A saas intelligent concurrent pushing method and system for private domain information

Abstract

The application relates to the technical field of information intelligent pushing, and discloses a SaaS intelligent concurrent pushing method and system for private domain information, which comprises the following steps: in response to a private domain information pushing task, a preset pushing prediction model is used to predict a task pushing quantity; based on the task pushing quantity, a resource expansion scheduling mechanism is configured to obtain a pushing resource set; according to a task feature vector, pushing information is logically fragmented to obtain a fragmented information set, a resource information matching mechanism is configured, the pushing resource set is matched with the fragmented information set, a first fragmented information set responsible for pushing by each pushing resource is obtained, and concurrent pushing is carried out; in the pushing process, according to a pushing state, an information set that fails to be pushed is decoupled, the decoupled information is matched with corresponding pushing resources, and the information is re-pushed; and through dynamic resource scheduling and load balancing, the application can improve the utilization rate of cloud server pushing resources and ensure that private domain information is on time pushed to corresponding users.

Description

Technical Field

[0001] This application relates to the field of intelligent information push technology, and more specifically to a SaaS intelligent concurrent push method and system for private domain information. Background Technology

[0002] Currently, the push of private domain information relies on SaaS tools. In the private domain ecosystem, which uses platforms such as WeChat Work, mini-programs, and social groups as carriers, enterprises need to push service notifications, marketing content, and rights reminders to users. Moreover, the push tasks have the characteristics of high concurrency, strong timeliness, and multiple scenarios. Existing information push methods are limited by server performance and technical difficulties, resulting in the inability to push information on time and failing to meet the requirements for concurrent push of private domain information.

[0003] Existing technologies suffer from the following problems: static configuration of cloud server resources based on historical data fails to analyze the characteristics of current push tasks in real time, leading to server overload during peak hours and resource idleness during off-peak hours; the single method of pushing information as a whole or in segments ignores the correlation between information, resulting in fragmented information push and poor user experience; and the use of a fixed push method during information push makes it impossible to adjust the push process in real time according to server resource usage, resulting in long push waiting times and low push efficiency. To address at least one of the above problems, this application proposes a SaaS intelligent concurrent push method and system for private domain information. Summary of the Invention

[0004] To address the shortcomings of existing technologies, the purpose of this application is to provide a SaaS intelligent concurrent push method and system for private domain information, which can effectively solve the problems in the background technology. The specific technical solution of this application is as follows:

[0005] A SaaS intelligent concurrent push method for private domain information includes:

[0006] In response to private domain information push tasks, the task feature vector is parsed out, and the number of task pushes is predicted through a preset push prediction model.

[0007] Based on the number of tasks pushed, a resource expansion scheduling mechanism is configured to schedule the push resources of multiple cloud servers to obtain a set of push resources.

[0008] The push information is logically segmented according to the task feature vector to obtain a segment information set. A resource information matching mechanism is configured to match the push resource set with the segment information set to obtain the first segment information set that each push resource is responsible for pushing, and concurrent push is performed.

[0009] During the push process, the set of information that failed to be pushed is decoupled according to the push status, and the decoupled information is matched with the corresponding push resources for re-push, so as to intelligently push private domain information concurrently.

[0010] Specifically, in response to the private domain information push task, the task feature vector is parsed, and the number of task pushes is predicted through a preset push prediction model, including:

[0011] In response to the private domain information push task, the task push time, task data and target user data are parsed to construct the task feature vector;

[0012] Based on the task feature vector, the information push volume of the current push task is predicted using a preset push prediction model to obtain the task push quantity.

[0013] Specifically, based on the number of tasks pushed, a resource expansion scheduling mechanism is configured to schedule push resources from multiple cloud servers, resulting in a set of push resources, including:

[0014] The target number of cloud servers is determined based on the number of tasks pushed and the cloud server's push performance.

[0015] Initialize cloud server instances, and select a target number of cloud server instances from multiple cloud server instances based on cloud server instance type and cost performance.

[0016] Configure a resource expansion scheduling mechanism to schedule the push resources of the cloud servers in the set of cloud server instances to obtain a set of push resources.

[0017] Specifically, the configured resource expansion scheduling mechanism schedules the push resources of cloud servers in the cloud server instance set to obtain a push resource set, including:

[0018] Analyze the push performance of each instance in the cloud server instance set and calculate the corresponding performance score;

[0019] Based on the performance scores, corresponding push resources are selected from each instance in the cloud server instance set. The instance status is analyzed based on the usage of push resources, and idle instances and overloaded instances are selected.

[0020] Dynamically migrate push resources from idle instances to balance the push content of overloaded instances, resulting in a set of push resources.

[0021] Specifically, the push information is logically segmented based on the task feature vector to obtain a segment information set. A resource information matching mechanism is configured to match the push resource set with the segment information set to obtain the first segment information set for each push resource. Concurrent push is then performed, including:

[0022] The correlation between push information is analyzed based on the task feature vector, a sharding key is generated, and the push information is logically sharded based on the sharding key to obtain a set of sharded information.

[0023] Configure a resource information matching mechanism to calculate the compatibility between push resources and segment information, match the push resource set with the segment information set to obtain the first segment information set that each push resource is responsible for pushing, and perform concurrent push.

[0024] Specifically, the step involves analyzing the correlation between push information based on task feature vectors, generating sharding keys, and logically sharding the push information based on these sharding keys to obtain a set of sharded information, including:

[0025] Based on the task feature vector, the information is classified according to the type of push information. Combining the logical and spatiotemporal relationships between each type of push information, the information correlation degree between push information categories is calculated.

[0026] Information with a relevance greater than a preset relevance threshold is selected from each push information category and used as the segmentation factor for the corresponding push information category;

[0027] Construct a sharding key by combining the sharding factors of each push message category;

[0028] The sharding key is hashed, and the push information is logically sharded to obtain a set of sharded information.

[0029] Specifically, the configuration resource information matching mechanism calculates the compatibility between push resources and shard information, matches the push resource set with the shard information set to obtain the first shard information set for each push resource, and performs concurrent pushes, including:

[0030] Analyze the type matching degree and distance similarity between the pushed resources and the segment information, calculate the corresponding fit degree, and construct the fit degree matrix;

[0031] Analyze the push timeliness of each segment of information and calculate the corresponding information timeliness value;

[0032] Based on the adaptation matrix, according to the information timeliness value from high to low, the push resource with the highest adaptation is matched for each piece of information to obtain the first matching result;

[0033] For each push resource, based on the information push capability of the push resource, the segment information in the first matching result that is greater than the information push capability is filtered out and matched with the push resource with the second highest suitability, until each push resource is matched with the first segment information set responsible for push, and concurrent push is performed.

[0034] Specifically, during the push process, the set of information that failed to be pushed is decoupled according to the push status, and the decoupled information is matched with corresponding push resources for re-pushing, including:

[0035] During the push process, the push failure type is analyzed based on the push status, the push failure information set is decoupled, and it is recombined based on the information timeliness value to obtain the recombined information set;

[0036] Analyze the matching degree between the recombined information set and the push resources, match the corresponding push resources for the recombined information set, and re-push.

[0037] Specifically, during the push process, the push failure type is analyzed based on the push status, the push failure information set is decoupled, and it is recombined based on the information timeliness value to obtain a reconstructed information set, including:

[0038] During the push process, the push failure type is analyzed based on the push status, including network failure, resource overload failure, and content error failure.

[0039] In the event of a network failure, the information set is decoupled according to the user's location to obtain the first decoupling information;

[0040] For resource overload failures, the information set is decoupled according to the amount of information pushed to obtain the second decoupling information;

[0041] For content errors and failures, invalid information is filtered out to decouple the information set and obtain third decoupling information;

[0042] Based on the information timeliness value, the first decoupling information, the second decoupling information, and the third decoupling information are recombined to obtain a reconstituted information set.

[0043] A SaaS intelligent concurrent push system for private domain information, used to implement the aforementioned SaaS intelligent concurrent push method for private domain information, includes:

[0044] The push quantity prediction module responds to the private domain information push task, parses the task feature vector, and predicts the task push quantity through the preset push prediction model.

[0045] The resource expansion scheduling module configures a resource expansion scheduling mechanism based on the number of tasks pushed, and schedules the push resources of multiple cloud servers to obtain a set of push resources;

[0046] The information matching and push module logically segments the push information according to the task feature vector to obtain a set of segmented information. It configures a resource information matching mechanism to match the push resource set with the set of segmented information to obtain the first set of segmented information that each push resource is responsible for pushing, and performs concurrent push.

[0047] The intelligent re-push module decouples the set of information that failed to be pushed during the push process based on the push status, and matches the corresponding push resources for the decoupled information to re-push, so as to intelligently push private domain information concurrently.

[0048] The beneficial effects of this application are as follows: Based on real-time predicted information push volume, the cloud server push resources are dynamically expanded and scheduled, and idle instance resources are migrated to overloaded instances to achieve resource load balancing; information correlation is analyzed, and related information is divided into the same segment. The compatibility between information segments and push resources is analyzed. Combining information timeliness and resource push capabilities, push resources with high compatibility are prioritized for information segments with high timeliness, and the matching results are dynamically adjusted to achieve concurrent information push. For information that fails to push, the matching resources are decoupled and re-matched based on information timeliness, and the corresponding information is pushed again; through dynamic resource scheduling and load balancing, the utilization rate of cloud server push resources can be improved, information push latency can be reduced, related information and segmented push can improve user experience, and intelligent concurrent push can ensure that private domain information is pushed to the corresponding users on time. Attached Figure Description

[0049] Figure 1 This is a flowchart illustrating the workflow of a SaaS intelligent concurrent push method for private domain information in an embodiment of this application.

[0050] Figure 2 This is a schematic diagram of the push information logical fragmentation process in the embodiments of this application;

[0051] Figure 3 This is a schematic diagram illustrating the matching process of push resources and fragment information in an embodiment of this application;

[0052] Figure 4 This is a schematic diagram of the structure of a SaaS intelligent concurrent push system for private domain information in an embodiment of this application. Detailed Implementation

[0053] The present application will be further described in detail below with reference to the accompanying drawings and embodiments. Identical components are indicated by the same reference numerals. It should be noted that the terms "front," "rear," "left," "right," "upper," and "lower" used in the following description refer to directions in the accompanying drawings, and the terms "bottom surface," "top surface," "inner," and "outer" refer to directions toward or away from the geometric center of a specific component, respectively.

[0054] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0055] Hereinafter, the terms "first," "second," and other generic terms are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0056] refer to Figure 1 The image shows a specific implementation of a SaaS intelligent concurrent push method for private domain information according to this application, including:

[0057] S101. In response to the private domain information push task, the task feature vector is parsed out, and the number of task pushes is predicted through the preset push prediction model.

[0058] S102. Based on the number of task pushes, configure a resource expansion scheduling mechanism to schedule push resources from multiple cloud servers to obtain a set of push resources.

[0059] S103. Logically segment the push information according to the task feature vector to obtain the segment information set, configure the resource information matching mechanism, match the push resource set with the segment information set to obtain the first segment information set that each push resource is responsible for pushing, and perform concurrent push.

[0060] S104. During the push process, the information set that failed to be pushed is decoupled according to the push status, and the corresponding push resources are matched for the decoupled information to be pushed again, so as to intelligently push private domain information concurrently.

[0061] When a user triggers a private domain information push task on the SaaS push platform, the task is parsed, and task feature data is extracted from the task configuration data and user profile database to construct a task feature vector. Task feature data includes, but is not limited to, task push time data, information type, information data volume, push priority, and user attribute data. The task feature data is sorted in a preset order to form the task feature vector. This task feature vector is then input into a preset push prediction model to predict the number of pushes. The push prediction model includes, but is not limited to, the XGBoost model. The XGBoost model is trained using a large amount of historically completed private domain push task data to obtain a pre-trained XGBoost model. The task feature vector is then input into the pre-trained XGBoost model, which predicts the number of pushes. By parsing the task features to construct the vector and combining it with the push prediction model, the number of pushes can be accurately predicted, providing data support for cloud server resource scheduling. This allows for advance expansion and scheduling of push resources, avoiding excessively long information queuing times during the push process and improving information push efficiency.

