Intelligent edge scheduling system and method for globalized audio and video service

Abstract

The invention discloses an intelligent edge scheduling system and method for globalized audio and video services, and relates to the technical field of audio and video distribution and edge computing. Comprising the following steps: 1, requesting analysis and region locking; step 2, resource filtering and node pre-selection; step 3, network matching and node determination; 4, performing cross-region expansion and degradation scheduling; and 5, connection establishment: returning the finally determined service node information to the user side, and guiding the user to establish audio and video service connection. According to the intelligent edge scheduling system and method oriented to the globalized audio and video service, node scheduling is carried out by comprehensively considering multiple factors such as user geographic positions, ISP information, network link quality, edge node resource conditions and future load trends, the limitation of a traditional scheduling strategy is broken through, and compared with a scheduling method only considering a single factor, the intelligent edge scheduling system and method have the advantages that the scheduling efficiency is improved; according to the method, the node performance can be evaluated more comprehensively and accurately, and the scientificity and rationality of a scheduling decision are improved.

Description

Technical Field

[0001] This invention relates to the field of audio and video distribution and edge computing technology, specifically to an intelligent edge scheduling system and method for global audio and video services. Background Technology

[0002] With the acceleration of global digitalization, audio and video services such as cross-border video conferencing, global live streaming, online education, and remote collaboration have become an important part of business operations and personal communication.

[0003] Early systems typically used centralized data centers or deployed service nodes within a single country / region. When users came from different regions around the world, all requests were directed to the central node. This meant that users far from the data center, especially cross-border users, had to transmit long distances through multiple public autonomous systems and international internet exits, inevitably introducing high network latency, significant jitter, and a high packet loss rate.

[0004] In addition, while some existing edge computing systems attempt to push computing power down to the network edge, their scheduling strategies are mostly based on simple polling or only consider the instantaneous CPU and memory load of nodes, lacking a comprehensive judgment on network link quality and future load trends. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent edge scheduling system and method for global audio and video services, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an intelligent edge scheduling method for global audio and video services, comprising:

[0007] Step 1: Request resolution and area locking. The unified signaling center receives the audio and video service request initiated by the user terminal and resolves the user's IP address in the audio and video service request to obtain its geographical location and network service provider (ISP) information. Based on the geographical location and ISP information, it queries the pre-built leased line topology relationship table to lock one or more candidate target areas. When determining the candidate target areas, the area that is geographically closest to the user and has a direct leased line connection is selected first. If no such area exists, the area that is on the same continent as the user and has deployed BGP nodes is selected.

[0008] Step 2: Resource filtering and node pre-selection. For each candidate target area locked in Step 1, collect real-time resource data and load prediction values ​​based on historical data of each edge node within it. Compare the real-time resource data and load prediction values ​​with preset thresholds to filter out all edge nodes that meet the health requirements and aggregate them to form an available node pool.

[0009] Step 3: Network matching and node determination. In the available node pool obtained in Step 2, nodes are first sorted by network priority, and nodes that are consistent with the user's ISP are selected first. Then, BGP nodes that support multiple ISPs are selected. For nodes with the same priority, their comprehensive score is calculated by a composite weight scoring model, and the node with the highest score is selected as the service node.

[0010] Step 4: Cross-region expansion and degraded scheduling. If, in step 2, no available node pool can be formed in all candidate target areas, cross-region expansion is initiated. That is, according to the leased line topology table, the neighboring area with a leased line connection to the current primary area and the lowest leased line latency is selected as the new scheduling target, and steps 2 and 3 are repeated. If a service node still cannot be selected through cross-region expansion, a global degraded strategy is triggered, and scheduling is performed to a global BGP node.

[0011] Step 5: Connection Establishment. The final service node information is returned to the user's end, guiding the user to establish an audio and video service connection.

[0012] Furthermore, the leased line topology table includes at least the following fields: area identifier, interconnection area identifier, leased line type, leased line bandwidth, real-time latency data, and congestion status identifier;

[0013] The leased line topology table is periodically updated by the system backend from the software-defined networking (SDN) controller, with an update frequency of no less than once every 5 minutes.

[0014] Furthermore, the method for calculating the load prediction value is as follows:

[0015] Collect historical time-series data on CPU utilization, memory utilization, and network bandwidth utilization of edge nodes within the candidate target area over the past 5-minute time window;

[0016] A linear regression algorithm is used to fit the historical time series data to predict the resource utilization values ​​of the node in the next minute.

[0017] If the predicted value of any resource utilization exceeds its set safety threshold, the node will be deemed unavailable for that scheduling cycle.

[0018] Furthermore, the calculation formula for the composite weighted scoring model is as follows:

[0019]

[0020] Where Ccur represents the current CPU utilization of the node, Cthr represents the CPU utilization threshold, RTTcur represents the current network round-trip delay between the node and the user, RTTthr represents the delay threshold configured for the service type, α and β are the CPU weight factor and delay weight factor, respectively, β>α, and α+β=1.

[0021] Furthermore, the latency threshold RTTthr is dynamically set according to the type of audio and video service. For real-time interactive services, it is set to 90-110ms, and for live streaming services, it is set to 250-350ms. The initial values ​​of α and β are 0.4 and 0.6, respectively, and are periodically adjusted by the system's dynamic optimization module based on the correlation analysis results of the service quality data over the past 8-12 minutes.

[0022] Furthermore, the cross-regional expansion specifically includes:

[0023] When there are no available nodes in the target area, all neighboring areas connected to the target area by dedicated lines are identified as candidate suboptimal areas based on the dedicated line topology table.

[0024] The candidate suboptimal regions are sorted in ascending order of dedicated line delay and in descending order of the number of available nodes within the region.

[0025] In the sorted order, the resource filtering and network matching phases are repeated in the top-ranked suboptimal regions until a usable node is found.

[0026] If no available nodes are found in any of the candidate suboptimal regions, the global BGP node scheduling process is triggered.

[0027] Furthermore, the method further includes the following step between step one and step two:

[0028] Query the global data compliance region mapping table to determine the compliant regions in which a user's data is allowed to be transmitted and processed, based on the user's geographic location identifier;

[0029] The candidate target area is compared with the compliant area range. If the candidate target area is not within the compliant range, the area is removed from the candidate list and the area locking operation is re-executed to ensure that the finally selected service node is within the compliant area.

[0030] An intelligent edge scheduling system for global audio and video services, comprising:

[0031] The unified signaling center is used to process user requests, execute scheduling decisions, and perform global state management.

[0032] A distributed edge network consists of clusters of edge nodes deployed in multiple regions around the world, used to carry media streaming data for audio and video services;

[0033] The data storage module is used to store the leased line topology table, the global data compliance region mapping table, edge node resource data, and historical link quality data;

[0034] The dynamic optimization module includes a load prediction unit, a weight adjustment unit, and a leased line status monitoring unit, which are used to perform load prediction, dynamic adjustment of scheduling weights, and leased line status monitoring.

[0035] Furthermore, the unified signaling center includes at least:

[0036] The user identifier resolution unit is used to extract and resolve the IP address from the user request to obtain the geographical location and ISP information;

[0037] The room management unit is used to maintain the lifecycle of audio and video rooms and the status of users within the rooms;

[0038] The scheduling decision unit is used to generate node scheduling instructions;

[0039] The compliance verification unit is used to call the data storage module during the scheduling process to constrain the data compliance of scheduling decisions.

[0040] Furthermore, each edge node in the distributed edge network is deployed with at least:

[0041] The resource acquisition unit is used to periodically collect and report the CPU, memory, bandwidth utilization, and GPU resource usage of this node.

[0042] The link detection unit is used to actively send probe packets to the user end or a specific target to detect and measure the RTT, packet loss rate and jitter indicators of the network link.

[0043] The cascaded forwarding unit is used to select the optimal path between edge nodes in different regions based on the leased line topology table and to relay media streams across regions using a private encryption protocol.

[0044] Compared with the prior art, the beneficial effects of the present invention are:

[0045] This intelligent edge scheduling system and method for global audio and video services comprehensively considers multiple factors such as user geographical location, ISP information, network link quality, edge node resource status, and future load trends for node scheduling. It breaks through the limitations of traditional scheduling strategies. Compared with scheduling methods that only consider a single factor, this method can more comprehensively and accurately evaluate node performance and improve the scientificity and rationality of scheduling decisions.

[0046] By using a linear regression algorithm based on historical data to predict the future resource utilization of edge nodes, we can understand the node load in advance and make scheduling decisions more forward-looking. Compared with scheduling methods that rely solely on instantaneous load, the dynamic load prediction method can make the resource utilization of nodes more balanced during peak business periods, reduce business interruptions caused by overload, and improve system stability.

[0047] Based on the key information in the user request and combined with the leased line topology table, suitable candidate target areas can be accurately identified, reducing the physical distance of data transmission and lowering network latency. Attached Figure Description

[0048] Figure 1 This is a flowchart of the method of the present invention;

[0049] Figure 2 This is a schematic diagram of the system structure of the present invention;

[0050] Figure 3 This is a first schematic diagram of the user interface of the present invention;

[0051] Figure 4 This is a second schematic diagram of the user interface of the present invention. Detailed Implementation

[0052] 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 embodiments of the present invention, and not all embodiments. 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.

[0053] like Figure 1 , Figure 3 and Figure 4 As shown, this invention provides a technical solution: an intelligent edge scheduling method for global audio and video services, comprising:

[0054] Step 1: Request parsing and area locking. This is executed by the unified signaling center. It receives audio and video service requests from user terminals, parses the key information within, and then uses this information to lock suitable candidate target areas from a pre-built leased line topology table. This lays the foundation for subsequent resource selection and node scheduling. Specifically, it includes:

[0055] S11. Request Reception and Preliminary Processing: As the core hub of the entire scheduling system, the unified signaling center possesses powerful request reception and processing capabilities. It can receive audio and video service requests initiated by users worldwide in real time and stably. These requests are typically transmitted in a specific data packet format, containing information related to the user-initiated service. Upon receiving a request, the signaling center first performs preliminary verification and parsing of the request data to ensure its integrity and accuracy, preparing for subsequent detailed parsing.

[0056] S12. User IP Address Resolution: After completing the initial processing of the request, the unified signaling center will perform in-depth analysis of the user IP address in the audio and video service request. The IP address is the unique identifier of a user on the Internet. Through professional IP address resolution technology, the user's geographical location information can be obtained from the IP address, accurate to the city or even a smaller area. At the same time, the network service provider (ISP) used by the user can also be determined, such as China Mobile, China Unicom, and AT&T in the United States. The accuracy of this step is crucial for subsequent area locking and node scheduling, because users in different geographical locations and with different ISPs have different needs and expectations for network latency, bandwidth and other performance indicators.

[0057] S13. Query the leased line topology table. Based on the parsed user geographical location and ISP information, the unified signaling center will query the pre-built leased line topology table. This table is an important data foundation for the entire scheduling system. It records in detail the leased line connections between various regions. The leased line topology table is periodically updated by the system backend from the software-defined network (SDN) controller, with an update frequency of no less than once every 5 minutes. The leased line topology table must contain at least the following key fields:

[0058] **Regional Identifier:** Used to uniquely identify different geographical regions, such as "Shanghai Region" or "New York Region"; **Interconnection Region Identifier:** Indicates other regions connected to this region via leased lines. This field clearly shows the network connectivity between regions; **Leased Line Type:** Records the specific type of leased line, such as fiber optic leased line or microwave leased line. Different types of leased lines may differ in transmission performance and stability; **Leased Line Bandwidth:** Specifies the bandwidth capacity of the leased line, measured in Mbps or Gbps. It directly affects the speed and efficiency of data transmission; **Real-Time Latency Data:** Reflects the real-time network latency between the regions at both ends of the leased line, measured in milliseconds. Lower latency means more timely data transmission and a better user experience; **Congestion Status Identifier:** Used to identify the current level of congestion on the leased line, generally categorized into different levels such as uncongested, lightly congested, moderately congested, and heavily congested, helping the scheduling system understand the real-time load of the leased line.

[0059] S14. After querying the leased line topology table, the unified signaling center will lock one or more candidate target areas according to certain rules. The specific rules are as follows:

[0060] Prioritizing regions geographically closest to the user and with dedicated direct lines minimizes physical data transmission distance, reduces network latency, and improves the real-time performance and smoothness of audio and video services. For example, if a user is located in Shanghai, the system will prioritize regions geographically closest to Shanghai with dedicated direct lines, such as Hangzhou and Nanjing, as candidate regions. If no such region exists, the system will select regions on the same continent as the user that have deployed BGP nodes. If no geographically close region with a dedicated direct line cannot be found, the system will consider routing the user to a region on the same continent with deployed BGP nodes. BGP nodes support multiple ISPs, improving network compatibility and reliability.

[0061] For example, suppose a user located in Beijing, China, initiates an audio / video service request using China Unicom's network. After receiving the request, the unified signaling center first resolves the user's IP address, determining that their geographical location is Beijing and their ISP is China Unicom. Then, the system queries the dedicated line topology table and finds that the regions geographically closest to Beijing and with direct dedicated line connections include Tianjin and Shijiazhuang. The system will prioritize these regions as candidate target regions. If, during subsequent scheduling, suitable services cannot be provided to the user due to certain reasons (such as insufficient edge node resources in these regions), the system will further expand the search scope. Since Beijing is located in Asia, the system will search for other regions in Asia with deployed BGP nodes, such as Tokyo, Japan, and Seoul, South Korea, and include these regions in the candidate target region list. Subsequent resource screening and node scheduling operations will continue to ensure that the user can successfully establish an audio / video service connection and obtain a good service experience.

[0062] Step Two: Resource Filtering and Node Pre-selection. For each candidate target area identified in Step One, the system will collect real-time resource data of each edge node within it and combine it with load prediction values ​​based on historical data. Through detailed comparison with preset thresholds, edge nodes that meet the health requirements are accurately selected and aggregated to form a pool of available nodes, providing high-quality candidates for the final selection of subsequent nodes. Specifically, this includes:

[0063] S21. Real-time resource data collection: For each locked candidate target area, the system will initiate a real-time resource data collection mechanism. By establishing stable data communication channels with each edge node, the system can obtain the key resource usage of the edge nodes in real time, mainly including the following aspects:

[0064] CPU utilization: This reflects the workload of the edge node's central processing unit. Excessively high or low CPU utilization can affect the processing efficiency of audio and video services. For example, when CPU utilization consistently approaches 100%, the node may be unable to process audio and video data encoding and decoding operations in a timely manner, leading to video stuttering or audio delay. Memory utilization: This reflects the usage status of the edge node's memory resources. Insufficient memory will cause the system to frequently perform memory swapping operations, increasing data processing latency and affecting the smoothness of audio and video services. For example, when processing high-definition video, if memory utilization is too high, problems such as blurry images or screen tearing may occur. Network bandwidth utilization: This measures the network bandwidth usage of the edge node. Network bandwidth is a critical channel for audio and video data transmission. Excessively high network bandwidth utilization may lead to data transmission congestion, causing audio and video stuttering, packet loss, and other phenomena. For example, when network bandwidth utilization is close to saturation, users may frequently encounter slow video loading or interruptions.

[0065] S22. Load Forecast Calculation: To more accurately assess the load situation of edge nodes over a future period, the system adopts a load forecasting method based on historical data, specifically including:

[0066] Historical Time-Series Data Acquisition: This feature collects historical time-series data on CPU utilization, memory utilization, and network bandwidth utilization of edge nodes within the candidate target area over a 5-minute time window. This data records the changing trends of node resource usage in time series form, providing rich information for subsequent predictions. For example, the system records the CPU utilization of each node every 10 seconds, recording 30 sets of data within 5 minutes. Linear Regression Algorithm Fitting: A linear regression algorithm is used to fit the historical time-series data of CPU utilization, memory utilization, and network bandwidth utilization. For example, for historical time-series data of CPU utilization, the linear regression algorithm finds an optimal fitting line that minimizes the sum of squared errors between the line and the actual data points. