal parameters. This method realizes parameter tuning even when parameters are incomplete, and can solve the problem of unreliable system performance caused by incomplete parameters. The set of parameters finally given by parameter tuning can better meet the user's needs and improve the performance of the business system 210; this parameter tuning method based on machine learning can reduce the dependence on product experts, save manpower, and shorten the time to determine suitable parameters. Moreover, different business systems 210 can call the tuning platform 320 for parameter tuning respectively, which is more targeted, simple, convenient, fast and efficient.
[0088] In one possible embodiment, the tuning platform 320 can also perform traditional parameter tuning methods. For information on traditional parameter tuning, please refer to [link to relevant documentation]. Figure 2 Related description. In parameter tuning scenarios, if all parameters affecting the performance of the business system can be adjusted, then the performance result is entirely determined by the adjusted parameters. Given reliable performance results, [the system can be optimized / adjusted]. Figure 2 Traditional parameter tuning methods can already recommend a suitable set of parameters, eliminating the need for fuzzification module 322, fuzzy tuning algorithm module 323, and defuzzification module 324 to implement fuzzy logic-based parameter tuning. In other words, the tuning platform 320 in this embodiment can provide two methods, one for cases with incomplete parameters and the other for cases with complete parameters, thus enhancing the tuning capabilities of the tuning platform 320.
[0089] Figure 7 This is a schematic diagram of the structure of a parameter tuning device 700 provided in an embodiment of this application. This device can be the tuning platform 320 mentioned above, such as... Figure 7As shown, the device includes an acquisition module 701, a blurring module 702, a determination module 703, and a deblurring module 704.
[0090] The acquisition module 701 is used to acquire the performance results of the business system running under certain parameters, wherein the certain parameters are a part of all parameters of the business system;
[0091] Fuzzification module 702 is used to fuzzify the performance results according to multiple fuzzy sets to obtain multiple membership degrees, wherein the multiple membership degrees correspond one-to-one with the multiple fuzzy sets;
[0092] The determining module 703 is used to determine multiple sets of fuzzy recommendation parameters based on the multiple membership degrees, wherein the multiple membership degrees and the multiple sets of fuzzy recommendation parameters correspond one-to-one;
[0093] The deblurring module 704 is used to deblur the multiple sets of fuzzy recommendation parameters to obtain target recommendation parameters, wherein the target recommendation parameters are used to update some of the parameters.
[0094] The various modules of the aforementioned parameter tuning device 700 are specifically used to implement... Figure 5 or Figure 6 The functions of the tuning platform 310 or business system 210 in the parameter tuning method embodiment will not be described in detail here for the sake of brevity.
[0095] Figure 8 This is a schematic diagram of the structure of a computing device 800 provided in an embodiment of this application. The computing device 800 may be the parameter tuning device 700 mentioned above. The computing device may be a laptop, tablet computer, or cloud server, etc., and this application does not impose any restrictions.
[0096] The computing device 800 includes a processor 801, a communication interface 802, and a memory 803. The computing device is used to implement the steps in the above-described parameter tuning method embodiments. The processor 801, communication interface 802, and memory 803 can be interconnected via an internal bus 804, or they can communicate via wireless transmission or other means. This embodiment takes the connection via bus 804 as an example. Bus 804 can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. Bus 804 can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 8 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.
[0097] Processor 801 may consist of at least one general-purpose processor, such as a Central Processing Unit (CPU), or a combination of a CPU and hardware chips. The aforementioned hardware chips may be Application-Specific Integrated Circuits (ASICs), Programmable Logic Devices (PLDs), or combinations thereof. The aforementioned PLDs may be Complex Programmable Logic Devices (CPLDs), Field-Programmable Gate Arrays (FPGAs), Generic Array Logic (GALs), or any combination thereof. Processor 801 executes various types of digital storage instructions, such as software or firmware programs stored in memory 803, enabling computing device 800 to provide a variety of services.
[0098] The memory 803 stores program code, which is executed by the processor 801 to perform the processing steps of the parameter tuning method in the above embodiments. The program code may include one or more software modules, which can be... Figure 7 The software modules provided in the embodiments, such as the acquisition module, the fuzzification module, the determination module, and the defuzzification module, can be specifically used to execute... Figure 5 Steps S501 to S504 in the embodiment will not be described again here.
[0099] It should be noted that this embodiment can be implemented using a general-purpose physical server, such as an ARM server or an x86 server, or it can be implemented using a virtual machine based on a general-purpose physical server combined with NFV technology. A virtual machine refers to a complete computer system simulated by software, possessing full hardware system functionality and running in a completely isolated environment. This application does not impose specific limitations on this. It should be understood that... Figure 8 The computing device shown can also be a computer cluster consisting of at least one server, which is not specifically limited in this application.
[0100] Memory 803 may include volatile memory, such as random access memory (RAM); memory 803 may also include non-volatile memory, such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid-state drive (SSD); memory 803 may also include combinations of the above types. Memory 803 may store program code, specifically including code for execution. Figure 5 or Figure 6 The program code for the steps described in the embodiments will not be repeated here.
