A grayscale release migration method and system for information system

Abstract

The present invention provides a grayscale release migration method and system for an information system, which relates to the field of information system migration. The method controls the grayscale ratio of each module by initializing a grayscale ratio vector, dynamically diverts user requests; monitors the abnormality and slow response of the system with migration based on time slices, and generates a system statistical vector; performs weighted statistics and standardization on the statistical vector using a sliding window to form a system state vector; combines health status information to generate a final system state vector reflecting the real-time operating status of the system; calculates the stability index of each module through a stability discrimination model, dynamically adjusts the grayscale ratio, prioritizes the transition of stable modules, and rolls back the grayscale ratio of unstable modules. The system supports automatic adjustment of grayscale ratios, real-time risk response, and multi-architecture compatibility, significantly reduces the risk of the migration process of the information system, and improves the efficiency and stability of grayscale release.

Description

Technical Field

[0001] The present invention belongs to the field of information system migration and system module code update, and specifically relates to an information system grayscale release migration method and system. Background Art

[0002] System informatization migration is an important task in the field of information technology. It is to upgrade, replace and optimize the existing information system by adopting the updated software and hardware products and technologies to be migrated. Informatization migration requires a smooth transition of the system from the legacy environment to the system to be migrated. However, for critical business systems with huge user volume and business volume, the migration process may bring risks of system unavailability, data loss or impaired user experience due to their requirements for high availability and stability. Therefore, the use of gray release, that is, letting the system to be migrated gradually take over some network requests and conduct small-scale pilots in batches, is a key means to effectively reduce the systemic risks of the migration process.

[0003] Grayscale release is a technical means to achieve a smooth transition in the process of switching between old and new versions of software. The current grayscale release method usually requires application developers and operation and maintenance personnel to specify the release strategy in advance, and monitor the grayscale release process through a monitoring center or manual intervention. It is difficult to dynamically adjust the grayscale strategy according to the real-time operating status of the system to be migrated during the grayscale release process. For example, the grayscale release method provided in the patent document with publication number CN117692327A relies on a fixed grayscale ratio, and cannot be flexibly adjusted according to the actual system performance, health status or abnormal conditions, resulting in the lack of scientificity and real-time setting of the grayscale ratio. The existing grayscale release process usually requires manual planning and management by operation and maintenance personnel, and lacks automated tools to generate dynamic release strategies based on the system operating status. This method that relies on manual operations is not only inefficient, but also prone to improper planning or delays in adjustment due to human factors.

[0004] When the system to be migrated is unstable or in an abnormal state, it is difficult for the existing technology to automatically determine the rollback strength of the grayscale ratio based on the system status and quickly execute the rollback. Response delays may cause the load of abnormal modules to continue to increase, which in turn has a greater impact on user experience and business continuity. The grayscale release strategy is mainly based on limited monitoring data, such as response time or simple abnormal statistics, and fails to fully cover health status indicators, such as database performance, message queue accumulation, middleware load, etc., and cannot effectively identify performance bottlenecks in complex scenarios. A grayscale release method and a grayscale release device are provided in the patent document with publication number CN111782260A. Although the policy matching results are determined according to the key element parameters and the preset grayscale policy library, there are deficiencies in dynamics, degree of automation, immediate response capability, and comprehensiveness of status analysis. Not only does it increase the risk of the grayscale release process, but it also reduces the migration efficiency and system stability. Summary of the invention

[0005] The purpose of the present invention is to propose a grayscale release migration method and system for an information system to solve one or more technical problems existing in the prior art and at least provide a beneficial option or create conditions.

[0006] The present invention provides a grayscale release migration method and system for an information system. The grayscale ratio of each module is controlled by initializing a grayscale ratio vector to dynamically divert user requests; the abnormal and slow response of the system with migration is monitored based on time slices to generate a system statistical vector; the statistical vector is weighted and standardized using a sliding window to form a system state vector; the final system state vector reflecting the real-time operating state of the system is generated in combination with health status information; the stability index of each module is calculated through a stability discrimination model, the grayscale ratio is dynamically adjusted, the stable module is transitioned first, and the grayscale ratio of the unstable module is rolled back. The system supports automatic adjustment of the grayscale ratio, real-time risk response and multi-architecture compatibility, significantly reduces the risk of the information system migration process, and improves the efficiency and stability of grayscale release.

[0007] The present invention provides a method and system for grayscale release migration of an information system.

[0008] In order to achieve the above object, according to one aspect of the present invention, a method for phased release migration of an information system is provided, the method comprising the following steps:

[0009] Initialize the grayscale ratio vector, where the grayscale ratio vector contains several values, each of which corresponds to a module and is used to indicate the grayscale ratio used by the module at the current stage;

[0010] Based on the grayscale ratio vector, each module is updated and the operation data of the system modules are collected in real time, including the number of exceptions and response time;

[0011] Based on the sliding window, the grayscale ratio vector is weighted and counted using the module's operating data to generate the system state vector;

[0012] Obtain the health status information of the information system and combine it with the system status vector to generate the final system status data;

[0013] The final system state data is input into the stability discrimination model to generate the module stability index of each module;

[0014] According to the module stability index, adjust the grayscale ratio of each module to generate the grayscale ratio vector of the next stage;

[0015] If the module stability index is lower than the threshold, the rollback mechanism is triggered for the grayscale ratio vector to restore it to the initial grayscale ratio vector.

[0016] Furthermore, the grayscale ratio vector is a floating point number vector having a length equal to the number of modules in the system, and is used to record the grayscale ratio used by each module in the system during the information system migration process.

[0017] Furthermore, the stability discriminant model refers to a discriminant model based on a logistic regression model, which projects the input system state data into a high-dimensional space through a polynomial feature expansion operation and then inputs the data into the logistic regression model to obtain the confidence that the system module is in a stable state, that is, the system module stability index.

[0018] Furthermore, based on the sliding window, the grayscale ratio vector is weighted statistically analyzed using the module's operating data to generate the system state vector:

[0019] Through the sliding window method, the grayscale proportion vector of each module is counted according to the number of abnormalities and the number of slow responses of each module to obtain the statistical vector of each module: the statistical weight of each module is calculated according to the number of abnormalities and the number of slow responses of each module; the statistical weight of each module is used to process the dimensions corresponding to each module in the grayscale proportion vector to obtain the statistical vector of each module; the statistical vectors of each module are combined to obtain the system state vector;

[0020] The specific value of the weight of the sliding window is equal to the exponential value of the weight growth rate over time, and the initialization value of the weight growth rate is greater than 1. The weight growth rate can be used to preferentially increase the weight of recent data.