[0062] Specifically, based on the predicted number of task pushes and the push performance of a single cloud server instance, the target number of cloud server instances is calculated. The cloud server instances are initialized, their types and performance are analyzed, and the target number of cloud server instances are selected. The push resources of the cloud servers are dynamically scheduled in conjunction with the real-time performance of the cloud server instances. By allocating and elastically scaling the computing resources of the cloud servers on demand, it is possible to avoid overload failures of push tasks on a single cloud server instance while ensuring that the requirements of private domain information push tasks are met. This ensures balanced server performance, improves the stability and high availability of the cloud servers, and enhances the efficiency and accuracy of information push.

[0063] Based on task feature vectors, the correlation between information is analyzed, and multimedia information of the same type is associated to generate sharding keys. A consistent hashing algorithm is used to calculate the sharding keys, distributing a large amount of information to be pushed into a fixed number of logical shards, resulting in a sharded information set. Logically grouping the information, with highly correlated information assigned to the same shard, optimizes push efficiency, avoids fragmentation of pushed information, and improves the completeness and internationality of pushed information. A resource information matching mechanism calculates the fit between push resources and information shards. Based on the fit, the push resource set is matched with the sharded information set, resulting in the first sharded information set that each push resource is responsible for pushing. Each push resource pushes its assigned shard set in parallel. By using correlated sharding, the duplication and redundancy of network requests are reduced, achieving efficient batch processing and improving the efficiency of concurrent pushes. Assigning sharded information to push resources with high matching degrees reduces network transmission latency and processing latency, optimizing push performance and user experience, enabling push information to be delivered quickly and accurately to the corresponding users.

[0064] During the push process, the return status codes and logs of each push are continuously monitored, and the push status and failure types are analyzed. Based on the failure type, the failed push information sets are decoupled, and the decoupled information is recombined according to the timeliness of the information. Then, the corresponding push resources are matched for re-push, enabling intelligent concurrent push of private domain information. Intelligent decoupling, recombining, and re-pushing improves the success rate of information push, ensuring the smooth delivery of critical information. Using different information decoupling strategies for different types avoids the significant impact of traditional simple full retries on other information push processes, improving the stability and accuracy of the information push process.

[0065] This application dynamically expands and schedules cloud server push resources based on real-time predicted information push volume, migrating idle instance resources to overloaded instances to achieve resource load balancing. It analyzes information correlation, dividing related information into the same shard, and analyzes the compatibility between information shards and push resources. Combining information timeliness and resource push capabilities, it prioritizes matching highly compatible push resources to high-timeliness information shards, dynamically adjusting the matching results to achieve concurrent information push. For information that fails to push, it decouples and re-matches resources based on information timeliness, re-pushing the corresponding information. Through dynamic resource scheduling and load balancing, it can improve the utilization rate of cloud server push resources, reduce information push latency, improve user experience by associating and sharding information for push, and ensure that private domain information is pushed to the corresponding users on time.

[0066] Furthermore, in response to the private domain information push task, the task feature vector is parsed, and the number of task pushes is predicted through a preset push prediction model, including:

[0067] S201. In response to the private domain information push task, the task push time, task data and target user data are parsed out, and a task feature vector is constructed.

[0068] S202. Based on the task feature vector, predict the amount of information pushed for the current push task using a preset push prediction model to obtain the task push quantity.

[0069] In this embodiment, when a push task arrives, it is parsed to extract the task push time, task data, and target user data. The task push time includes, but is not limited to, timestamps, whether it is a weekday, and the time period. The task data includes, but is not limited to, the information format type, information size, and the keywords contained in the content. The target user data includes, but is not limited to, the size of the target user group, the geographical distribution density of users, and the historical behavioral characteristics of users. The extracted feature data is standardized and normalized, and the normalized features are concatenated in sequence to construct the task feature vector. By extracting comprehensive feature information, comprehensive and accurate data support is provided for predicting the number of push tasks, improving the accuracy of the prediction results, and providing accurate data support for the expansion of cloud servers.

[0070] Specifically, the task feature vector is input into a pre-defined push prediction model to predict the information push volume of the current push task, thus obtaining the task push quantity. The push prediction model includes, but is not limited to, the XGBoost model. A pre-trained XGBoost model is trained using a large amount of historically completed private domain push task data. The task feature vector is then input into the pre-trained XGBoost model, which predicts the task push quantity, i.e., the total amount of information expected to be pushed for the task. Model prediction can accurately predict the scale of push tasks. The predicted task push quantity provides a precise quantitative basis for subsequent cloud server resource planning and scheduling, and concurrent pushes. This is a prerequisite for elastic scaling and on-demand resource allocation of cloud servers, effectively avoiding push delays due to insufficient resource preparation or cost waste caused by excessive resource reservation, thereby improving information push efficiency and resource utilization.

[0071] Furthermore, based on the number of task pushes, a resource expansion scheduling mechanism is configured to schedule push resources from multiple cloud servers, resulting in a push resource set, including:

[0072] S301. Determine the target number of cloud servers based on the number of tasks pushed and the cloud server's push performance;

[0073] S302. Initialize the cloud server instances, and select a target number of cloud server instances from multiple cloud server instances based on the cloud server instance type and cost performance.

[0074] S303. Configure a resource expansion scheduling mechanism to schedule the push resources of cloud servers in the set of cloud server instances to obtain a set of push resources.

[0075] In this embodiment, the target number of cloud servers is determined based on the number of task pushes and the cloud server push performance. The cloud server push performance is calculated by combining the instance's computing power, memory bandwidth, and network throughput, and then weighting the calculated push performance value. The predicted total number of task pushes is divided by the cloud server's push performance value, and a performance redundancy coefficient is used for weighting to calculate the target number of cloud servers. The performance redundancy coefficient is dynamically adjusted according to the task's criticality level and the stability requirements of historical push success rates. By calculating the target number of cloud servers, server push resources can be accurately planned, avoiding cost waste caused by resource redundancy and avoiding push performance bottlenecks or task failures due to insufficient push resources, thereby improving the efficiency of information push tasks.

[0076] Specifically, the process involves initializing cloud server instances, selecting the target number of cloud server instances from multiple instances based on instance type and cost-performance ratio; initializing the cloud server instances again, obtaining all available cloud server instance types and their corresponding cost-performance ratios within the current available area, where cost-performance ratio is calculated by the amount of information pushed per unit cost; sorting all cloud server instance types from highest to lowest according to the calculated cost-performance ratio, constructing a priority queue; starting from the top of the priority queue, if there are sufficient available instances of the most cost-effective instance type, directly selecting the target number of instances of that type; if a single instance type cannot meet the target number requirement, a hybrid strategy is adopted, prioritizing the most cost-effective instance type, followed by the next most cost-effective instance type, until the target number of cloud server instances is obtained; and finally, combining all successfully applied and initialized instances to obtain the cloud server instance set.

[0077] It is important to emphasize that by selecting and combining cloud server instances based on their type, cost, and performance, the most cost-effective set of cloud server instances can be chosen from cloud service resources. This can reduce server resource costs while meeting performance requirements. By dynamically adjusting the set of cloud server instances in real time, push resources can be adjusted according to changes in cloud server push resources, thereby improving the adaptability and intelligence of the information push process.

[0078] Specifically, a resource expansion scheduling mechanism is configured to schedule push resources from cloud servers in the cloud server instance set, resulting in a push resource set. Multi-dimensional performance metrics for each cloud server instance are obtained, and the weighted sum of these metrics is used to calculate the real-time performance status score for each instance. Based on the performance status score, cloud server instances are classified into different status categories. According to the status of the cloud server instances, the push resources are expanded and scheduled using the resource expansion scheduling mechanism. For cloud server instances in an overloaded state, the load is migrated from their pending push queue to idle cloud server instances. This load migration prevents instances from becoming performance bottlenecks and avoids any instance resources being wasted due to idleness.

[0079] It is important to emphasize that dynamically adjusting push resources and load based on the real-time status of cloud server instances and elastically scaling the resource pool can ensure high availability of push resources. This avoids single-point overload caused by load concentration on a single cloud server instance, which could lead to information push delays or crashes, thus ensuring the stability and low latency of the overall information push process. At the same time, it avoids idle push resources and improves the utilization rate of push resources.

[0080] Furthermore, configure a resource expansion scheduling mechanism to schedule push resources from cloud server instances in the cloud server instance set, resulting in a push resource set, including:

[0081] S401. Analyze the push performance of each instance in the cloud server instance set and calculate the corresponding performance score;

[0082] S402. Based on performance scores, select corresponding push resources from each instance in the cloud server instance set, analyze the instance status based on the usage of push resources, and filter out idle instances and overloaded instances.

[0083] S403. Dynamically migrate push resources in idle instances to balance the push content of overloaded instances and obtain a set of push resources.

[0084] In this embodiment, the push performance of each instance in the cloud server instance set is analyzed, and the corresponding performance score is calculated. Real-time monitoring is performed on the CPU utilization, memory usage and availability, network interface throughput, data packet transmission and reception error rate, and disk I / O operation queue length of the cloud server instances to obtain multi-dimensional performance indicators. Based on the impact of each performance indicator on information push, corresponding weights are assigned to each dimension, and a weighted sum is calculated to obtain the performance score for each cloud server instance. Through multi-dimensional performance indicator monitoring and quantitative calculation, the push performance of instances is quantified, providing accurate data support for push resource scheduling. Scheduling is based on instance load levels, improving the utilization rate of push resources and the efficiency of information push.

[0085] Specifically, based on performance scores, corresponding push resources are selected from each instance in the cloud server instance set. The instance status is analyzed based on the usage of these push resources, identifying idle and overloaded instances. A lower idle threshold and a higher overload threshold are set according to the utilization demand of push resources, with the overload threshold being greater than the idle threshold. If the performance score of a cloud server instance is less than the idle threshold, it indicates extremely low resource utilization, and it is classified as an idle instance. If the performance score of a cloud server instance is greater than the overload threshold, it indicates that its resources are approaching saturation, its information processing capacity has reached a bottleneck, and it may cause request delays or even service unavailability, thus being classified as an overloaded instance. By filtering idle and overloaded instances, overloaded instances that cause push delays and idle instances with low push resource utilization are identified, providing targets for resource expansion scheduling, achieving load balancing, and improving the stability of the push process.

[0086] Specifically, push resources in idle instances are dynamically migrated to balance the push content of overloaded instances, resulting in a push resource set. Push tasks on overloaded instances are analyzed, including their urgency and workload. Network transmission and time latency for migrating push tasks to idle instances are calculated to determine the migration overhead. Push tasks are then migrated to idle instances using hot migration technology, prioritizing tasks with the highest possible overhead. This process does not affect ongoing tasks that have not yet been migrated, ensuring the continuity of information push. By migrating a portion of the load from overloaded instances to idle instances, the load level of the entire instance set is balanced, allowing all instances to run within their efficient operating range. This avoids localized performance bottlenecks caused by uneven task allocation, improving push resource utilization and the stability of the push process.

[0087] Furthermore, the push information is logically segmented based on the task feature vector to obtain a segment information set. A resource information matching mechanism is configured to match the push resource set with the segment information set, obtaining the first segment information set for each push resource, and concurrent push is performed, including:

[0088] S501. Analyze the correlation between push information based on task feature vectors, generate sharding keys, and logically shard the push information based on the sharding keys to obtain a set of sharded information.

[0089] S502. Configure the resource information matching mechanism to calculate the compatibility between push resources and segment information, match the push resource set with the segment information set to obtain the first segment information set that each push resource is responsible for pushing, and perform concurrent push.

[0090] In this embodiment, as Figure 2As shown, the correlation between push information is analyzed based on task feature vectors to generate sharding keys. The push information is then logically sharded based on these keys, resulting in a set of sharded information. Furthermore, the similarity of user attributes, including geographical proximity and shared user groups, is analyzed based on task feature vectors to link information corresponding to user groups with common characteristics. The similarity of information content themes or consistency of multimodal types is analyzed to extract sharding keys from highly correlated information groups. These keys are then hashed, and information with the same or similar sharding keys is grouped into one shard. The push information is then logically sharded again, resulting in a set of sharded information. Through multi-dimensional correlation analysis, related information is grouped into the same shard for batch processing, optimizing the batch processing efficiency of push information. Aggregating information destined for the same region, user group, or with similar content into the same shard reduces the overhead of repeatedly establishing connections during push, achieving efficient concurrent push.

