Network traffic scheduling method and device based on multi-cloud environment, equipment and medium

Abstract

The invention relates to the technical field of cloud platforms, and discloses a network traffic scheduling method and device based on a multi-cloud environment, equipment and a medium, and the method comprises the steps: monitoring the real-time load state and real-time traffic data of each cloud platform in the multi-cloud environment, analyzing dynamic network traffic characteristics of each cloud platform according to the real-time load state and the real-time traffic data; classifying a traffic feature mapping table in a multi-cloud environment according to the dynamic network traffic features, and analyzing traffic demand data of each cloud platform based on the classified traffic feature mapping table; and analyzing a network traffic scheduling queue of each cloud platform by using the traffic demand data and backup traffic data in a backup cloud platform in the multi-cloud environment, and performing traffic scheduling on each cloud platform according to the network traffic scheduling queue. According to the invention, the network traffic scheduling accuracy under the condition of fixed-level traffic distribution can be improved.

Description

Technical Field

[0001] This invention relates to the field of cloud platform technology, and in particular to a network traffic scheduling method, apparatus, device and medium based on a multi-cloud environment. Background Technology

[0002] With the acceleration of cloud computing technology and enterprise digital transformation, multi-cloud architecture has become the mainstream model for enterprise IT deployment. Network traffic scheduling is the core hub of multi-cloud environment. Therefore, in order to achieve balanced scheduling of network traffic in multi-cloud environment, it is necessary to analyze traffic applications in multi-cloud environment to achieve the accuracy of traffic scheduling in multi-source environment.

[0003] Existing traffic scheduling technologies mostly rely on DNS resolution to redirect domain name requests to different IP addresses to achieve traffic distribution. This can only be done at the domain name level and cannot detect dynamic indicators such as real-time traffic load and network latency. It is difficult to cope with sudden traffic changes, resulting in low accuracy of network traffic scheduling under fixed-level traffic distribution conditions. Summary of the Invention

[0004] This invention provides a network traffic scheduling method, apparatus, device, and medium based on a multi-cloud environment to solve the problem of low accuracy in network traffic scheduling under fixed-level traffic distribution.

[0005] Firstly, a network traffic scheduling method based on a multi-cloud environment is provided, including: Monitor the real-time load status and real-time traffic data of each cloud platform in a multi-cloud environment, and analyze the dynamic network traffic characteristics of each cloud platform based on the real-time load status and real-time traffic data. Based on the dynamic network traffic characteristics, the traffic feature mapping table in the multi-cloud environment is classified, and the traffic demand data of each cloud platform is analyzed based on the classified traffic feature mapping table. The network traffic scheduling queue of each cloud platform is analyzed using the traffic demand data and backup traffic data in the backup cloud platform under the multi-cloud environment, and traffic scheduling is performed on each cloud platform according to the network traffic scheduling queue.

[0006] Optionally, the step of analyzing the dynamic network traffic characteristics of each cloud platform based on the real-time load status and the real-time traffic data includes: Extract the real-time load status sequence and real-time traffic data sequence corresponding to the preset time window; The real-time load status sequence is time-aligned with the real-time traffic data sequence to obtain the aligned data timing sequence. Analyze the flow characteristic factors at each time point in the aligned data time series; A dynamic traffic feature sequence for each cloud platform is generated based on the traffic feature factors. The dynamic network traffic characteristics of each cloud platform are determined by the dynamic traffic characteristic sequence.

[0007] Optionally, before classifying the traffic feature mapping table in a multi-cloud environment based on the dynamic network traffic characteristics, the method further includes: The mapping table structure for each cloud platform is determined based on pre-defined mapping table field data; The fields of the mapping table structure are standardized to obtain a unified mapping table structure. The common traffic fields in a multi-cloud environment are determined based on the unified mapping table structure. By integrating the unified mapping table structure with the common traffic field, a traffic feature mapping table for a multi-cloud environment is obtained.

[0008] Optionally, classifying the traffic feature mapping table in a multi-cloud environment based on the dynamic network traffic characteristics includes: Write the dynamic network traffic characteristics into the traffic field of the traffic characteristic mapping table in the multi-cloud environment; The dynamic network traffic characteristics corresponding to the traffic field are dynamically refreshed, and a traffic characteristic time chain is generated based on the refreshed dynamic network traffic characteristics. Extract the target network traffic features corresponding to the traffic feature time chain list based on the preset timestamp; Based on the bidirectional feature categories corresponding to the target network traffic characteristics, the traffic feature table in the multi-cloud environment is classified to obtain a positive traffic feature sub-table and a negative traffic feature sub-table.

[0009] Optionally, the analysis of traffic demand data for each cloud platform based on the classified traffic feature mapping table includes: The positive traffic data of each cloud platform is determined based on the positive traffic feature sub-table in the classified traffic feature mapping table, and the negative traffic data of each cloud platform is determined based on the negative traffic feature sub-table in the classified traffic feature mapping table. The traffic demand dimension for each cloud platform is determined based on the positive traffic data and the negative traffic data. The traffic demand data for each cloud platform is determined based on the aforementioned traffic demand dimensions.

[0010] Optionally, the step of analyzing the network traffic scheduling queue of each cloud platform using the traffic demand data and backup traffic data in the backup cloud platform under a multi-cloud environment includes: The total demand traffic under the multi-cloud environment is calculated based on the traffic demand data of each platform, and the total traffic surplus under the multi-cloud environment is calculated based on the positive traffic data. When the total demand traffic is less than or equal to the total traffic reserve, a network traffic scheduling sequence for each cloud platform is generated based on the total demand traffic. When the total demand traffic exceeds the total remaining traffic, extract the backup traffic data of the backup cloud platform in the multi-cloud environment; A network traffic scheduling queue for each cloud platform is generated based on the demand traffic data and the backup traffic data.

[0011] Optionally, generating a network traffic scheduling queue for each cloud platform based on the demand traffic data and the backup traffic data includes: The demand traffic data of each cloud platform is initially sorted in descending order to obtain the first demand traffic sequence. Then, traffic is allocated to each cloud platform one by one according to the total traffic surplus and the first demand traffic sequence. The demand traffic data corresponding to the unallocated cloud platforms is re-sorted to obtain a second demand traffic sequence. Traffic is then allocated to the unallocated cloud platforms one by one according to the backup traffic data and the second demand traffic sequence. Based on the allocated cloud platform statistics, the backup traffic balance corresponding to the backup traffic data is divided according to the required traffic data of each cloud platform and the preset linked list element threshold, so as to obtain the partitioned linked list element of each cloud platform. The network traffic scheduling queue for each cloud platform is formed by dividing the linked list elements according to each cloud platform.