[0021] Furthermore, the method of obtaining the health status information of the information system and concatenating it with the system status vector to generate the final system status data is as follows:

[0022] Obtain health status information of the system, where the health status information is a vector that can be spliced ​​into a system status vector, and the health status information includes a numerical value for indicating a database status, a numerical value for indicating a message queue status, a numerical value for indicating a middleware status, a numerical value for indicating a memory status, and a numerical value for indicating the number of active threads. The vector consisting of the numerical value for indicating a database status, the numerical value for indicating a message queue status, the numerical value for indicating a middleware status, the numerical value for indicating a memory status, and the numerical value for indicating the number of active threads is spliced ​​into the system status vector as final system status data.

[0023] Furthermore, the final system state data is input into the stability discrimination model to generate the module stability index of each module as follows:

[0024] The logistic regression model maps the final system state data through polynomial expansion to generate the stability index of each module, wherein the value range of the stability index of each module is [0,1], and the polynomial expansion of the logistic regression model includes a combination of quadratic and cubic features;

[0025] The stability index of the module can be used to determine whether the module is in a stable state.

[0026] Furthermore, according to the module stability index, the grayscale ratio of each module is adjusted to generate the grayscale ratio vector of the next stage, which specifically includes:

[0027] Calculate the difference between the module stability index and the stability decision boundary;

[0028] The sum of the current grayscale ratio, the difference between the module stability index and the set stability decision boundary, and the response index is the adjusted grayscale ratio.

[0029] Among them, the module response index represents the correlation between the current grayscale ratio and the module grayscale baseline.

[0030] Furthermore, the grayscale ratio rollback mechanism includes the following steps: when the module stability index continues to be lower than the decision boundary for more than a preset time threshold, a rollback is triggered.

[0031] The present invention also provides an information system grayscale release migration system, which runs on any computing device of a desktop computer, a laptop computer or a cloud data center, and the computing device includes: a processor, a memory, and a computer program stored in the memory and running on the processor, and the processor implements the steps of the information system grayscale release migration method when executing the computer program, and the system includes a service registration center and an automated grayscale release framework:

[0032] The service registration center can be used to store grayscale ratio vectors and support grayscale publishing in a distributed environment;

[0033] The automated grayscale release framework includes:

[0034] The reverse proxy module can be used to dynamically control the grayscale release of each module according to the grayscale ratio vector and record the operation data of the system module;

[0035] The information cache module can be used to update the code of the cache module in stages over time according to the grayscale ratio;

[0036] The monitoring center can be used to monitor the system operation status, handle abnormal and slow response information, and dynamically calculate and update the grayscale ratio vector;

[0037] The system monitoring module can be used to monitor the running status in real time, push abnormal information to the message queue, and provide a monitoring interface for the health status, including the health status of the database, message queue, and middleware, as well as memory usage and the number of active threads.

[0038] The beneficial effects of the present invention are as follows: the present invention provides an information system grayscale release migration method and system, which collects and counts abnormal statistical information, slow response statistical information, health status information and other indicators of each system module in the system to be migrated and realizes fine-grained monitoring of the system to be migrated by means of sliding window weighted statistical data, provides a variety of reference indicator data, ensures that the grayscale release process of the system to be migrated has a comprehensive perception of the real-time status of the system to be migrated, is conducive to the instant discovery of the unstable or abnormal status of each system module of the system to be migrated and rolls back the grayscale ratio, minimizes the impact of the unstable or abnormal status of the system modules of the system to be migrated on the user experience, and reduces the risk of the grayscale release process to be migrated; and by calculating the grayscale ratio used by each module in the next stage according to the real-time status of each module of the system to be migrated and updating it in real time, directly calculates and provides the module grayscale ratio, and does not require operation and maintenance personnel to independently plan the grayscale release process, solves the problem that the transition and rollback strength of the grayscale release is difficult to confirm and the grayscale release process is insufficiently automated, and improves the efficiency of the grayscale release process to be migrated.

[0039] Traditional grayscale release methods usually use fixed-ratio traffic switching, which cannot be dynamically adjusted according to the operating status of the system module. The lack of refined grayscale ratio control at the module level may lead to the expansion of single-point failures. The module-level grayscale ratio control in the present invention records the grayscale ratio of each module through the grayscale ratio vector to achieve refined grayscale release. The grayscale ratio is dynamically adjusted according to the stability index, the stable module is transitioned first, and the grayscale ratio of the unstable module is rolled back to reduce the switching risk. The gradual transition strategy can avoid the system crash or sharp performance degradation that may be caused by a one-time traffic switch.

[0040] The system status analysis of the prior art is mostly based on static data or global statistics, which makes it difficult to reflect the real-time dynamic changes of the system. The lack of priority analysis for recent status can easily lead to delayed response to sudden problems. The sliding window mechanism of the present invention focuses on recent data and makes the status analysis more in line with the dynamic changes of the system through flexible adjustment of the window size. Weights are given to the status data in the sliding window to highlight the impact of recent data and improve the sensitivity of the analysis. Based on the above mechanism, the system can respond to current environmental changes more quickly and effectively reduce the impact of performance fluctuations on grayscale releases.

[0041] Traditional methods often require analysis of all historical data, which has high computational overhead and slow response speed. In addition, data processing lacks unified standards, and the dimensional differences of different indicators affect the accuracy of analysis. The cache forwarding strategy of the present invention greatly reduces repeated calculations and improves request processing efficiency by caching user IP and module forwarding strategies. Standardize multi-dimensional indicators, eliminate dimensional differences, and improve the consistency of model input and the accuracy of analysis results. The feature expansion of the logistic regression model can be combined with polynomial feature expansion to project the system state into a high-dimensional space, thereby improving the accuracy and robustness of stability analysis.

[0042] In the prior art, the analysis of the system operation status is mainly focused on limited indicators such as abnormal statistics or response time, and fails to fully cover the system health status. It is impossible to effectively identify the operation bottlenecks directly related to hardware resources and software performance. The health status information splicing of the present invention introduces key performance indicators such as database status, message queue health, middleware load, memory usage, and number of active threads into the state vector. Combining health status and operation data, a more comprehensive system status description is generated, covering multi-dimensional operation characteristics, and providing strong support for stability judgment. Through health status information, system performance bottlenecks can be quickly identified, and the bottleneck identification ability is enhanced, so that measures can be taken in advance. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The above and other features of the present invention will become more obvious by describing in detail the embodiments shown in the accompanying drawings. The same reference numerals in the accompanying drawings of the present invention represent the same or similar elements. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other accompanying drawings can be obtained based on these accompanying drawings without creative work. In the accompanying drawings:

[0044] Figure 1 The figure shows a flow chart of a grayscale release migration method of an information system;

[0045] Figure 2 The figure shows a flow chart of grayscale diversion in a grayscale release migration system of an information system. DETAILED DESCRIPTION

[0046] The following will be combined with the embodiments and drawings to clearly and completely describe the concept, specific structure and technical effects of the present invention, so as to fully understand the purpose, scheme and effect of the present invention. It should be noted that the embodiments and features in the embodiments of this application can be combined with each other without conflict.

[0047] In the description of the present invention, "several" means one or more, "more" means more than two, "greater than", "less than", "exceed" etc. are understood as not including the number itself, and "above", "below", "within" etc. are understood as including the number itself. If there is a description of "first" or "second", it is only used for the purpose of distinguishing the technical features, and cannot be understood as indicating or implying the relative importance or implicitly indicating the number of the indicated technical features or implicitly indicating the order of the indicated technical features.

[0048] After long-term use, information systems may have accumulated and emerged operational problems. In order to solve these problems, it is necessary to continuously build and release updated information systems and gradually cover the original information systems. Information systems need to be upgraded and migrated. In the process of information system upgrade and migration, the information system is composed of codes of many different modules, and each module is gradually updated and replaced bit by bit.

[0049] In some embodiments, the initial value of the grayscale ratio vector can be preset to be initialized, and the grayscale ratio vector is defined as a tensor used to represent the grayscale ratio of each module in the system to be migrated, for example, using a neural network framework or numpy to initialize the value. According to the grayscale ratio vector, user requests can also be distributed to the original information system such as the legacy system through forwarding rules. Among them, the response time and abnormal state (including the number of abnormalities and the number of slow responses) of the user request can be recorded in real time. In some embodiments, the grayscale ratio vector of each module is, preferably, the embedding vector of the code of each module is obtained after the code of all modules is vectorized using a vectorization embedding algorithm.

[0050] like Figure 1 The flowchart of the information system grayscale release migration method according to the present invention is shown below. Figure 1 The present invention provides an information system grayscale release migration method and system according to an embodiment of the present invention. The present invention provides an information system grayscale release migration method, and the method specifically includes the following steps:

[0051] Initialize the grayscale ratio vector, where the grayscale ratio vector contains several values, each of which corresponds to a module and is used to indicate the grayscale ratio used by the module at the current stage; based on the grayscale ratio vector, update each module and collect the operation data of the system module in real time, including the number of exceptions and response time;

[0052] Based on the sliding window, the module's operating data is used to perform weighted statistics on the grayscale ratio vector to generate a system state vector; the health status information of the information system is obtained and spliced ​​with the system state vector to generate the final system state data;

[0053] The final system state data is input into the stability discrimination model to generate the module stability index of each module;

[0054] According to the module stability index, the grayscale ratio of each module is adjusted to generate the grayscale ratio vector of the next stage; if the module stability index is lower than the threshold, the rollback mechanism is triggered for the grayscale ratio vector to restore it to the initial grayscale ratio vector.

[0055] In some embodiments provided by the present invention, it specifically includes:

[0056] S100, grayscale ratio vector initialization, grayscale diversion and response timing operations are performed on user requests, wherein the grayscale ratio vector is a floating point vector with a length equal to the number of system modules, and is used to record the grayscale ratio used by each module of the system to be migrated, wherein the system to be migrated refers to a system developed based on the existing system (legacy system) and developed using the technology and equipment to be migrated and having the same functions and modules. The structure of the grayscale ratio vector is as follows:

[0057] ，

[0058] Where n is the number of system modules to be migrated, g j The grayscale ratio used by the jth module in the system to be migrated in the next stage, where j is an integer greater than 0 and less than or equal to n;

[0059] Among them, grayscale diversion refers to forwarding user network requests to the legacy system and the system to be migrated according to the module grayscale ratio of each module in the grayscale ratio vector, and response timing refers to the operation of timing the time taken by the system to be migrated to respond to the user network request, wherein the system module is a number of modules divided according to the interface request path in the system to be migrated.

[0060] S100 provides an implementation method for grayscale ratio vector initialization, grayscale diversion and response timing. The grayscale ratio vector is used to control request diversion and gradually guide users to switch from the legacy system to the system to be migrated to reduce the risk during the switching process. Through response timing, the performance of the system to be migrated is monitored in real time to provide data support for subsequent grayscale ratio adjustment. Using floating-point vectors to record grayscale ratios, the ratio can be adjusted independently for each module to achieve refined grayscale release. Through response timing, problems such as slow response can be captured in a timely manner to avoid large-scale impact of unstable performance on user experience. The goal of grayscale release is to reduce risks. By gradually adjusting the grayscale ratio, the traffic can be smoothly transitioned from the legacy system to the system to be migrated, which can reduce the uncertainty of the system. Real-time response timing provides a feedback loop of the system operation status, enabling the system to adjust the diversion strategy according to real-time data and improve response efficiency.

[0061] Grayscale diversion refers to forwarding network requests for accessing the system module to the legacy system service and the system service to be migrated according to the grayscale ratio of each module, including the following steps:

[0062] S101, query whether there is a corresponding forwarding policy for the user IP address in the user access system module in the cache, wherein the cache is data stored in the memory space or storage space in the computer system, the system module refers to a number of modules divided according to the interface request path in the system to be migrated, and the forwarding policy refers to a key-value pair of a user IP address, the accessed system module and the forwarding information corresponding to the IP address, and the structure of the forwarding policy is as follows:

[0063] ，

[0064] Among them, user is a part of the key in the forwarding policy key-value pair, which stores the user's IPv4 address. The IPv4 address is a digital label used in the fourth version of the Internet Protocol to uniquely identify devices connected to the Internet. It consists of 32 bits (4 bytes) and is usually expressed in dotted decimal format, such as: 192.168.1.1. Module is another part of the key in the forwarding policy key-value pair, which is used to identify the system module. Server is the value in the forwarding policy key-value pair, which records whether the network request from the user's IP address should be forwarded to the system to be migrated or the legacy system. If server is original, it indicates that the network request of the user's IP address needs to be forwarded to the legacy system. If server is new, it indicates that the network request of the user's IP address needs to be forwarded to the system to be migrated. In summary, the forwarding policy key-value pair can be summarized as follows: the network request from the user to access the module module needs to be forwarded to the server;

[0065] S102, if the forwarding policy exists in the cache, adding the value information in the forwarding policy to the request header, wherein the request header refers to a set of key-value pair information sent by the client (such as a browser) to the server in an HTTP request, wherein the HTTP request is a network request based on the Hypertext Transfer Protocol, which is the basis for communication between the client (such as a browser) and the server, allowing the client to request resources on the server, such as web pages, pictures, videos, etc.;

[0066] S103, if there is no forwarding strategy, a random number of 0 to 1 is generated and compared with the module grayscale ratio to obtain the forwarding strategy of the user IP address when accessing the corresponding system module. After obtaining the forwarding strategy used, the forwarding strategy corresponding to the IP address is cached and the key information of the forwarding strategy is added to the request header, wherein the module grayscale ratio is the grayscale ratio used by the corresponding module obtained from the grayscale ratio vector. The forwarding strategy of the IP address when accessing the corresponding system module obtained by comparison refers to determining the user IP address by comparing the size, as shown below:

[0067] The user's IP address is 192.168.1.1, and the module being accessed is the product module. The currently generated random number is 0.3. The grayscale ratio of the module corresponding to the product module is 0.4, which is 0.3<0.4. Therefore, when the user's IP accesses this system module, the user's network request is forwarded to the legacy system. The generated forwarding strategy key-value pair is as follows:

[0068] ，

[0069] This forwarding policy key-value pair indicates that network requests to access the product module from IP address 192.168.1.1 will be forwarded to the legacy system (original);

[0070] S104, starting response timing, forwarding the network request to the corresponding module of the corresponding system (the system to be migrated and the legacy system) according to the forwarding policy information in the request header, wherein the response timing refers to the operation of timing the time taken for the system to be migrated to respond to the user's network request;

[0071] S105, obtain the system response, end the response timing and return the response to the user, and record the slow response situation, where the slow response means that the response time of the system to be migrated to a network request exceeds the threshold defined by the developer. If the slow response threshold defined by the developer is 3 seconds, then when the response time of the system to be migrated to a network request exceeds 3 seconds, the response is regarded as a slow response.

[0072] Embodiments S101-S103 provide a method for caching and dynamically generating forwarding strategies, caching the forwarding strategies corresponding to the user IP, avoiding repeated calculations of diversion decisions, and improving the efficiency of request processing. When there is no cache strategy, the strategy is dynamically generated by random number generation and grayscale ratio comparison to ensure system flexibility. The cache strategy reduces the computational overhead of each request, and significantly improves performance in high-concurrency scenarios. The dynamic generation strategy ensures that the system can quickly adapt to new diversion requirements even after the grayscale ratio is adjusted. The cache reduces repeated operations and reduces computational complexity, while the dynamic generation mechanism enables the system to flexibly respond to changes in grayscale ratios. This combination achieves a balance between efficiency and flexibility.

[0073] S200, counting the number of abnormalities and slow responses obtained by monitoring each module of the system to be migrated within the time slice, forming a system statistical vector with fine granularity of monitoring data, and storing the system statistical vector in a statistical queue, wherein the time slice value is a parameter set and adjusted by the operation and maintenance personnel, and is used to define the time interval for counting the number of abnormalities and slow responses of each module of the system to be migrated, wherein the monitoring abnormalities in the system to be migrated include the developer's custom abnormalities and the system abnormalities of the system to be migrated, including:

[0074] Custom exceptions are exceptions created by developers based on specific application requirements, such as access permission exceptions, resource not found exceptions, data validation exceptions, etc.

[0075] System exceptions refer to exceptions thrown by the operating environment of the system to be migrated, such as null pointer exceptions, database connection exceptions, database query statement syntax exceptions, etc.

[0076] The slow response in the system to be migrated means that the response time of the system to be migrated to a network request exceeds the threshold defined by the developer. If the slow response threshold defined by the developer is 3 seconds, then when the response time of the system to be migrated to a network request exceeds 3 seconds, the response is considered a slow response. The system statistics vector is used to record the number of abnormalities and slow responses of each module of the system to be migrated within the time slice. It is a two-dimensional array of (number of modules of the system to be migrated * number of monitoring indicators), and its structure is as follows:

[0077] ，

[0078] Where n is the number of system modules to be migrated, m is the number of monitoring indicators, and c 1,j Indicates the number of slow responses of the jth module in the system to be migrated in the time slice, where j is an integer greater than 0 and less than n, and c i,j represents the number of occurrences of the i-th type of anomaly in the j-th module of the system to be migrated within the time slice, where i is an integer greater than 1 and less than m;

[0079] The statistics queue is a queue for storing system statistics vectors, and is used to record system statistics vectors of all time slices that have passed during the operation of the system to be migrated.

[0080] S300, extracting the system statistical vector of the sliding window size in the statistical queue, weighting and standardizing the system statistical vector in the sliding window to form a system state vector, wherein the sliding window size is a parameter set and adjusted by the operation and maintenance personnel, which determines how many system state vectors counted in consecutive time slices are considered each time the weighted statistics are performed, the weighted statistics is to multiply the system state vector in the sliding window by the weight and then add them up, and the standardization is to convert the system state vector into a distribution with a mean of 0 and a standard deviation of 1.

[0081] Embodiment S200 provides a method for monitoring data statistics, collects abnormal and slow response data of each module, and provides key input for subsequent system status analysis. Time slice granularity division records data by time slice to ensure the real-time and continuity of data. Time slice granularity makes monitoring data more timely, which helps to find and solve problems in time. Module independence analysis allows the abnormal data of each module to be recorded independently, which is convenient for analyzing the operating status of a single module. The risk of grayscale release mainly comes from the instability of the system. By monitoring abnormalities and slow responses, problem modules can be discovered in time, thereby adjusting the grayscale ratio to avoid complete system failure.

[0082] Embodiment S300 provides a sliding window weighted statistics and standardization method, wherein the sliding window provides the recent change trend of the system state. Weighted statistics give greater weight to recent data to reflect the current state of the system. The sliding window makes the system state analysis more flexible and can adapt to the dynamic changes of traffic and performance. Standardization eliminates the dimensional differences between different indicators and improves the analysis accuracy. The sliding window mechanism is combined with weighted statistics to ensure that the system is sensitive to recent changes while reducing the interference of historical data. Standardization further improves the comparability of data between different modules.