[0101] The communication interface 802 can be a wired interface (e.g., an Ethernet interface), an internal interface (e.g., a Peripheral Component Interconnect express (PCIe) bus interface), a wired interface (e.g., an Ethernet interface), or a wireless interface (e.g., a cellular network interface or a wireless LAN interface), for communicating with other devices or modules.
[0102] It needs to be explained that, Figure 8 This is merely one possible implementation of an embodiment of this application. In practical applications, the computing device 800 may include more or fewer components, and this is not a limitation. For any content not shown or described in the embodiments of this application, please refer to the foregoing. Figure 5 or Figure 6 The relevant descriptions in the embodiments will not be repeated here.
[0103] This application also provides a computer-readable storage medium storing instructions that, when executed on a processor. Figure 5 The method flow shown is thus implemented.
[0104] This application also provides a computer program product that, when run on a processor, provides a solution for... Figure 5 or Figure 6 The method flow shown is thus implemented.
[0105] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.
[0106] The above-disclosed embodiments are merely preferred embodiments of this application and should not be construed as limiting the scope of this application. Those skilled in the art will understand that all or part of the processes for implementing the above embodiments and equivalent variations made in accordance with the claims of this application are still within the scope of this application.
Claims
1. A parameter tuning method, characterized in that, The method includes: Step 1: Obtain the performance results of the business system running under certain parameters, wherein the certain parameters are a subset of all parameters of the business system; Step 2: Perform fuzzification processing on the performance results according to multiple fuzzy sets to obtain multiple membership degrees, wherein the multiple membership degrees correspond one-to-one with the multiple fuzzy sets; Step 3: Determine multiple sets of fuzzy recommendation parameters based on the multiple membership degrees, wherein the multiple membership degrees and the multiple sets of fuzzy recommendation parameters correspond one-to-one; Step 4: Defuzzify the multiple sets of fuzzy recommendation parameters to obtain the first target recommendation parameters, wherein the first target recommendation parameters are used to update the partial parameters; The performance results of the business system under the first target recommendation parameters are obtained based on the first target recommendation parameters. Steps one to four are repeated based on the performance results of the first target recommendation parameters until a termination condition is met, wherein the termination condition includes the performance result reaching a predetermined value.
2. The method according to claim 1, characterized in that, The step of determining multiple sets of fuzzy recommendation parameters based on the multiple membership degrees includes: Based on the multiple membership degrees, a fuzzy optimization algorithm is executed to obtain multiple sets of fuzzy recommendation parameters. The fuzzy optimization algorithm includes one or more of Bayesian optimization, genetic algorithm, and fuzzy logic.
3. The method according to claim 2, characterized in that, The performance results include one or more of the following: first screen time, average response time, and transaction success rate.
4. The method according to claim 2 or 3, characterized in that, Before obtaining the performance results of the business system running under certain parameters, the method further includes: The business system displays an optimization interface, which is used to obtain optimization settings information input by the user. The optimization settings information is used to set the fuzzification, the fuzz optimization algorithm, and the defuzzification.
5. The method according to claim 4, characterized in that, The optimization settings information also include the identifier of the business system.
6. A parameter tuning device, characterized in that, The device includes: The acquisition module is used to acquire the performance results of the business system running under certain parameters, wherein the certain parameters are a part of all parameters of the business system; The fuzzification module is used to fuzzify the performance results according to multiple fuzzy sets to obtain multiple membership degrees, wherein the multiple membership degrees correspond one-to-one with the multiple fuzzy sets; The determination module is used to determine multiple sets of fuzzy recommendation parameters based on the multiple membership degrees, wherein the multiple membership degrees and the multiple sets of fuzzy recommendation parameters correspond one-to-one; A deblurring module is used to deblur the multiple sets of fuzzy recommendation parameters to obtain target recommendation parameters, wherein the target recommendation parameters are used to update some of the parameters; The iterative module is used to obtain the performance result of the business system running under the first target recommendation parameter based on the first target recommendation parameter, and to instruct the acquisition module, fuzzification module and determination module to execute the corresponding steps again according to the performance result of the first target recommendation parameter until the termination condition is met, wherein the termination condition includes the performance result reaching a predetermined value.
7. The apparatus according to claim 6, characterized in that, The determining module is specifically used for: Based on the multiple membership degrees, a fuzzy optimization algorithm is executed to obtain multiple sets of fuzzy recommendation parameters. The fuzzy optimization algorithm includes one or more of Bayesian optimization, genetic algorithm, and fuzzy logic.
8. The apparatus according to claim 7, characterized in that, The performance results include one or more of the following: first screen time, average response time, and transaction success rate.
9. The apparatus according to claim 7 or 8, characterized in that, The device also includes a display module; The display module is used to display the optimization interface, wherein the optimization interface is used to obtain optimization settings information input by the user, and the optimization settings information is used to set the fuzzification, the fuzz optimization algorithm and the defuzzification.
10. The apparatus according to claim 9, characterized in that, The optimization settings information also include the identifier of the business system.
11. A computer device, characterized in that, Including processor and memory; The memory is used to store computer programs; The processor is configured to execute a computer program stored in the memory, so that the computer device performs the method as described in any one of claims 1-5.
12. A computer-readable storage medium, characterized in that, Includes a program or instructions, which, when executed on a computer device, cause the method as described in any one of claims 1-5 to be performed.