[0012] Secondly, a network traffic scheduling device based on a multi-cloud environment is provided, including: The dynamic network traffic characteristic analysis module is used to monitor the real-time load status and real-time traffic data of each cloud platform in a multi-cloud environment, and analyze the dynamic network traffic characteristics of each cloud platform based on the real-time load status and real-time traffic data. The traffic demand data analysis module is used to classify the traffic feature mapping table in the multi-cloud environment according to the dynamic network traffic characteristics, and analyze the traffic demand data of each cloud platform based on the classified traffic feature mapping table. The traffic scheduling module is used to analyze the network traffic scheduling queue of each cloud platform using the traffic demand data and the backup traffic data in the backup cloud platform in the multi-cloud environment, and to perform traffic scheduling for each cloud platform according to the network traffic scheduling queue.

[0013] Thirdly, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the network traffic scheduling method based on a multi-cloud environment described above.

[0014] Fourthly, a computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements the steps of the network traffic scheduling method based on a multi-cloud environment described above.

[0015] This invention dynamically senses fluctuations in business traffic by monitoring the real-time load status and traffic data of each cloud platform. Based on a real-time data-classified traffic feature mapping table, it can more accurately analyze the actual needs of each cloud platform. Network traffic scheduling queues, through priority sorting and hierarchical allocation, ensure efficient matching of demand traffic with available resources. Sorting and hierarchically allocating traffic according to demand from largest to smallest prioritizes resource supply to high-demand platforms, reducing the risk of interruptions to critical services. When all queue positions on a cloud platform fail to transmit, a backup cloud platform is automatically used, completing failover within seconds. The failed traffic data is written to the common traffic field of the traffic feature mapping table for post-failure analysis. Therefore, the network traffic scheduling method, apparatus, device, and medium proposed in this invention, based on a multi-cloud environment, can solve the problem of low accuracy in network traffic scheduling under fixed-level traffic distribution conditions. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a schematic diagram of an application environment for a network traffic scheduling method based on a multi-cloud environment according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating a network traffic scheduling method based on a multi-cloud environment according to an embodiment of the present invention. Figure 3 This is a functional block diagram of a network traffic scheduling device based on a multi-cloud environment provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of a computer device according to an embodiment of the present invention; Figure 5 This is another structural schematic diagram of a computer device according to one embodiment of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] The network traffic scheduling method based on a multi-cloud environment provided in this invention can be applied to, for example... Figure 1 In this application environment, the client communicates with the server via a network. The server can monitor the real-time load status and traffic data of each cloud platform in a multi-cloud environment through the client. Based on the real-time load status and traffic data, it analyzes the dynamic network traffic characteristics of each cloud platform. It then classifies the traffic characteristic mapping table in the multi-cloud environment according to the dynamic network traffic characteristics, analyzes the traffic demand data of each cloud platform based on the classified traffic characteristic mapping table, and analyzes the network traffic scheduling queue of each cloud platform using the traffic demand data and backup traffic data in the backup cloud platform in the multi-cloud environment. It then performs traffic scheduling for each cloud platform according to the network traffic scheduling queue and feeds the scheduled traffic back to the client. In this invention, through... The system monitors the real-time load status and traffic data of each cloud platform to dynamically sense fluctuations in business traffic. Based on a real-time data-classified traffic feature mapping table, it can more accurately analyze the actual needs of each cloud platform. Network traffic scheduling queues, through priority sorting and hierarchical allocation, ensure efficient matching of demand traffic with available resources. Sorting and hierarchically allocating traffic according to demand from largest to smallest prioritizes resource supply to high-demand platforms, reducing the risk of interruptions to critical services. When all queue positions on a cloud platform fail to transmit, it automatically replaces with a backup cloud platform, completing failover within seconds. The failed traffic data is written to the common traffic field of the traffic feature mapping table for post-failure analysis. Clients can be, but are not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server can be implemented using a standalone server or a server cluster consisting of multiple servers. The invention will be described in detail below through specific embodiments.

[0020] Reference Figure 2 The diagram shown is a flowchart illustrating a network traffic scheduling method based on a multi-cloud environment according to an embodiment of the present invention. In this embodiment, the network traffic scheduling method based on a multi-cloud environment includes: S1. Monitor the real-time load status and real-time traffic data of each cloud platform in a multi-cloud environment, and analyze the dynamic network traffic characteristics of each cloud platform based on the real-time load status and real-time traffic data.

[0021] In one practical application scenario of this invention, the unified underlying CPU, GPU, storage, and network resources in a multi-cloud environment, combined with large models and business data, provide flexible and reliable infrastructure such as GPU computing, container scheduling, secure isolated networks, unified distributed storage, and database caching for upper-layer large model pre-training or inference applications. This enables rapid deployment and simplified management of large models, facilitating the efficient delivery and stable operation of AI applications.

[0022] In this embodiment of the invention, a multi-cloud environment refers to an IT architecture that uses two or more cloud service providers simultaneously, with resources distributed across different regions and physical nodes of different service providers, and data collection needs to cross network boundaries; while the real-time load status refers to the resource usage pressure and operating status of the cloud platform at the current moment, mainly reflecting the occupancy of core resources such as computing, storage, and network; and the real-time traffic data refers to the amount of data traffic transmitted in real time in the cloud platform network.

[0023] In detail, real-time load status is monitored through cloud-native monitoring services of each cloud platform in a multi-cloud environment, and real-time traffic data is monitored through network traffic acquisition devices, such as mirroring traffic from a specified port to the monitoring device through a switch to achieve real-time traffic analysis.

[0024] Specifically, edge probes are deployed to the egress nodes of each cloud platform to collect network latency, packet loss rate, and bandwidth utilization, and to filter out timeouts and duplicate data, reducing the processing pressure on the cloud. Through SDK embedding into business systems, traffic is tagged with business type, priority, and network quality indicators. For example, core business: financial payment (priority 0.8, SLA: latency ≤100ms, packet loss rate ≤0.1%); important business: e-commerce order placement (priority 0.7, SLA: latency ≤300ms, packet loss rate ≤0.5%); non-core business: log synchronization (priority 0.3, SLA: latency ≤1s, packet loss rate ≤5%).

[0025] Furthermore, by analyzing the load on application resources, scaling up and down can be performed according to specific strategies. Therefore, in a multi-cloud environment, load balancing and network traffic scheduling can be intelligently implemented to improve application response speed and reliability.

[0026] In this embodiment of the invention, the dynamic network traffic characteristics refer to the changing patterns, characteristics, and correlation with resource load over time, such as the difference between the load and traffic of a cloud platform.

[0027] In this embodiment of the invention, the step of analyzing the dynamic network traffic characteristics of each cloud platform based on the real-time load status and the real-time traffic data includes: Extract the real-time load status sequence and real-time traffic data sequence corresponding to the preset time window; The real-time load status sequence is time-aligned with the real-time traffic data sequence to obtain the aligned data timing sequence. Analyze the flow characteristic factors at each time point in the aligned data time series; A dynamic traffic feature sequence for each cloud platform is generated based on the traffic feature factors. The dynamic network traffic characteristics of each cloud platform are determined by the dynamic traffic characteristic sequence.

[0028] In detail, the window size is selected according to the analysis purpose, such as 5 minutes, 10 minutes, etc. Different monitoring tools may have different collection frequencies, such as load data collected every second and traffic data collected every 5 seconds. To unify the timestamps, load data is unified by interpolation or resampling, such as the time granularity of data collected every second and traffic data collected every 5 seconds, so as to obtain the aligned data time series. Then, the traffic characteristic factors at each time point in the aligned data time series are analyzed. The traffic characteristic factors are quantitative indicators that describe traffic characteristics and are generated by mathematical transformation of load data and traffic data.

[0029] Specifically, LSTM (Long Short-Term Memory) networks are used to train traffic data sequences to extract long-term fluctuation patterns (such as traffic peaks from 9:00 to 12:00 on weekdays and troughs from 2:00 to 4:00 AM), outputting trend slopes (e.g., a slope greater than 0.6 indicates a rapid increase in traffic, and a slope less than -0.4 indicates a sharp drop in traffic) and periodic coefficients (e.g., e-commerce platforms experience periodic peaks every Saturday at 10:00 AM), thereby obtaining time-series trend characteristics. The correlation between traffic and CPU / GPU / storage can be calculated using mutual information entropy. For example, when the mutual information entropy between video transcoding service traffic and CPU utilization is >0.8, it is determined that the two are strongly correlated, and CPU resources need to be adjusted synchronously during scheduling.

[0030] In this embodiment of the invention, the analysis of the traffic characteristic factors at each time point in the aligned data time series includes: Extract the load and traffic data corresponding to each time point in the aligned data time series; The load data and the traffic data are unified; Calculate the difference characteristics between the unified load data and traffic data at each time point; The dynamic flow characteristic factor for each time point is determined based on the difference characteristics.

[0031] In detail, load data, such as data collected per second and traffic data, is unified through interpolation or resampling, with a time granularity of data collected every 5 seconds. This transforms load data with different dimensions and distributions, such as CPU utilization percentage and traffic data, such as bandwidth in Mbps, into a comparable unified scale. This can be achieved through standardization or normalization. Hidden patterns are then discovered through the relative differences in the unified data. For example, the absolute difference is calculated by subtracting traffic data from load data, and the relative difference is calculated by subtracting traffic data from load data and then dividing by traffic data. The difference characteristic reflects the imbalance between resource consumption and data transmission. For instance, when the difference is positive and continuously increasing, it may indicate that computing resources are overloaded but traffic is not increasing synchronously. When the difference is negative and the absolute value increases, it may indicate that network bandwidth has become a bottleneck and network resources need to be expanded. Thus, the difference characteristic is determined as a dynamic traffic characteristic factor corresponding to each time point. The value of the traffic characteristic factor varies over time, meaning that the traffic characteristic factor is constantly changing.

[0032] Specifically, the characteristic factors at each time point are combined sequentially to form a dynamic flow characteristic sequence. The dynamic flow characteristic sequence refers to the characteristic factors corresponding to each time point, such as... The time-feature factor is ,exist The time-feature factor is , The time-feature factor is Then the dynamic flow characteristic sequence is { , , This allows us to determine the dynamic network traffic characteristics of each cloud platform, such as the dynamic network traffic characteristics of cloud platform A being {}. , , }

[0033] Furthermore, using the Isolation Forest algorithm (training set consisting of normal traffic data from the past 30 days), it detects sudden traffic spikes (such as a sudden increase in traffic exceeding 150% or a sudden increase in packet loss exceeding 5%) in real time, outputs anomaly confidence scores (such as a confidence score >0.95 triggering an audible and visual warning, and <0.7 marking it as a suspected anomaly), and locates the source of the anomaly (such as bandwidth congestion caused by a surge in user access in a certain region). It calculates the adaptation coefficient by combining the network quality indicators of the business, and has a built-in feature evaluator that calculates the predictive contribution of each feature weekly based on 100,000 scheduling records (such as retaining time-series trend feature contribution >0.6 and switching algorithms if correlation feature contribution <0.2, such as changing from mutual information entropy to Pearson correlation coefficient), ensuring continuous improvement in feature accuracy. This allows for the filtering of traffic feature factors to obtain more accurate dynamic smoothness feature factors.

[0034] Furthermore, in a multi-cloud environment, traffic on different cloud platforms has inherent differences. If traffic is uniformly scheduled based on average traffic, it may lead to resource waste, such as reserving excessive bandwidth for sudden traffic or service degradation. Therefore, it is necessary to clarify the relationship between the traffic characteristics and resource utilization of each cloud platform.

[0035] S2. Classify the traffic feature mapping table in the multi-cloud environment according to the dynamic network traffic characteristics, and analyze the traffic demand data of each cloud platform based on the classified traffic feature mapping table.

[0036] In this embodiment of the invention, in a multi-cloud environment, the traffic data structures of different cloud platforms may differ. Therefore, before traffic management and scheduling, it is necessary to construct a unified traffic feature mapping table. The traffic feature mapping table is a standardized data structure used to uniformly store and manage the traffic feature data of each cloud platform in a multi-cloud environment. Through field unification and common field extraction, the original traffic data of different platforms are mapped to the same model to achieve consistent expression of cross-platform data.

[0037] In this embodiment of the invention, for efficient traffic management and scheduling in a multi-cloud environment, before classifying the traffic feature mapping table in the multi-cloud environment according to the dynamic network traffic characteristics, the method further includes: The mapping table structure for each cloud platform is determined based on pre-defined mapping table field data; The fields of the mapping table structure are standardized to obtain a unified mapping table structure. The common traffic fields in a multi-cloud environment are determined based on the unified mapping table structure. By integrating the unified mapping table structure with the common traffic field, a traffic feature mapping table for a multi-cloud environment is obtained.

[0038] In detail, the storage structure and field composition of the raw traffic data of each cloud platform are obtained. This involves collecting the traffic data interface documents, log formats, or database table structures provided by each cloud platform, extracting the raw fields, such as the cloud platform identifier field, load status field, and traffic status field. This eliminates the differences in field naming, data types, and units among different cloud platforms, forming a standardized structure. Fields with the same semantics but different names across different platforms are uniformly named. For common fields not provided by some cloud platforms, default values ​​or empty fields are reserved in the unified structure. This solves the data semantic gap in a multi-cloud environment, enabling data from different platforms to be processed and analyzed uniformly.

[0039] Specifically, by selecting common fields that can be provided by various cloud platforms from a unified structure to ensure data compatibility, common traffic fields refer to traffic data available in a multi-cloud environment, which can be used for traffic scheduling by all cloud platforms in a multi-cloud environment. Then, by integrating the standardized structure and common fields, a unified cross-platform data model is formed. With a unified mapping table structure as the framework, common traffic fields are made mandatory, and some extended fields are reserved to be compatible with future added platforms or indicators. Thus, the traffic feature mapping table becomes a common language for cross-cloud platform data interaction, supports the unified formulation of subsequent traffic classification and scheduling strategies, can eliminate data format differences in a multi-cloud environment, and facilitate centralized monitoring and analysis.

[0040] In addition, by adding core fields to the unified mapping table structure to achieve full lifecycle recording, the core fields include the cloud platform ID, traffic timestamp, and bandwidth requirement from the basic fields, while retaining the original fields and supplementing the timestamp with millisecond precision; the business association fields include business type, priority, and network quality indicators, thereby associating business attributes and supporting classification dimensions; and the fault and energy consumption fields include the number of failures and fault type, thereby recording historical faults (link congestion / node downtime) and cloud platform energy consumption.

[0041] Furthermore, in order to cope with the complexity of multi-cloud architecture, the dynamic characteristics of traffic changes, and the need for refined management, and to achieve accurate traffic scheduling on the cloud platform, it is necessary to classify the traffic feature mapping table.

[0042] In this embodiment of the invention, bidirectional feature classification based on positive and negative values ​​is a lightweight solution for traffic management. It can quickly achieve coarse-grained filtering of traffic by using numerical symbols, and upgrade the positive and negative values ​​to intelligent classification driven by business semantics, thereby further reducing the misjudgment rate and improving management accuracy.

[0043] In this embodiment of the invention, classifying the traffic feature mapping table in a multi-cloud environment based on the dynamic network traffic characteristics includes: Write the dynamic network traffic characteristics into the traffic field of the traffic characteristic mapping table in the multi-cloud environment; The dynamic network traffic characteristics corresponding to the traffic field are dynamically refreshed, and a traffic characteristic time chain is generated based on the refreshed dynamic network traffic characteristics. Extract the target network traffic features corresponding to the traffic feature time chain list based on the preset timestamp; Based on the bidirectional feature categories corresponding to the target network traffic characteristics, the traffic feature table in the multi-cloud environment is classified to obtain a positive traffic feature sub-table and a negative traffic feature sub-table.

[0044] In detail, the traffic characteristics collected in real time by the cloud platform are structured and stored in preset fields of the mapping table. For example, dynamic network traffic characteristics correspond to traffic characteristic fields in the mapping table, and the traffic field data is updated according to preset periods, overwriting old data or adding new records to ensure data real-time performance. For example, the traffic characteristic field of a certain cloud platform is refreshed every 10 seconds, recording the current traffic characteristic factors, and the historically refreshed traffic characteristics are linked into a linked list structure in the order of timestamps, such as timestamp 1-characteristic A → timestamp 2-characteristic B → timestamp 3-characteristic C, forming a time series of traffic characteristics, thereby preserving the historical change trajectory of traffic characteristics. Dynamic refresh ensures that the traffic characteristic mapping table always reflects the current network status and avoids classification errors caused by data lag, such as attack traffic ending but old data still being classified as normal.

[0045] Specifically, based on the specified time range according to the analysis requirements, the corresponding traffic characteristic data is located through timestamps. Based on the traffic characteristic data, positive and negative traffic characteristic sub-tables are generated according to bidirectional characteristic categories. The bidirectional characteristic category means that if the network traffic characteristic is positive, it is classified as a positive traffic characteristic sub-table; if the network traffic characteristic is negative, it is classified as a negative traffic characteristic sub-table. That is, the target characteristic of each timestamp is bidirectionally marked. Finally, all positive characteristics are stored in the positive sub-table, and negative characteristics are stored in the negative sub-table, achieving binary management of traffic data. For example, if the network traffic characteristic values ​​of cloud platform A and cloud platform B are positive, then cloud platform A and cloud platform B are classified as positive traffic characteristic sub-tables; if the network traffic characteristic values ​​of cloud platform C and cloud platform D are negative, then cloud platform C and cloud platform D are classified as negative traffic characteristic sub-tables.

[0046] In addition, a three-level classification dimension can be designed based on business value, traffic characteristics, and resource status, with each dimension further subdivided into 3-4 subcategories to achieve personalized classification: First-level dimension: Business Priority (weight 0.4): Core Business (0.8-1.0), Important Business (0.5-0.7), Non-Core Business (0.1-0.4), dynamically adjusted based on business revenue contribution (e.g., e-commerce order priority temporarily increased to 0.9 during Double 11). Second-level dimension: Traffic Sensitivity (weight 0.3): High Sensitivity: Latency ≤ 100ms, Packet Loss Rate ≤ 0.1% (e.g., financial payments, industrial control); Medium Sensitivity: Latency ≤ 500ms, Packet Loss Rate ≤ 1% (e.g., e-commerce product page loading); Low Sensitivity: Latency ≤ 1s, Packet Loss Rate ≤ 5% (e.g., log synchronization, data backup). Level 3 Dimension: Real-time volatility (weight 0.3): High volatility: Standard deviation of traffic in the past 5 minutes > 50Mbps (e.g., flash sales); Medium volatility: Standard deviation 20-50Mbps (e.g., weekday office traffic); Low volatility: Standard deviation < 20Mbps (e.g., nighttime data synchronization). Then, a classification rule driven by reinforcement learning is used to classify the traffic feature mapping table in a multi-cloud environment. The Q-Learning algorithm is used, with the network quality indicator satisfaction rate, resource utilization rate, and energy consumption reduction rate after scheduling as the reward function. The classification dimension weights are optimized in real-time to solve the problem of rule solidification, i.e., the weights are initialized, with the weight allocation as business priority 0.4, traffic sensitivity 0.3, and volatility 0.3; the reward function is R = 0.5 × network quality indicator satisfaction rate + 0.3 × resource utilization rate + 0.2 × (1 - mean of energy consumption reduction rate) (PUE). (The lower the value, the higher the reward). When the network quality index satisfaction rate of a certain category sub-table (such as core-high sensitivity-high volatility) is <90%, the business priority weight is automatically increased to 0.5 and the volatility weight is reduced to 0.2. When the resource utilization rate is <60%, the volatility weight is increased to 0.4, and high volatility traffic is prioritized to fill the resource gap.

[0047] Furthermore, in a multi-cloud environment, the traffic demands of different platforms may vary significantly. For example, the main platform carries core business, while the sub-platforms handle edge traffic. By analyzing demand data, bandwidth resources and load balancing strategies can be dynamically adjusted to avoid resource waste or congestion. For instance, when a platform experiences a surge in positive traffic, more inbound bandwidth can be allocated in a timely manner or a traffic diversion mechanism can be activated.

[0048] In this embodiment of the invention, the traffic demand data refers to the specific indicators of network traffic resources required by each cloud platform to meet the needs of business operation in a multi-cloud environment.

[0049] In this embodiment of the invention, the analysis of traffic demand data for each cloud platform based on the classified traffic feature mapping table includes: The positive traffic data of each cloud platform is determined based on the positive traffic feature sub-table in the classified traffic feature mapping table, and the negative traffic data of each cloud platform is determined based on the negative traffic feature sub-table in the classified traffic feature mapping table. The traffic demand dimension for each cloud platform is determined based on the positive traffic data and the negative traffic data. The traffic demand data for each cloud platform is determined based on the aforementioned traffic demand dimensions.

[0050] In detail, the categorized positive traffic feature sub-table stores records with positive network traffic characteristics. Positive traffic indicates sufficient cloud platform traffic, meaning the actual traffic exceeds a preset threshold or baseline level. The negative traffic feature sub-table stores negative traffic characteristics. Negative traffic indicates insufficient cloud platform traffic, requiring traffic scheduling, meaning the actual traffic is below a preset threshold or baseline level. Based on the positive and negative traffic data, the traffic demand dimension for each cloud platform is determined. This traffic demand dimension includes demand intensity, demand volatility, and demand priority, where demand intensity is represented by the positive traffic value minus the negative traffic value. A positive result indicates sufficient overall traffic, but optimization of excess traffic should be considered. A negative result indicates insufficient overall traffic, requiring priority resource allocation for replenishment. Demand volatility is calculated as the positive traffic standard deviation / positive traffic mean minus the negative traffic standard deviation / negative traffic mean. A larger positive value indicates unstable traffic availability, requiring the reservation of elastic resources. A larger negative value indicates persistent traffic shortage, requiring long-term resource expansion. Demand priority is calculated as business value weight × duration of traffic shortage / resource allocation cost, used to determine the order of traffic allocation. For example, traffic shortages for core businesses should be addressed first.

[0051] Specifically, based on the demand intensity in the traffic demand dimension, the initial traffic demand data of each cloud platform can be determined as negative traffic data, while based on the demand volatility, the reserved traffic data of each cloud platform can be determined. Then, the negative traffic data and the reserved traffic data are superimposed on the traffic demand data of each cloud platform, and the demand priority can determine the order of traffic scheduling of each cloud platform based on the traffic demand data.

[0052] In addition, the Transformer-Time Series (TST) model can be used to predict traffic demand in the next 5 / 10 / 15 minutes by inputting dynamic traffic characteristic factors, thus solving the problem of temporary resource shortage. Using multi-dimensional traffic data from the past 90 days as the training set, it is divided into training / validation / test sets in a 7:2:1 ratio, with the optimization goal of prediction error <8% (the error on the test set is stable at 5-7%). The output is the demand value plus a confidence interval, for example, a payment service traffic demand of 120±6Mbps in the next 10 minutes with a confidence level of 93%. When the predicted demand is greater than 90% of the current reserve, advance reservation is triggered (automatically initiating a resource lock request to the backup cloud platform; the reservation duration = prediction duration + 5 minutes, to avoid resource preemption), thereby determining the traffic demand data for each cloud platform.

[0053] Furthermore, in a multi-cloud environment, the traffic demand of each platform changes dynamically. The scheduling queue can fill the traffic gap in real time, avoiding service lag, interruption or performance degradation due to insufficient traffic. For example, when an e-commerce platform experiences a surge in traffic during a promotion, the scheduling queue can quickly allocate traffic from other platforms or backup resources to ensure business continuity.

[0054] S3. Analyze the network traffic scheduling queue of each cloud platform using the traffic demand data and the backup traffic data in the backup cloud platform under the multi-cloud environment, and perform traffic scheduling for each cloud platform according to the network traffic scheduling queue.

[0055] In this embodiment of the invention, the network traffic scheduling queue refers to a traffic allocation execution list generated in a multi-cloud environment based on the traffic demand (negative traffic) and available resources (positive traffic and backup traffic) of each cloud platform, according to certain rules such as priority, traffic gap size, and service importance. The queue contains the traffic scheduling order, allocation amount, and resource source of each platform, which comes from internal surplus or backup platforms, and is a specific execution scheme for realizing dynamic traffic scheduling.

[0056] In this embodiment of the invention, the step of analyzing the network traffic scheduling queue of each cloud platform using the traffic demand data and backup traffic data in the backup cloud platform under a multi-cloud environment includes: The total demand traffic under the multi-cloud environment is calculated based on the traffic demand data of each platform, and the total traffic surplus under the multi-cloud environment is calculated based on the positive traffic data. When the total demand traffic is less than or equal to the total traffic reserve, a network traffic scheduling sequence for each cloud platform is generated based on the total demand traffic. When the total demand traffic exceeds the total remaining traffic, extract the backup traffic data of the backup cloud platform in the multi-cloud environment; A network traffic scheduling queue for each cloud platform is generated based on the demand traffic data and the backup traffic data.

[0057] In detail, the sum of negative traffic data for all cloud platforms is calculated. Negative traffic represents the traffic gap of each platform, that is, the insufficient traffic. Therefore, the total demand reflects the total amount of traffic that needs to be supplemented in the entire multi-cloud environment. The sum of positive traffic data for all cloud platforms is also calculated. Positive traffic represents the remaining available traffic of each platform. The total surplus reflects the total amount of schedulable traffic resources in the current multi-cloud environment. The demand and surplus are compared to determine the scheduling strategy.

[0058] Specifically, when the total demand traffic is less than or equal to the total traffic reserve, it indicates that the positive traffic in the current multi-cloud environment, i.e., the reserve, is sufficient to meet the negative traffic demand of all platforms, i.e., the traffic gap. There is no need to call backup resources. At this time, a scheduling sequence is generated directly based on the demand traffic of each platform, and the reserve traffic is allocated according to priority or preset rules. When the total demand traffic is greater than the total traffic reserve, it indicates that the current reserve cannot meet all the demand, and additional traffic needs to be obtained from the backup cloud platform. At this time, the backup traffic data of the backup cloud platform, i.e., the pre-reserved redundant traffic resources, is extracted and combined with the existing demand traffic to recalculate the schedulable resources. Then, the demand traffic, reserve traffic, and backup traffic are integrated to determine the traffic scheduling order and allocation of each cloud platform, forming the final scheduling queue.

[0059] In addition, the total predicted demand and real-time surplus of all cloud platforms are calculated. If demand is less than or equal to surplus, the surplus is allocated using a greedy algorithm based on business priority and adaptability coefficient (core and highly sensitive businesses are prioritized), such as allocating 50Mbps to payment services and 40Mbps to e-commerce orders, ensuring that high-value businesses are prioritized. When demand exceeds surplus (e.g., 200Mbps > 180Mbps, resulting in a 20Mbps gap), 3-5 candidate backup cloud nodes are selected (e.g., AWS Tokyo, Alibaba Cloud Shanghai, Huawei Cloud Guangzhou), and optimization objectives are constructed: Min (transmission latency): real-time detection of latency for each node based on edge probes; Min (resource cost): real-time unit price is obtained by calling the cloud platform billing API; Max (resource utilization): nodes with a utilization rate of 60-80% are selected (avoiding resource waste of utilization rate <50% or overload risk of utilization rate >90%). The Pareto optimal solution is solved using NSGA-II, and finally, the Alibaba Cloud Shanghai node with a latency of 30ms, cost of 0.5 yuan / GB, and utilization rate of 75% is selected to fill the 20Mbps gap.

[0060] Furthermore, the initial allocation is from largest to smallest to ensure that platforms with high demand receive resources first. If the allocation is from smallest to largest, there may be a situation where small demand accumulates and exhausts the reserve, leaving no resources available for platforms with high demand. When the main reserve is insufficient, backup traffic is introduced as a second resource pool to avoid scheduling interruption due to the depletion of a single resource pool.

[0061] In this embodiment of the invention, the backup cloud platform is divided into two parts of traffic: backup traffic data and cloud platform fault traffic data. By using the hierarchical allocation of primary spare capacity and backup traffic, dynamic fluctuations in traffic demand can be addressed, and scheduling failures caused by the failure of a single resource pool can be avoided.

[0062] In this embodiment of the invention, to prevent scheduling terminals from occurring during traffic scheduling, an anti-interruption traffic scheduling sequence needs to be generated. The step of generating a network traffic scheduling queue for each cloud platform based on the demand traffic data and the backup traffic data includes: The demand traffic data of each cloud platform is initially sorted in descending order to obtain the first demand traffic sequence. Then, traffic is allocated to each cloud platform one by one according to the total traffic surplus and the first demand traffic sequence. The demand traffic data corresponding to the unallocated cloud platforms is re-sorted to obtain a second demand traffic sequence. Traffic is then allocated to the unallocated cloud platforms one by one according to the backup traffic data and the second demand traffic sequence. Based on the allocated cloud platform statistics, the backup traffic balance corresponding to the backup traffic data is divided according to the required traffic data of each cloud platform and the preset linked list element threshold, so as to obtain the partitioned linked list element of each cloud platform. The network traffic scheduling queue for each cloud platform is formed by dividing the linked list elements according to each cloud platform.

[0063] In detail, the traffic demand data of each cloud platform is sorted from largest to smallest to generate a first traffic demand sequence. Traffic is then allocated sequentially using the total available traffic capacity according to this sequence. For example, if platform A demands 100MB, platform B demands 80MB, and platform C demands 50MB, the order is A→B→C. If the total available capacity is 150MB, then 100MB is allocated to platform A first, the remaining 50MB is allocated to platform B, and no traffic is allocated to platform C. Priority is given to platforms with large traffic demands to avoid the accumulation of small demands consuming available capacity, which could lead to service interruptions for platforms with large demands. For cloud platforms that do not receive traffic, such as platform C, the demand data is re-sorted to generate a second traffic demand sequence, and then allocated sequentially using backup traffic data. If the backup traffic is 100MB and platform C demands 50MB, then 50MB is allocated, leaving 50MB of backup traffic. If the primary capacity is insufficient, backup traffic is used to supplement the remaining capacity, ensuring that the demands of all platforms are at least partially covered, avoiding the risk of service interruption due to zero allocation.

[0064] Specifically, the remaining backup traffic after allocation is calculated. For example, if there is 50MB remaining, the backup traffic is divided into three linked list elements based on the traffic demand of each platform and the preset linked list threshold. For instance, if the linked list element threshold is set to three, the backup traffic is divided into three 30MB linked list elements. By dividing the traffic according to the threshold, the backup traffic is split into the smallest unit that can be flexibly scheduled. This ensures that even if the remaining traffic is insufficient, a minimum traffic allowance can be allocated to each platform according to priority to prevent complete traffic outage. The allocation results of each platform, including the allocation of the main remaining traffic, the allocation of backup traffic, and the linked list elements, are combined into a queue according to the scheduling order. The traffic source, allocation amount, and scheduling priority of each platform are clearly defined. Thus, the network traffic scheduling queue of platform D is 30MB-30MB-30MB-30MB. If the network traffic demand of platform D exceeds 30MB, the traffic in the next position in the scheduling queue will be called to add to the previous 30MB to support the network traffic demand of platform D. In this way, the dynamic fluctuation of traffic demand can be dealt with through the hierarchical calling of the main remaining traffic and backup traffic, and scheduling failure caused by the failure of a single resource pool can be avoided.

[0065] Furthermore, to avoid service interruptions due to momentary network failures and to ensure the success rate of traffic scheduling, a fault-tolerant closed loop for traffic scheduling in a multi-cloud environment is constructed through a multi-layered design of sequential scheduling, retry mechanism, and backup switching.

[0066] In this embodiment of the invention, the dynamic adjustment of the queue head and the writing of backup data enable the cloud platform scheduling to have the adaptive ability to cope with sudden anomalies, and the persistence of failure data supports problem location and strategy optimization, forming a management closed loop of scheduling, monitoring and optimization.

[0067] In this embodiment of the invention, the step of scheduling traffic for each cloud platform according to the network traffic scheduling queue includes: The network traffic data at the head of the network traffic scheduling queue is transmitted to the cloud platform; The cloud platform is polled to check whether the network traffic data has been successfully transmitted by polling the next queue position corresponding to the head of the network traffic scheduling queue. When network traffic data transmission fails, the next queue position corresponding to the head position is updated to the head position, and the process returns to the step of transmitting the network traffic data at the head position in the network traffic scheduling sequence to the cloud platform, until all queue positions in the network traffic scheduling queue have been transmitted to the cloud platform. When all queue positions in the network traffic scheduling queue have been transmitted to the cloud platform, and all network traffic data in the cloud platform fails to be transmitted, the cloud platform is replaced by a backup cloud platform. After successful network traffic data transmission, the untransmitted network traffic data in the network traffic scheduling queue is written to the backup cloud platform, or the failed network traffic data is written to the common traffic field of the traffic feature mapping table in the multi-cloud environment.

[0068] In detail, queue-based scheduling ensures high-priority traffic is transmitted first, avoiding disruptions to critical services due to random scheduling. This involves retrieving traffic data, such as allocated traffic values ​​or linked list elements from the head of the network traffic scheduling queue, and transmitting it to the target cloud platform. For example, if the first element of the queue is "C platform - 20MB", 20MB of traffic will be scheduled from the backup platform to platform C. Network transmission may fail due to momentary congestion or packet loss; polling and retries improve the success rate and prevent scheduling interruptions due to single failures. Therefore, the next queue position after the first element is transmitted is checked to confirm whether the traffic data has successfully reached the target cloud platform, i.e., through heartbeat detection and AC (Acceptance and Reset) checks. The K response mechanism or API callback confirms the transmission status. If the transmission fails, the next queue position is promoted to the head and the transmission is re-executed until all queue positions have been tried. For example, if the transmission of 20MB to platform C fails, the next element, C platform -20MB, is set to the head and the transmission is re-executed. In other words, if a queue position continues to fail to transmit, it is switched to the next position in a timely manner to ensure that other schedulable resources are used first. This avoids blocking the overall scheduling due to a single point of failure and enables fine-grained scheduling of traffic. When the transmission of a linked list element fails, the system can quickly switch to the next element to retry, avoiding the interruption of the entire scheduling process due to a single failure. The retry efficiency is improved by more than 50%.

[0069] Specifically, in extreme cases such as network link failure or server crashes on the main cloud platform, where all elements in the queue fail to transmit, the target cloud platform will be switched to the backup cloud platform to ensure service continuity. If the main cloud platform fails to schedule all traffic, the backup platform will take over the traffic to prevent a complete service interruption. The backup platform must be physically isolated from the main platform to avoid sharing the fault domain. If traffic in the network traffic scheduling queue is successfully transmitted, the untransmitted traffic data will be written to the backup cloud platform as redundancy. If traffic in the network traffic scheduling queue fails to transmit, the failed data will be written to a common field in the traffic feature mapping table after successful transmission. This facilitates subsequent analysis of the failure causes. By analyzing the failure records in the mapping table, firewall rules can be adjusted, network bandwidth can be expanded, or the scheduling queue order can be optimized.

[0070] Furthermore, based on three-level fault tolerance and fault self-diagnosis, the fault recovery time is shortened from seconds to milliseconds. The three-level fault tolerance includes: Level 1 Retry (Rapid Recovery within the Same Region): When transmission at the head of the queue fails (e.g., a sudden increase in packet loss rate > 5% at the AWS Beijing node), other nodes within the same region are prioritized for retry, with ≤ 3 retries and 200ms intervals between each retry. After a successful retry, the fault field in the traffic feature mapping table is updated to prevent subsequent scheduling from selecting the faulty node again. Level 2 Switchover (Cross-Region Backup): After a retry failure, the system immediately switches to the backup node selected in step 3.2, and simultaneously triggers link detection (using traceroute to locate the faulty segment, such as congestion in the Beijing-Shanghai Telecom backbone network), and synchronizes the detection results to the intelligent diagnosis module. Level 3 Degradation (Non-Core Business Rate Limiting): If cross-region switching still fails (e.g., a sudden outage of the Alibaba Cloud Shanghai node), tiered rate limiting will be implemented for non-core businesses (e.g., log synchronization traffic limited to 50%, internal office traffic limited to 30%), releasing resources to prioritize core businesses (e.g., 100% retention of payment business traffic); Degradation will also trigger emergency resource expansion (calling the cloud platform's elastic bandwidth API to temporarily request 100Mbps bandwidth, with an expansion response time of <3 seconds).

[0071] In addition, failure data (failure time, node, business type, link status) is input into the decision tree diagnostic model (training set consists of 100,000 historical failure records) to automatically generate a diagnostic report, for example: Failure type: cross-regional link congestion; Affected business: e-commerce order placement (latency increased to 600ms); Solution: 1. Temporarily switch to Huawei Cloud Guangzhou node; and integrate the remaining small bandwidth resources (such as 10Mbps, 15Mbps) after allocation by each cloud platform into a virtual resource pool, and allocate them to businesses with small demand (such as internal office systems requiring 20Mbps, integrating AWS Beijing 10Mbps + Alibaba Cloud Beijing 10Mbps) to improve resource utilization.

[0072] Furthermore, end-to-end dynamic analysis and scheduling enables on-demand resource allocation and automatic fault migration in multi-cloud environments, reducing manual intervention costs. Intelligent scheduling can reduce maintenance manpower costs by 40% while improving resource scheduling efficiency by more than 30%. Traffic demand analysis based on real-time data can help enterprises expand or shrink cloud resources on demand, avoiding over-purchasing. For example, by dynamically adjusting traffic allocation across platforms, idle cloud resource costs can be reduced by 20%, which is especially suitable for internet business scenarios with large traffic fluctuations.

[0073] This invention dynamically senses fluctuations in business traffic by monitoring the real-time load status and traffic data of each cloud platform. Based on a real-time data-classified traffic feature mapping table, it can more accurately analyze the actual needs of each cloud platform. Network traffic scheduling queues, through priority sorting and hierarchical allocation, ensure efficient matching of demand traffic with available resources. Sorting and hierarchically allocating traffic according to demand from largest to smallest prioritizes resource supply to high-demand platforms, reducing the risk of interruptions to critical services. When all queue positions on a cloud platform fail to transmit, a backup cloud platform is automatically used, completing failover within seconds. The failed traffic data is written to the common traffic field of the traffic feature mapping table for post-failure analysis. Therefore, the network traffic scheduling method, apparatus, device, and medium proposed in this invention, based on a multi-cloud environment, can solve the problem of low accuracy in network traffic scheduling under fixed-level traffic distribution conditions.

[0074] like Figure 3 The diagram shown is a functional block diagram of a network traffic scheduling device based on a multi-cloud environment provided in an embodiment of the present invention.

[0075] The network traffic scheduling device 100 based on a multi-cloud environment described in this invention can be installed in an electronic device. Depending on the functions implemented, the network traffic scheduling device 100 based on a multi-cloud environment may include a dynamic network traffic characteristic analysis module 101, a traffic demand data analysis module 102, and a traffic scheduling module 103. The module described in this invention can also be called a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, and is stored in the memory of the electronic device.

[0076] In this embodiment, the functions of each module / unit are as follows: The dynamic network traffic feature analysis module 101 is used to monitor the real-time load status and real-time traffic data of each cloud platform in a multi-cloud environment, and analyze the dynamic network traffic features of each cloud platform based on the real-time load status and the real-time traffic data. The traffic demand data analysis module 102 is used to classify the traffic feature mapping table in the multi-cloud environment according to the dynamic network traffic characteristics, and analyze the traffic demand data of each cloud platform based on the classified traffic feature mapping table. The traffic scheduling module 103 is used to analyze the network traffic scheduling queue of each cloud platform using the traffic demand data and the backup traffic data in the backup cloud platform in the multi-cloud environment, and to perform traffic scheduling for each cloud platform according to the network traffic scheduling queue.

[0077] Specific limitations regarding the network traffic scheduling device in a multi-cloud environment can be found in the limitations of the network traffic scheduling method in a multi-cloud environment described above, and will not be repeated here. Each module in the aforementioned network traffic scheduling device in a multi-cloud environment can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in the computer device in hardware form, or stored in the memory of the computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0078] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 4 As shown, the computer device includes a processor, memory, network interface, and database connected via a device bus. The processor provides computing and control capabilities. The memory includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage media stores operating devices, computer programs, and a database. The internal memory provides an environment for the operation of the operating devices and computer programs stored in the non-volatile storage media. The network interface is used to communicate with external clients via a network connection. When the computer program is executed by the processor, it implements the functions or steps of a server-side network traffic scheduling method based on a multi-cloud environment.

[0079] In one embodiment, a computer device is provided, which may be a client, and its internal structure diagram may be as follows: Figure 5 As shown, the computer device includes a processor, memory, network interface, display screen, and input devices connected via a device bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores operating devices and computer programs. The internal memory provides an environment for the operation of the operating devices and computer programs stored in the non-volatile storage media. The network interface is used to communicate with an external server via a network connection. When executed by the processor, the computer program implements client-side functions or steps of a network traffic scheduling method based on a multi-cloud environment.

[0080] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps: Monitor the real-time load status and real-time traffic data of each cloud platform in a multi-cloud environment, and analyze the dynamic network traffic characteristics of each cloud platform based on the real-time load status and real-time traffic data. Based on the dynamic network traffic characteristics, the traffic feature mapping table in the multi-cloud environment is classified, and the traffic demand data of each cloud platform is analyzed based on the classified traffic feature mapping table. The network traffic scheduling queue of each cloud platform is analyzed using the traffic demand data and backup traffic data in the backup cloud platform under the multi-cloud environment, and traffic scheduling is performed on each cloud platform according to the network traffic scheduling queue.

[0081] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor: Monitor the real-time load status and real-time traffic data of each cloud platform in a multi-cloud environment, and analyze the dynamic network traffic characteristics of each cloud platform based on the real-time load status and real-time traffic data. Based on the dynamic network traffic characteristics, the traffic feature mapping table in the multi-cloud environment is classified, and the traffic demand data of each cloud platform is analyzed based on the classified traffic feature mapping table. The network traffic scheduling queue of each cloud platform is analyzed using the traffic demand data and backup traffic data in the backup cloud platform under the multi-cloud environment, and traffic scheduling is performed on each cloud platform according to the network traffic scheduling queue.

[0082] It should be noted that the functions or steps that can be implemented by the computer-readable storage medium or computer device described above can be referred to the relevant descriptions on the server side and client side in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0083] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0084] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0085] It should be noted that if any software tools or components not belonging to our company appear in the embodiments of this application, they are merely for illustrative purposes and do not represent actual use.

[0086] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A network traffic scheduling method based on a multi-cloud environment, characterized in that, include: Monitor the real-time load status and real-time traffic data of each cloud platform in a multi-cloud environment, and analyze the dynamic network traffic characteristics of each cloud platform based on the real-time load status and real-time traffic data. Based on the dynamic network traffic characteristics, the traffic feature mapping table in the multi-cloud environment is classified, and the traffic demand data of each cloud platform is analyzed based on the classified traffic feature mapping table. The network traffic scheduling queue of each cloud platform is analyzed using the traffic demand data and backup traffic data in the backup cloud platform under the multi-cloud environment, and traffic scheduling is performed on each cloud platform according to the network traffic scheduling queue.

2. The network traffic scheduling method based on a multi-cloud environment as described in claim 1, characterized in that, The step of analyzing the dynamic network traffic characteristics of each cloud platform based on the real-time load status and the real-time traffic data includes: Extract the real-time load status sequence and real-time traffic data sequence corresponding to the preset time window; The real-time load status sequence is time-aligned with the real-time traffic data sequence to obtain the aligned data timing sequence. Analyze the flow characteristic factors at each time point in the aligned data time series; A dynamic traffic feature sequence for each cloud platform is generated based on the traffic feature factors. The dynamic network traffic characteristics of each cloud platform are determined by the dynamic traffic characteristic sequence.

3. The network traffic scheduling method based on a multi-cloud environment as described in claim 1, characterized in that, Before classifying the traffic feature mapping table in a multi-cloud environment based on the dynamic network traffic characteristics, the method further includes: The mapping table structure for each cloud platform is determined based on pre-defined mapping table field data; The fields of the mapping table structure are standardized to obtain a unified mapping table structure. The common traffic fields in a multi-cloud environment are determined based on the unified mapping table structure. By integrating the unified mapping table structure with the common traffic field, a traffic feature mapping table for a multi-cloud environment is obtained.

4. The network traffic scheduling method based on a multi-cloud environment as described in claim 3, characterized in that, The classification of traffic feature mapping tables in a multi-cloud environment based on the dynamic network traffic characteristics includes: Write the dynamic network traffic characteristics into the traffic field of the traffic characteristic mapping table in the multi-cloud environment; The dynamic network traffic characteristics corresponding to the traffic field are dynamically refreshed, and a traffic characteristic time chain is generated based on the refreshed dynamic network traffic characteristics. Extract the target network traffic features corresponding to the traffic feature time chain list based on the preset timestamp; Based on the bidirectional feature categories corresponding to the target network traffic characteristics, the traffic feature table in the multi-cloud environment is classified to obtain a positive traffic feature sub-table and a negative traffic feature sub-table.

5. The network traffic scheduling method based on a multi-cloud environment as described in claim 4, characterized in that, The analysis of traffic demand data for each cloud platform based on the classified traffic feature mapping table includes: The positive traffic data of each cloud platform is determined based on the positive traffic feature sub-table in the classified traffic feature mapping table, and the negative traffic data of each cloud platform is determined based on the negative traffic feature sub-table in the classified traffic feature mapping table. The traffic demand dimension for each cloud platform is determined based on the positive traffic data and the negative traffic data. The traffic demand data for each cloud platform is determined based on the aforementioned traffic demand dimensions.

6. The network traffic scheduling method based on a multi-cloud environment as described in claim 5, characterized in that, The analysis of the network traffic scheduling queue of each cloud platform using the traffic demand data and backup traffic data in the backup cloud platform under a multi-cloud environment includes: The total demand traffic under the multi-cloud environment is calculated based on the traffic demand data of each platform, and the total traffic surplus under the multi-cloud environment is calculated based on the positive traffic data. When the total demand traffic is less than or equal to the total traffic reserve, a network traffic scheduling sequence for each cloud platform is generated based on the total demand traffic. When the total demand traffic exceeds the total remaining traffic, extract the backup traffic data of the backup cloud platform in the multi-cloud environment; A network traffic scheduling queue for each cloud platform is generated based on the demand traffic data and the backup traffic data.

7. The network traffic scheduling method based on a multi-cloud environment as described in claim 1, characterized in that, The step of generating a network traffic scheduling queue for each cloud platform based on the demand traffic data and the backup traffic data includes: The demand traffic data of each cloud platform is initially sorted in descending order to obtain the first demand traffic sequence. Then, traffic is allocated to each cloud platform one by one according to the total traffic surplus and the first demand traffic sequence. The demand traffic data corresponding to the unallocated cloud platforms is re-sorted to obtain a second demand traffic sequence. Traffic is then allocated to the unallocated cloud platforms one by one according to the backup traffic data and the second demand traffic sequence. Based on the allocated cloud platform statistics, the backup traffic balance corresponding to the backup traffic data is divided according to the required traffic data of each cloud platform and the preset linked list element threshold, so as to obtain the partitioned linked list element of each cloud platform. The network traffic scheduling queue for each cloud platform is formed by dividing the linked list elements according to each cloud platform.

8. A network traffic scheduling device based on a multi-cloud environment, characterized in that, include: The dynamic network traffic characteristic analysis module is used to monitor the real-time load status and real-time traffic data of each cloud platform in a multi-cloud environment, and analyze the dynamic network traffic characteristics of each cloud platform based on the real-time load status and real-time traffic data. The traffic demand data analysis module is used to classify the traffic feature mapping table in the multi-cloud environment according to the dynamic network traffic characteristics, and analyze the traffic demand data of each cloud platform based on the classified traffic feature mapping table. The traffic scheduling module is used to analyze the network traffic scheduling queue of each cloud platform using the traffic demand data and the backup traffic data in the backup cloud platform in the multi-cloud environment, and to perform traffic scheduling for each cloud platform according to the network traffic scheduling queue.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the network traffic scheduling method based on a multi-cloud environment as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the network traffic scheduling method based on a multi-cloud environment as described in any one of claims 1 to 7.

Images