[0083] S400, request to obtain the health status information of the system to be migrated, splice it to the system state vector to obtain the final system state vector, input the system stability judgment model to obtain the system stability index vector, wherein the health status information of the system to be migrated includes but is not limited to the health status information indicators such as the database health status, message queue health status, and middleware health status of the system to be migrated, as well as the memory consumption of the system to be migrated and the interest rate of its memory limit, the number of active threads, etc., wherein splicing to the system state vector to obtain the final system state vector means combining the health status information indicators of the system to be migrated, and directly splicing them to the end of the system state vector to form the final system state vector, and the structure of the final system state vector is as follows:

[0084] ，

[0085] Where n is the number of system modules to be migrated, m is the number of monitoring indicators, g is the number of health indicators of the system to be migrated, and c is the number of monitoring indicators of the system to be migrated. 1,j Indicates the number of slow responses of the jth module in the system to be migrated in the time slice, where j is an integer greater than 0 and less than or equal to n, and c i,j represents the number of occurrences of the i-th type of anomaly in the j-th module of the system to be migrated within the time slice, where i is an integer greater than 1 and less than or equal to m, and h p represents the pth health status information indicator of the system to be migrated, where p is an integer greater than 0 and less than or equal to g;

[0086] The system stability discriminant model refers to a discriminant model based on a logistic regression model. The input system module state vector is projected into a high-dimensional space through a polynomial feature expansion operation and then input into the logistic regression model to obtain the confidence that the system module is in a stable state, that is, the system module stability index. The system module state vector is a number of vectors obtained after the final system vector is separated according to the module. The structure of the system module state vector is as follows:

[0087] ，

[0088] Where m is the number of monitoring indicators, j is the module number of the module state vector, g is the number of health indicators of the system to be migrated, and c is the number of monitoring indicators of the module state vector. 1,j Indicates the number of slow responses of the jth module in the system to be migrated in the time slice, where j is an integer greater than 0 and less than or equal to n, where n is the number of system modules to be migrated, and c i,j represents the number of occurrences of the i-th type of anomaly in the j-th module of the system to be migrated within the time slice, where i is an integer greater than 1 and less than or equal to m, and h p represents the pth health status information indicator of the system to be migrated, where p is an integer greater than 0 and less than or equal to g;

[0089] After inputting all system module state vectors into the system stability judgment model, the system module number and system module stability index are combined into a system stability index vector. The structure of the system stability index vector is as follows:

[0090] ，

[0091] Where n is the number of system modules to be migrated, s jThe system module stability index of the j-th module in the system to be migrated, where j is an integer greater than 0 and less than or equal to n.

[0092] Example S400 provides an implementation method for splicing health status information. The health status information supplements the deficiencies of monitoring data and provides a more comprehensive input for stability analysis. Among them, the key performance indicators covered include health information such as databases and message queues, which directly reflect the operating bottlenecks of the system. After splicing the health status information, the system state vector covers multidimensional data from operating performance to resource usage. The introduction of health status information improves the input quality of the system stability judgment model and makes the grayscale ratio adjustment more scientific. Health status information (such as memory usage and the number of active threads) has a direct impact on system performance. After splicing it to the state vector, the stability of the module can be predicted more accurately, and the grayscale ratio can be adjusted accordingly.

[0093] S500, calculating the grayscale ratio vector of the next stage through the system stability index vector and the current grayscale ratio vector, and updating the grayscale ratio of each module of the system to be migrated, wherein the structure of the grayscale ratio vector is as follows:

[0094] ，

[0095] Where n is the number of system modules to be migrated, g j The grayscale ratio used by the j-th module in the system to be migrated in the next stage, where j is an integer greater than 0 and less than or equal to n.

[0096] Example S500 provides an implementation method for stability determination and dynamic adjustment of grayscale ratio, wherein stability assessment is the core basis for grayscale adjustment, and the module stability index is calculated through a logistic regression model to provide decision support for grayscale ratio adjustment. According to the stability index, the grayscale ratio is dynamically adjusted to avoid the risks caused by premature or late switching. The module stability index directly affects the grayscale ratio adjustment speed, ensuring that the high-stability module is transitioned first. The dynamic adjustment mechanism can quickly respond to unstable states, roll back the grayscale ratio in time, and ensure system operation. The stability of the module determines its ability to handle traffic. Dynamic adjustment of the grayscale ratio is based on stability, which not only avoids excessive load on unstable modules, but also makes full use of the resources of stable modules.

[0097] Furthermore, the grayscale ratio vector is a floating point number vector having a length equal to the number of modules in the system, and is used to record the grayscale ratio used by each module in the system during the information system migration process.

[0098] Furthermore, the stability discriminant model refers to a discriminant model based on a logistic regression model, which projects the input system state data into a high-dimensional space through a polynomial feature expansion operation and then inputs the data into the logistic regression model to obtain the confidence that the system module is in a stable state, that is, the system module stability index.

[0099] Furthermore, based on the sliding window, the grayscale ratio vector is weighted statistically analyzed using the module's operating data to generate the system state vector:

[0100] Through the sliding window method, the grayscale proportion vector of each module is counted according to the number of abnormalities and the number of slow responses of each module to obtain the statistical vector of each module: the statistical weight of each module is calculated according to the number of abnormalities and the number of slow responses of each module; the statistical weight of each module is used to process the dimensions corresponding to each module in the grayscale proportion vector to obtain the statistical vector of each module; the statistical vectors of each module are combined to obtain the system state vector;

[0101] The specific value of the weight of the sliding window is equal to the exponential value of the weight growth rate over time, and the initialization value of the weight growth rate is greater than 1. The weight growth rate can be used to preferentially increase the weight of recent data.

[0102] In some embodiments, extracting the system statistical vector of the sliding window size, weighted summing and normalizing the system statistical vector in the sliding window to form the system state vector includes:

[0103] Extract the system statistical vector of the sliding window size in the statistical queue, and form the system state vector v by weighted statistics of the system statistical vector in the sliding window according to the following formula m :

[0104] ,

[0105] Where n is the size of the sliding window, v i is the i-th system statistics vector in the sliding window, w i For v i The statistical weight used represents the weight used for the system statistical vector at the i-th position in the sliding window when performing weighted statistics on the system statistical vector in the sliding window, w i The calculation is done by the following formula:

[0106] ,

[0107] where w min is the minimum statistical weight, which indicates the minimum weight used when performing weighted statistics on the system statistical vector in the sliding window, linspace(0,1,n) irepresents the i-th element in the linear space from 0 to 1, r is the weight growth rate, which indicates the importance of later system state vectors in the weighted statistical process. The larger the weight growth rate r is, the later system statistical vectors in the sliding window will be given a greater weight, while the earlier system statistical vectors will be given a smaller weight accordingly, so that the statistical results can more sensitively reflect the recent changes in the system state.

[0108] The system operating environment is dynamic, and the sliding window can centrally analyze the status data within a specific time range to avoid excessive dispersion of information in global statistics. The status changes within the sliding window can reflect recent trends and provide a higher guarantee for the real-time performance of the system status.

[0109] The sliding window limits the scope of the analyzed data, which helps to reduce the computing cost, especially for long-running systems, avoiding the storage and computing burden brought by the full amount of data. Recent data is more representative of the current state of the system. By giving greater weight to the recent state through weighted statistics, the operating trend of the current system can be captured more accurately. Weighted statistics integrate the data at different time points within the window, which can avoid the impact of fluctuations at a single time point on the overall state assessment.

[0110] The value ranges and units of different system modules and indicators may be different. Standardization can unify the dimensions and facilitate subsequent calculations and comparisons. The standardized state data is evenly distributed and has a moderate value range, which helps to improve the discrimination ability of subsequent models (such as logistic regression models).

[0111] The sliding window allows the system status analysis to focus on recent changes and respond more sensitively to current environmental changes. By limiting the window size, only data within a fixed range is analyzed, reducing the amount of data and computational complexity. The weight increases over time, and the priority focus on recent status data can better capture the dynamic changes of the system. Smaller weights are given to early data to reduce the impact of early anomalies on current analysis.

[0112] Standardization improves data consistency, unifies indicators of different dimensions into standard ranges, and improves the generalization ability of analytical models. The standardized data distribution is more suitable for model input and reduces the interference of extreme values.

[0113] The range of the sliding window limits the timeliness of the analysis and can reflect the recent state changes of the system more quickly. Limiting the window size reduces storage and computing overhead, and is more suitable for scenarios with higher real-time requirements than global statistics. Using the weight-increasing method to assign a larger weight to recent state data can more accurately capture the current state. By controlling the weight growth rate r, the sensitivity of the system to recent data can be flexibly adjusted. The weight reduction has a reduced impact on early data, making the system more tolerant to historical fluctuations. The numerical ranges of different indicators vary greatly, and direct calculation may cause some indicators to dominate. Standardization eliminates this effect. Standardized data is more in line with the input requirements of common models (such as logistic regression), improving stability and convergence speed.

[0114] Furthermore, the method of obtaining the health status information of the information system and concatenating it with the system status vector to generate the final system status data is as follows:

[0115] Obtain health status information of the system, where the health status information is a vector that can be spliced ​​into a system status vector, and the health status information includes a numerical value for indicating a database status, a numerical value for indicating a message queue status, a numerical value for indicating a middleware status, a numerical value for indicating a memory status, and a numerical value for indicating the number of active threads. The vector consisting of the numerical value for indicating a database status, the numerical value for indicating a message queue status, the numerical value for indicating a middleware status, the numerical value for indicating a memory status, and the numerical value for indicating the number of active threads is spliced ​​into the system status vector as final system status data.

[0116] In some embodiments, the values ​​corresponding to each module for indicating the database status may include the occupancy ratio of the connection pool corresponding to each module and the occupancy ratio of the query response time; the values ​​corresponding to each module for indicating the message queue status may include the ratio of the message accumulation amount corresponding to each module to its total accumulation capacity and the ratio of the processing delay to the maximum delay;

[0117] The values ​​corresponding to each module used to represent the middleware status, including the thread load ratio and resource utilization rate corresponding to each module;

[0118] Numeric values ​​corresponding to each module that represent the memory status, including the ratio of memory usage to memory limit (for example, the ratio of memory consumption to memory limit);

[0119] A numerical value indicating the number of active threads, including the ratio of active threads to the total number of threads;

[0120] The health status information is spliced ​​with the system status data generated based on the sliding window to form the final system status data.

[0121] Furthermore, the final system state data is input into the stability discrimination model to generate the module stability index of each module as follows:

[0122] The logistic regression model maps the final system state data through polynomial expansion to generate the stability index of each module, wherein the value range of the stability index of each module is [0,1], and the polynomial expansion of the logistic regression model includes a combination of quadratic and cubic features;

[0123] The stability index of the module can be used to determine whether the module is in a stable state.

[0124] Furthermore, according to the module stability index, the grayscale ratio of each module is adjusted to generate the grayscale ratio vector of the next stage, which specifically includes:

[0125] Calculate the difference between the module stability index and the stability decision boundary (for example, the stability decision boundary can be the average value of the grayscale ratio of each module in the current stage);

[0126] The current grayscale ratio is added to the difference between the module stability index and the set stability decision boundary and multiplied by the response index. The sum is the adjusted grayscale ratio.

[0127] The module response index is a floating point number greater than or equal to 1, indicating the correlation degree (eg, ratio) between the current grayscale ratio and the module grayscale baseline.

[0128] In some embodiments, when the module stability index is lower than the set stability decision boundary d, the grayscale ratio vector is dynamically adjusted. The grayscale ratio adjustment method is to calculate the difference C_d between the module stability index and the set stability decision boundary, where G is the current grayscale ratio, k is the response index, and G' is the adjusted grayscale ratio. Preferably, the current grayscale ratio G is added to the product of C_d multiplied by the response index k, and the sum obtained is the adjusted grayscale ratio G'.

[0129] Preferably, if the module is unstable, a grayscale ratio rollback mechanism is triggered to restore the grayscale ratio to the baseline value b.

[0130] k is the module response index, which is a floating point number greater than or equal to 1, indicating the correlation (e.g., ratio) between the current grayscale ratio G and the module grayscale baseline (the minimum grayscale ratio of each module). The initial value of the module grayscale ratio is calculated based on the grayscale ratio vector of each module, and is used to indicate the ratio of the code of each module that needs to be replaced at the current moment of information migration, wherein the specific value of the module grayscale ratio can be changed later.

[0131] Furthermore, the grayscale ratio rollback mechanism includes the following steps: when the module stability index continues to be lower than the decision boundary for more than a preset time threshold, a rollback is triggered.

[0132] In some embodiments, the step of calculating the grayscale ratio vector of the next stage by using the system stability index vector and the current grayscale ratio vector, and updating the grayscale ratio of each module of the system to be migrated, includes:

[0133] Separate the system stability index vector to obtain the system module stability index corresponding to each module of the system to be migrated, separate the grayscale ratio vector to obtain the current grayscale ratio g of each module of the system to be migrated;

[0134] The grayscale ratio used in the next stage of the module is calculated by the following formula:

[0135]

[0136]

[0137]

[0138]

[0139] Among them, min function and max function represent taking the minimum value and taking the maximum value respectively, for is used to indicate the use condition of the formula, g is the current grayscale ratio of the module, which is a floating point number from 0 to 1, indicating the grayscale ratio the module is using, s is the system module stability index, which is a floating point number from 0 to 1, indicating the confidence that the system module is in a stable state, calculated by the system stability discrimination model, d is the module stability decision boundary, which is a floating point number from 0 to 1, indicating the threshold for judging whether the module is in a stable state, that is, when the module grayscale stability index s is greater than the module stability decision boundary, the module is considered to be in a stable state and the grayscale ratio can be transitioned forward, otherwise, the module is considered to be in an unstable state and the grayscale ratio should be rolled back. This parameter is set and adjusted by the operation and maintenance personnel, b is the module grayscale baseline, which is a floating point number from 0 to 1 and less than or equal to the current grayscale ratio g of the module, indicating the starting grayscale ratio of the module's grayscale ratio transition process, affecting the speed of the grayscale ratio transition, and the current grayscale ratio g of the module is The closer to the module grayscale baseline b, the slower the grayscale ratio transition speed will be used. Conversely, the farther the module current grayscale ratio g is from the module grayscale baseline b, the faster the grayscale ratio transition speed will be used. k is the module response index, which is a floating point number greater than or equal to 1, indicating the correlation between the module grayscale ratio transition speed and the difference between the module current grayscale ratio g and the module grayscale baseline b. The larger the module response index k is, the slower the transition speed is used for the early stage of the grayscale ratio transition process, and the faster the transition speed is used for the late stage of the grayscale ratio transition. This parameter is set and adjusted by the operation and maintenance personnel. l is the upper limit of the module grayscale adjustment, which is a floating point number from 0 to 1, indicating the upper limit of each adjustment of the module grayscale ratio. This parameter is set and adjusted by the operation and maintenance personnel. p is the upper limit of the module baseline adjustment, which is a floating point number close to 0. It is used to adjust the upper limit of the initial speed adjustment of the grayscale ratio process after the grayscale ratio is rolled back. g' is the module grayscale ratio used in the next stage of the module, and b' is the module grayscale baseline used in the next stage of the module.

[0140] To facilitate storage management, the grayscale ratios of each module in the next stage are combined to form a grayscale ratio vector used by the system in the next stage. The grayscale ratio vector is a floating-point vector with a length equal to the number of system modules. It is used to record the grayscale ratios used by each module of the system to be migrated. The structure of the grayscale ratio vector is as follows:

[0141] ，

[0142] Where n is the number of system modules to be migrated, g j The grayscale ratio used by the j-th module in the system to be migrated in the next stage, where j is an integer greater than 0 and less than or equal to n.

[0143] In order to achieve the above purpose, an embodiment of the present invention further provides a low-risk grayscale release system to be migrated, including:

[0144] The service registration center is used to automatically register the services of each module in the grayscale release system to be migrated, and to register the system to be migrated and the legacy system to the service registration center through manual registration. It also provides a global configuration function for storing grayscale ratio vectors, so that the grayscale release process can be run in a distributed environment, ensuring the availability of the grayscale release system to be migrated, and providing business support for the grayscale release of the system to be migrated in microservice architecture and monolithic architecture. Microservice architecture refers to an architecture in which an application is developed as a set of small services and deployed separately. Each service implements a specific business function and can be deployed independently. Monolithic architecture refers to an application that is built as a single, independent unit. In this architecture, all functions are tightly coupled and deployed in the same service.

[0145] The automated grayscale release framework consists of a reverse proxy module, an information cache module, and a monitoring center, including:

[0146] The reverse proxy module is used to grayscale divert user requests and provide response timing and slow response information submission functions. It dynamically reads the global configuration in the service registration center to obtain the module grayscale ratio to achieve the transition and rollback of the module grayscale ratio, and then realizes grayscale release. Among them, grayscale diversion refers to forwarding user network requests to the legacy system and the system to be migrated according to the module grayscale ratio of each module in the grayscale ratio vector. Response timing refers to the operation of timing the time it takes for the system to be migrated to respond to the user's network request. Slow response information submission refers to the reverse proxy module submitting the module information of the system module with slow response to the message queue;

[0147] Information cache module, used to cache the forwarding strategy corresponding to the user IP;

[0148] The monitoring center is used to monitor the status of the system to be migrated, process the abnormal information submitted by the system to be migrated and the slow response information submitted by the reverse proxy module, calculate the module grayscale ratio used by each system module in the next stage, and update the calculated grayscale ratio vector to the global configuration of the service registration center, thereby realizing the automatic calculation of the grayscale ratio;

[0149] The system monitoring module is embedded in the system to be migrated and is used to monitor the running status of the system to be migrated, push the specified exception information to the message queue, and provide the system health status monitoring interface to the monitoring center, where the exception information contains the developer's custom exception and related information of the system exception of the system to be migrated, including but not limited to the name of the exception, the parameters of the exception, the call stack information of the exception, etc. The health status monitoring interface is a network interface that can provide health information of the system to be migrated to the outside world, including but not limited to the database health status of the system to be migrated, the message queue health status, the middleware health status and other health status information indicators.

[0150] The information system grayscale release migration system runs on any computing device such as a desktop computer, a laptop computer, a mobile phone, a PDA or a cloud data center. The computing device includes: a processor, a memory and a computer program stored in the memory and running on the processor. When the processor executes the computer program, the steps in the information system grayscale release migration method are implemented. The executable system may include, but is not limited to, a processor, a memory, and a server cluster.

[0151] An embodiment of the present invention provides an information system grayscale release migration system, such as Figure 2 As shown, an information system grayscale release migration system of this embodiment includes: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps in the above-mentioned information system grayscale release migration method embodiment are implemented. The processor executes the computer program to run in the modules of the following system, mainly including a service registration center and an automated grayscale release framework:

[0152] The service registration center can be used to store grayscale ratio vectors and support grayscale publishing in a distributed environment;

[0153] The automated grayscale release framework includes:

[0154] The reverse proxy module can be used to dynamically control the grayscale release of each module according to the grayscale ratio vector and record the operation data of the system module;

[0155] The information cache module can be used to update the code of the cache module in stages over time according to the grayscale ratio;

[0156] The monitoring center can be used to monitor the system operation status, handle abnormal and slow response information, and dynamically calculate and update the grayscale ratio vector;

[0157] The system monitoring module can be used to monitor the running status in real time, push abnormal information to the message queue, and provide a monitoring interface for the health status, including the health status of the database, message queue, and middleware, as well as memory usage and the number of active threads.

[0158] Among them, preferably, all undefined variables in the present invention, if not clearly defined, can be manually set thresholds.

[0159] Among them, dimensionless numerical calculations are used between physical quantities of different units.

[0160] The information system grayscale release migration system can be run on computing devices such as desktop computers, laptops, mobile phones, PDAs, and cloud data centers. The information system grayscale release migration system includes, but is not limited to, processors and memories. Those skilled in the art can understand that the example is only an example of an information system grayscale release migration method and system, and does not constitute a limitation on an information system grayscale release migration method and system. It may include more or fewer components than the example, or a combination of certain components, or different components. For example, the information system grayscale release migration system may also include input and output devices, network access devices, buses, etc.

[0161] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete component gate circuits or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the information system grayscale release migration system, and uses various interfaces and lines to connect the various sub-areas of the entire information system grayscale release migration system.

[0162] The memory can be used to store the computer program and / or module, and the processor implements the various functions of the information system grayscale release migration method and system by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0163] The present invention provides an information system grayscale release migration method and system, which builds a full-process guarantee mechanism for grayscale release of information systems to be migrated based on fine-grained system monitoring, automatic grayscale ratio calculation and system module level grayscale release strategy; the present invention implements a fine-grained system monitoring mechanism by monitoring multiple indicators of each module of the system to be migrated, calculates system module stability indicators through a system stability discrimination model, obtains the grayscale ratio used by each module of the system to be migrated in the next stage through automatic calculation, implements automatic grayscale ratio calculation, and implements low-risk grayscale release of each module of the system to be migrated according to the system module level grayscale release strategy. The stability and system availability of the grayscale release process of information migration are guaranteed, the risk of the grayscale release process is reduced, and the efficiency of migration of the system to be migrated is improved.

[0164] Although the description of the present invention has been quite detailed and has been described in particular with respect to several described embodiments, it is not intended to be limited to any of these details or embodiments or any particular embodiment, so as to effectively cover the intended scope of the present invention. In addition, the present invention is described above with the embodiments foreseeable by the inventors, and its purpose is to provide a useful description, and those non-substantial changes to the present invention that are not currently foreseen may still represent equivalent changes of the present invention.

Claims

1. A grayscale release migration method for an information system. During the information system migration process, the information system is divided into multiple different modules, and the code of each module is updated in stages over time according to the grayscale ratio. The method is characterized in that: The method comprises the following steps: Initialize the grayscale ratio vector, where the grayscale ratio vector contains several values, each of which corresponds to a module and is used to indicate the grayscale ratio used by the module at the current stage; Based on the grayscale ratio vector, each module is updated and the operation data of the system modules are collected in real time, including the number of exceptions and response time; Based on the sliding window, the grayscale ratio vector is weighted and counted using the module's operating data to generate the system state vector; Obtain the health status information of the information system and combine it with the system status vector to generate the final system status data; The final system state data is input into the stability discrimination model to generate the module stability index of each module; According to the module stability index, adjust the grayscale ratio of each module to generate the grayscale ratio vector of the next stage; If the module stability index is lower than the threshold, the rollback mechanism is triggered for the grayscale ratio vector to restore it to the initial grayscale ratio vector.

2. The information system phased release migration method according to claim 1, characterized in that: in, The grayscale ratio vector is a floating point number vector whose length is equal to the number of modules in the system, and is used to record the grayscale ratio used by each module in the system during the information system migration process.

3. The information system phased release migration method according to claim 1, characterized in that: in, The stability discriminant model refers to a discriminant model based on the logistic regression model. The input system state data is projected into a high-dimensional space through a polynomial feature expansion operation and then input into the logistic regression model to obtain the confidence that the system module is in a stable state, that is, the system module stability index.

4. The information system phased release migration method according to claim 1, characterized in that: Based on the sliding window, the grayscale ratio vector is weighted statistically analyzed using the module's operating data to generate the system state vector: By means of sliding windows, the grayscale ratio vector of each module is counted according to the number of abnormalities and the number of slow responses of each module to obtain the statistical vector of each module: the statistical weight of each module is calculated according to the number of abnormalities and the number of slow responses of each module; Using the statistical weights of each module respectively, the dimensions corresponding to each module in the grayscale ratio vector are processed to obtain the statistical vectors of each module; the statistical vectors of each module are combined to obtain the system state vector; The specific value of the weight of the sliding window is equal to the exponential value of the weight growth rate over time, and the initialization value of the weight growth rate is greater than 1. The weight growth rate can be used to preferentially increase the weight of recent data.

5. The information system phased release migration method according to claim 4, characterized in that: The method of obtaining the health status information of the information system and concatenating it with the system status vector to generate the final system status data is as follows: Obtain health status information of the system, where the health status information is a vector that can be spliced ​​into a system status vector, and the health status information includes a numerical value for indicating a database status, a numerical value for indicating a message queue status, a numerical value for indicating a middleware status, a numerical value for indicating a memory status, and a numerical value for indicating the number of active threads. The vector consisting of the numerical value for indicating a database status, the numerical value for indicating a message queue status, the numerical value for indicating a middleware status, the numerical value for indicating a memory status, and the numerical value for indicating the number of active threads is spliced ​​into the system status vector as final system status data.

6. The information system phased release migration method according to claim 5, characterized in that: The final system state data is input into the stability judgment model to generate the module stability index of each module as follows: The logistic regression model maps the final system state data through polynomial expansion to generate the stability index of each module, wherein the value range of the stability index of each module is [0,1], and the polynomial expansion of the logistic regression model includes a combination of quadratic and cubic features; The stability index of the module can be used to determine whether the module is in a stable state.

7. The information system phased release migration method according to claim 5, characterized in that: in, According to the module stability index, the grayscale ratio of each module is adjusted to generate the grayscale ratio vector of the next stage, which specifically includes: Calculate the difference between the module stability index and the stability decision boundary; The current grayscale ratio is added to the difference between the module stability index and the set stability decision boundary and multiplied by the response index. The sum is the adjusted grayscale ratio. Among them, the module response index represents the correlation between the current grayscale ratio and the module grayscale baseline.

8. The information system phased release migration method according to claim 7, characterized in that: in, The grayscale ratio rollback mechanism includes the following steps: when the module stability index continues to be lower than the decision boundary for more than a preset time threshold, a rollback is triggered.

9. An information system grayscale release migration system, characterized in that: The information system grayscale release migration system runs in any computing device such as a desktop computer, a laptop computer or a cloud data center. The computing device includes: a processor, a memory and a computer program stored in the memory and running on the processor. When the processor executes the computer program, the steps in the information system grayscale release migration method according to any one of claims 1 to 8 are implemented. The system includes a service registration center and an automated grayscale release framework: The service registration center can be used to store grayscale ratio vectors and support grayscale publishing in a distributed environment; The automated grayscale release framework includes: The reverse proxy module can be used to dynamically control the grayscale release of each module according to the grayscale ratio vector and record the operation data of the system module; The information cache module can be used to update the code of the cache module in stages over time according to the grayscale ratio; The monitoring center can be used to monitor the system operation status, handle abnormal and slow response information, and dynamically calculate and update the grayscale ratio vector; The system monitoring module can be used to monitor the running status in real time, push abnormal information to the message queue, and provide a monitoring interface for the health status, including the health status of the database, message queue, and middleware, as well as memory usage and the number of active threads.

Images