[0091] like Figure 3 As shown, a resource information matching mechanism is configured to calculate the compatibility between push resources and segment information. The push resource set is matched with the segment information set to obtain the first segment information set that each push resource is responsible for pushing, and concurrent pushes are performed. The compatibility between push resources and segment information is calculated based on the functional type and location distance between the push resources and segment information. The timeliness requirements of segment information are analyzed. According to the timeliness requirements, the push resource with the highest current compatibility score is prioritized to match segment information with high timeliness requirements, ensuring that tasks with high timeliness requirements receive the optimal resources. Matching is performed step by step. If the matched tasks exceed the push capacity limit of the push resource, the excess segment information is adjusted to the push resource with the second best compatibility, until all segment information is matched with the corresponding push resource, and each push resource obtains the first segment information set that is responsible for pushing. After matching, the push resources begin to execute information push tasks in parallel and efficiently.

[0092] It is important to emphasize that by matching push resources and shard information, it can be ensured that each push task is executed by a highly matched push resource, reducing network transmission latency and data processing latency during task execution, thereby shortening the overall completion time of push tasks and improving push efficiency. At the same time, load balancing can avoid resource bottlenecks caused by improper matching and improve the stability of the information push process.

[0093] Furthermore, based on the task feature vector analysis, the correlation between push information is generated, a sharding key is produced, and the push information is logically sharded based on the sharding key to obtain a set of sharded information, including:

[0094] S601. Based on the task feature vector, classify the push information according to the type, and calculate the information correlation degree between the push information categories by combining the logical and spatiotemporal correlations between each type of push information.

[0095] S602. Select information with a relevance greater than a preset relevance threshold from each push information category and use it as the segmentation factor for the corresponding push information category.

[0096] S603. Combine the sharding factors of each push information category to construct a sharding key;

[0097] S604. Perform hash calculation on the sharding key, logically shard the push information, and obtain a set of sharded information.

[0098] In this embodiment, push information is categorized according to its type based on task feature vectors. The information correlation between categories is calculated by combining the logical and spatiotemporal relationships between each type of push information. Based on task features in the task feature vectors, including but not limited to content format and topic keywords, the push information is clustered into different categories. The logical and spatiotemporal relationships between each type of push information are analyzed. For logical relationships, the business relevance between user historical data and the information is analyzed. For spatiotemporal relationships, the geographical location information in the task push time and target user data is analyzed to calculate distance similarity. The information correlation between categories is calculated using co-occurrence statistics by analyzing the frequency and pattern of simultaneous or sequential occurrences of different categories of information in historical successful push tasks. By analyzing the correlation between each type of push information, data support is provided for information sharding, ensuring that information assigned to the same logical shard is not only similar in surface type but also highly correlated in logical and spatiotemporal dimensions. This enables batch optimization processing during push notifications, improving efficiency in the push phase.

[0099] Specifically, information with a relevance greater than a preset relevance threshold is selected from each push notification category and used as the sharding factor for that category. The relevance threshold is set based on the overall scale of the push task and the timeliness requirements. For the currently processed push notification category A, its associated category list is traversed, and all information categories (denoted as B and C) with a relevance score exceeding the threshold are selected. The selected information categories with a strong relevance to category A are used as the sharding factor for category A. When constructing sharding keys for category A, not only the characteristics of category A are considered, but also the characteristics of strongly related categories B and C. By combining the strong relevance between categories, it is possible to ensure that information of category A is allocated to similar or identical shards with its strongly related information during sharding, improving the relevance of information within the same shard, optimizing sharding quality, and thus improving resource matching and concurrent execution efficiency.

[0100] Simultaneously, a sharding key is constructed by combining the sharding factors of each push information category. For category A, the sharding factors of category A are obtained, including strongly related categories B and C. Features of the corresponding categories are extracted from category A, B, and C, and the extracted features are concatenated into a string in sequence to obtain the sharding key. By constructing sharding keys by fusing multi-dimensional features, and sharding information based on the sharding keys, it is possible to ensure that information units that are physically divided into the same shard also have the highest relevance in business logic, thereby achieving batch processing optimization and resource optimization for information push.

[0101] Specifically, the sharding key is hashed, and the pushed information is logically sharded to obtain a set of sharded information. A consistent hashing algorithm is used to hash the sharding key of each type of information, mapping it to an integer range. The range size is set according to the information sharding precision requirements, dividing the integer range into several consecutive ranges. Each range corresponds to a logical shard. The sharding index of each range's logical shard is calculated using modulo. Each piece of information is routed to its corresponding logical shard, and information routed to the same logical shard is considered as one shard, resulting in a set of sharded information. Through hashing and information sharding, information with the same sharding key is guaranteed to be assigned to the same shard, improving the correlation of information within a shard. Range partitioning and sharding avoid data skew, improving the stability of the information push process.

[0102] Furthermore, a resource information matching mechanism is configured to calculate the compatibility between push resources and shard information, matching the push resource set with the shard information set to obtain the first shard information set for each push resource, and performing concurrent pushes, including:

[0103] S701. Analyze the type matching degree and distance similarity between the pushed resources and the segment information, calculate the corresponding fit degree, and construct the fit degree matrix.

[0104] S702. Analyze the push timeliness of each segment of information and calculate the corresponding information timeliness value;

[0105] S703. Based on the adaptability matrix, according to the information timeliness value from high to low, match the push resource with the highest adaptability for each piece of information to obtain the first matching result;

[0106] S704. For each push resource, based on the information push capability of the push resource, filter out the segment information in the first matching result that is greater than the information push capability, and match it with the push resource with the second highest suitability, until each push resource is matched with the first segment information set responsible for push, and perform concurrent push.

[0107] In this embodiment, the type matching degree and distance similarity between push resources and fragment information are analyzed to calculate the corresponding fit degree and construct a fit degree matrix. For type matching degree, the inherent characteristics of push resources are compared with the content characteristics of fragment information. Inherent characteristic feature vectors and content feature vectors are constructed according to the corresponding dimensions. The type matching degree is obtained by calculating the cosine similarity between the inherent characteristic feature vector and the content feature vector. For distance similarity, the physical distance between the cloud server location and the location of the target user in the fragment information is analyzed, and the reciprocal of the distance is used as the distance similarity. The type matching degree and distance similarity are summed to obtain the corresponding fit degree. A fit degree matrix is ​​constructed with push resources as rows and fragment information as columns. Quantitatively calculating the fit degree between push resources and fragment information provides accurate data support for matching push resources and fragment information, improving the resource utilization of the cloud server.

[0108] Simultaneously, the push timeliness of each segment of information is analyzed, and the corresponding information timeliness value is calculated. The business attributes of the segmented information are analyzed, and each business attribute is assigned a corresponding business attribute value. Combining the push time plan in the task feature vector with the current system time, the remaining pushable time window is calculated. The business attribute value and the pushable time window are then weighted and summed to obtain the information timeliness value. By calculating the information timeliness value, the urgency of tasks is analyzed, ensuring that critical tasks with high timeliness requirements receive priority access to high-quality resources during resource allocation, thus meeting their timeliness requirements and guaranteeing timely information delivery.

[0109] Specifically, based on the adaptability matrix, each segment of information is matched with the push resource with the highest adaptability, ranked from highest to lowest information timeliness value, to obtain the first matching result. Then, all segments are sorted from highest to lowest information timeliness value to form a processing priority queue. Processing begins with the segment with the highest timeliness value at the head of the queue. The push resource with the highest adaptability score is searched in the adaptability matrix, and the segment is assigned to the corresponding push resource. After matching, the process continues with the next highest-priority segment until all segments are matched with corresponding push resources, yielding the first matching result. Prioritizing the most urgent push tasks to receive the most suitable push resources ensures that the system's limited, optimal-performance push resources are used for the most critical tasks, thereby reducing push latency for high-priority tasks and optimizing the user experience.

[0110] Specifically, for each push resource, based on its information push capability, the fragment information exceeding the information push capability in the first matching result is filtered out and matched with the push resource with the second highest suitability, until each push resource is matched with the first fragment information set responsible for push, and concurrent push is performed. For each push resource, the total workload of all fragments allocated to the push resource in the first matching result is calculated. If the total workload exceeds the push capability limit of the resource, the fragment information with relatively low suitability on the resource is filtered out. For the filtered fragment information, the second highest suitability push resource with remaining capacity is searched in the suitability matrix, and matched again, until all fragments find their assigned resources, and the total workload undertaken by each resource does not exceed its capacity limit, resulting in the final first fragment information set responsible for each resource; each push resource then begins to execute concurrent push tasks. By re-matching fragment information that exceeds the push capability of the push resource, the stability of the information push process can be improved while ensuring optimal matching, avoiding performance bottlenecks or even task failures due to overload of a single resource, thereby improving the efficiency of concurrent push.

[0111] Furthermore, during the push process, based on the push status, the set of information that failed to push is decoupled, and the decoupled information is matched with corresponding push resources for re-push, including:

[0112] S801. During the push process, the push failure type is analyzed according to the push status, the push failure information set is decoupled, and it is recombined based on the information timeliness value to obtain the recombined information set.

[0113] S802. Analyze the matching degree between the recombined information set and the push resources, match the corresponding push resources for the recombined information set, and push again.

[0114] In this embodiment, during the push process, the push failure type is analyzed based on the push status, the information set of push failures is decoupled, and recombined based on the information timeliness value to obtain a reconstructed information set. The return status code, error message, and system log of each push request are monitored in real time, and the failure type is analyzed. Failure types include network failure, resource overload failure, and content error failure. Information decoupling is performed separately for different types of failure sets. After decoupling, the urgency of the information is reassessed based on the information timeliness value and the latest status of the information fragments. Information fragments with the same or similar timeliness level are recombined to obtain a reconstructed information set. By analyzing the causes of failures and performing differentiated decoupling processing for different failure types, common fault points in the failure information set can be avoided. Information recombining ensures the rationality of the re-push task set and improves the information delivery rate.

[0115] Specifically, the matching degree between the recombined information set and the push resources is analyzed, and corresponding push resources are matched for the recombined information set and re-push. The status of all currently available push resources and the characteristics of each recombined information set are reassessed, and the matching degree between each recombined set and each available push resource is recalculated. Based on the timeliness requirements of the recombined information, from high to low, the push resource with the highest matching degree is matched for each recombined information and re-push. Through dynamic secondary matching, the success rate of re-push can be improved, effectively avoiding the original path or resource that may lead to further failure, thus improving the reliability and robustness of the system and optimizing the overall resource utilization efficiency.

[0116] Furthermore, during the push process, the push failure type is analyzed based on the push status, the information set of push failures is decoupled, and then recombined based on the information timeliness value to obtain a reconstructed information set, including:

[0117] S901. During the push process, analyze the push failure type based on the push status. The push failure types include network failure, resource overload failure, and content error failure.

[0118] S902. For network failures, decouple the information set according to the user's location to obtain the first decoupling information;

[0119] S903. For resource overload failure, decouple the information set according to the amount of information pushed to obtain the second decoupling information.

[0120] S904. For content errors and failures, filter invalid information to decouple the information set and obtain third decoupling information;

[0121] S905. Based on the information timeliness value, the first decoupling information, the second decoupling information, and the third decoupling information are recombined to obtain a recombined information set.

[0122] In this embodiment, during the push process, push failure types are analyzed based on the push status. These failure types include network failure, resource overload failure, and content error failure. Response data for each push request is collected in real time, and the status information and failure type of push failure are analyzed. Network failures include consecutive connection timeouts, DNS resolution failures, or widespread failures on specific carrier network segments. Resource overload failures include errors such as server busy, rate limiting, or queue full returned by the receiving end service. Content error failures include invalid parameters, format validation failures, non-existent users, or violations of content security policies. Classifying push failure data into several typical types provides a clear decision-making basis for subsequent implementation of differentiated and targeted recovery and re-pushing strategies, avoiding the blindness of uniform retries and improving the success rate of re-pushing.

[0123] For network failures, the information set is decoupled according to the user's location to obtain the first decoupling information. Based on user registration information, IP address resolution, or the network access point of the last successful connection, the geographical location of the target user for each failure message is determined. The failure information set is then grouped according to geographical location tags, and information destined for users in the same region or network segment is aggregated together to obtain the first decoupling information. Decoupling failure information by geographical location can avoid global re-push due to network jitter in a certain location. This allows subsequent re-push to select the optimal resources for specific fault domains, thereby greatly improving the success rate and efficiency of re-push and reducing unnecessary cross-network scheduling.

[0124] For resource overload failures, the information set is decoupled according to the amount of information pushed, resulting in second decoupling information. The push workload represented by each message in the failure information set is analyzed, including the size of the data packet, the required processing complexity, or the estimated processing time at the receiving end. Based on historical experience or by setting a safe batch threshold much smaller than the threshold causing overload, the large batch of failures is divided into multiple smaller information units with lower processing overhead, resulting in second decoupling information. By decomposing the information set into multiple small information units that the receiving end can process quickly, the flow control protection of the receiving end or causing overload can be avoided during re-push, significantly improving the re-push success rate when facing unstable or limited-processing-capacity receivers.

[0125] For content errors leading to failure, invalid information is filtered to decouple the information set, resulting in third decoupling information. Each message in the failure set is analyzed individually to identify invalid information, including: target user identifiers with non-standard formats, deactivated user accounts, illegal characters or content in the message body, and messages that have become invalid due to changes in business status. Invalid information is filtered out and isolated from the failure information set, and no further re-pushing attempts are made on it. This information, after removing invalid information, serves as the third decoupling information. Removing invalid information prevents system resources from being wasted on repeated push attempts, saving network and computing resources and improving the efficiency and reliability of re-pushing operations.

[0126] Specifically, based on information timeliness values, the first, second, and third decoupling information are recombined to obtain a reconstituted information set. The information timeliness values ​​of these three information sets are then analyzed and calculated. Information with similar timeliness values, ranked from highest to lowest, is aggregated to obtain the reconstituted information set. This information reconstituted based on timeliness values ​​ensures that subsequent re-push resource scheduling prioritizes tasks with high timeliness requirements, reducing the impact of push failures on business operations and ensuring accurate delivery of pushed information.

[0127] like Figure 4 As shown, a SaaS intelligent concurrent push system for private domain information is used to implement a SaaS intelligent concurrent push method for private domain information, including:

[0128] The push quantity prediction module responds to the private domain information push task, parses the task feature vector, and predicts the task push quantity through the preset push prediction model.

[0129] The resource expansion scheduling module configures a resource expansion scheduling mechanism based on the number of tasks pushed, and schedules the push resources of multiple cloud servers to obtain a set of push resources;

[0130] The information matching and push module logically segments the push information based on the task feature vector to obtain a set of segmented information. It then configures a resource information matching mechanism to match the push resource set with the set of segmented information to obtain the first set of segmented information that each push resource is responsible for pushing, and performs concurrent push.

[0131] The intelligent re-push module decouples the set of information that failed to be pushed during the push process based on the push status, and matches the corresponding push resources for the decoupled information to re-push, so as to intelligently push private domain information concurrently.

[0132] In this embodiment, the push quantity prediction module responds to the private domain information push task. By analyzing the task configuration data and user profile database in a targeted manner, it extracts three types of information: task push time, task data, and target user data. It then transforms the unstructured data into a standardized task feature vector according to a unified quantification rule. The task feature vector is input into a preset push prediction model to calculate and output the task push quantity after excluding invalid users. The predicted task push quantity provides data support for resource scheduling and concurrent push.

[0133] The resource expansion scheduling module calculates the target number of cloud servers to meet the push requirements based on the number of tasks pushed and the stable push performance of each cloud server. It initializes the cloud server instances, selects the target number of instances from the initialized instances, and schedules the push resources of the instance set through multi-dimensional performance evaluation and dynamic load balancing to obtain the push resource set. It accurately matches resource requirements with configurations to avoid cost waste or insufficient capacity caused by static resource configurations. Through dynamic load balancing, it adjusts the resource load status in real time to ensure that each server is always in a high-efficiency and stable operating range. It also elastically schedules the push resources of the cloud servers, adjusting the push resources according to the real-time push load to improve the utilization rate and work efficiency of push resources.

[0134] The information matching and push module analyzes the logical and spatiotemporal relationships of push information based on task feature vectors, generates sharding keys, and performs logical sharding of information through hash calculation, resulting in a balanced set of sharded information with strong correlation. Through a resource information matching mechanism, it calculates the type matching degree and distance similarity between push resources and sharded information, constructing a fit matrix. Combining the timeliness value of sharded information, it matches the information shard with the push resource with the highest fit, while adjusting the matching results according to the resource push capability to ensure reasonable resource load. It drives each push resource to start multi-threaded concurrent push, and logical sharding of information based on correlation ensures that the same sharded information has related processing logic. Precise matching of resources and shards reduces push failures caused by type incompatibility or long geographical distances, improving the information push success rate.

[0135] During concurrent push operations, the intelligent re-push module analyzes push failure types by monitoring the status feedback data of each push resource in real time; it performs differentiated decoupling for different failure types; it reassembles the decoupled valid information into a time-priority set based on the information timeliness value; it calculates the multi-dimensional matching degree between the reassembled set and the push resource, matches and adapts the resource, and re-pushes to ensure that private domain information can be accurately delivered.

[0136] The above description is merely a preferred embodiment of this application. The scope of protection of this application is not limited to the above embodiments. All technical solutions falling within the scope of this application's concept are within the scope of protection of this application. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of this application should also be considered within the scope of protection of this application.

Claims

1. A SaaS intelligent concurrent push method for private domain information, characterized in that, include: In response to private domain information push tasks, the task feature vector is parsed out, and the number of task pushes is predicted through a preset push prediction model. Based on the number of tasks pushed, a resource expansion scheduling mechanism is configured to schedule the push resources of multiple cloud servers to obtain a set of push resources. The push information is logically segmented according to the task feature vector to obtain a segment information set. A resource information matching mechanism is configured to match the push resource set with the segment information set to obtain the first segment information set that each push resource is responsible for pushing, and concurrent push is performed. During the push process, the set of information that failed to be pushed is decoupled according to the push status, and the decoupled information is matched with the corresponding push resources for re-push, so as to intelligently push private domain information concurrently.

2. The SaaS intelligent concurrent push method for private domain information according to claim 1, characterized in that, The response to the private domain information push task involves parsing the task feature vector and predicting the number of task pushes using a preset push prediction model, including: In response to the private domain information push task, the task push time, task data and target user data are parsed to construct the task feature vector; Based on the task feature vector, the information push volume of the current push task is predicted using a preset push prediction model to obtain the task push quantity.

3. The SaaS intelligent concurrent push method for private domain information according to claim 1, characterized in that, Based on the number of tasks pushed, a resource expansion scheduling mechanism is configured to schedule push resources from multiple cloud servers, resulting in a set of push resources, including: The target number of cloud servers is determined based on the number of tasks pushed and the cloud server's push performance. Initialize cloud server instances, and select a target number of cloud server instances from multiple cloud server instances based on cloud server instance type and cost performance. Configure a resource expansion scheduling mechanism to schedule the push resources of the cloud servers in the set of cloud server instances to obtain a set of push resources.

4. The SaaS intelligent concurrent push method for private domain information according to claim 3, characterized in that, The configured resource expansion scheduling mechanism schedules the push resources of cloud servers in the set of cloud server instances to obtain a set of push resources, including: Analyze the push performance of each instance in the cloud server instance set and calculate the corresponding performance score; Based on the performance scores, corresponding push resources are selected from each instance in the cloud server instance set. The instance status is analyzed based on the usage of push resources, and idle instances and overloaded instances are selected. Dynamically migrate push resources from idle instances to balance the push content of overloaded instances, resulting in a set of push resources.

5. The SaaS intelligent concurrent push method for private domain information according to claim 1, characterized in that, The push information is logically segmented based on the task feature vector to obtain a segment information set. A resource information matching mechanism is configured to match the push resource set with the segment information set to obtain the first segment information set for each push resource. Concurrent push is then performed, including: The correlation between push information is analyzed based on the task feature vector, a sharding key is generated, and the push information is logically sharded based on the sharding key to obtain a set of sharded information. Configure a resource information matching mechanism to calculate the compatibility between push resources and segment information, match the push resource set with the segment information set to obtain the first segment information set that each push resource is responsible for pushing, and perform concurrent push.

6. The SaaS intelligent concurrent push method for private domain information according to claim 5, characterized in that, The process involves analyzing the correlation between push notifications based on task feature vectors, generating sharding keys, and logically sharding the push notifications based on these sharding keys to obtain a set of sharded information, including: Based on the task feature vector, the information is classified according to the type of push information. Combining the logical and spatiotemporal relationships between each type of push information, the information correlation degree between push information categories is calculated. Information with a relevance greater than a preset relevance threshold is selected from each push information category and used as the segmentation factor for the corresponding push information category; Construct a sharding key by combining the sharding factors of each push message category; The sharding key is hashed, and the push information is logically sharded to obtain a set of sharded information.

7. The SaaS intelligent concurrent push method for private domain information according to claim 5, characterized in that, The configuration resource information matching mechanism calculates the compatibility between push resources and fragment information, matches the push resource set with the fragment information set to obtain the first fragment information set for each push resource, and performs concurrent push, including: Analyze the type matching degree and distance similarity between the pushed resources and the segment information, calculate the corresponding fit degree, and construct the fit degree matrix; Analyze the push timeliness of each segment of information and calculate the corresponding information timeliness value; Based on the adaptation matrix, according to the information timeliness value from high to low, the push resource with the highest adaptation is matched for each piece of information to obtain the first matching result; For each push resource, based on the information push capability of the push resource, the segment information in the first matching result that is greater than the information push capability is filtered out and matched with the push resource with the second highest suitability, until each push resource is matched with the first segment information set responsible for push, and concurrent push is performed.

8. The SaaS intelligent concurrent push method for private domain information according to claim 1, characterized in that, During the push process, based on the push status, the set of information that failed to push is decoupled, and the decoupled information is matched with corresponding push resources for re-pushing, including: During the push process, the push failure type is analyzed based on the push status, the push failure information set is decoupled, and it is recombined based on the information timeliness value to obtain the recombined information set; Analyze the matching degree between the recombined information set and the push resources, match the corresponding push resources for the recombined information set, and re-push.

9. The SaaS intelligent concurrent push method for private domain information according to claim 8, characterized in that, During the push process, the push failure type is analyzed based on the push status, the push failure information set is decoupled, and it is recombined based on the information timeliness value to obtain a reconstructed information set, including: During the push process, the push failure type is analyzed based on the push status, including network failure, resource overload failure, and content error failure. In the event of a network failure, the information set is decoupled according to the user's location to obtain the first decoupling information; For resource overload failures, the information set is decoupled according to the amount of information pushed to obtain the second decoupling information; For content errors and failures, invalid information is filtered out to decouple the information set and obtain third decoupling information; Based on the information timeliness value, the first decoupling information, the second decoupling information, and the third decoupling information are recombined to obtain a reconstituted information set.

10. A SaaS intelligent concurrent push system for private domain information, characterized in that, A SaaS intelligent concurrent push method for private domain information as described in any one of claims 1 to 9 includes: The push quantity prediction module responds to the private domain information push task, parses the task feature vector, and predicts the task push quantity through the preset push prediction model. The resource expansion scheduling module configures a resource expansion scheduling mechanism based on the number of tasks pushed, and schedules the push resources of multiple cloud servers to obtain a set of push resources; The information matching and push module logically segments the push information according to the task feature vector to obtain a set of segmented information. It configures a resource information matching mechanism to match the push resource set with the set of segmented information to obtain the first set of segmented information that each push resource is responsible for pushing, and performs concurrent push. The intelligent re-push module decouples the set of information that failed to be pushed during the push process based on the push status, and matches the corresponding push resources for the decoupled information to re-push, so as to intelligently push private domain information concurrently.

Images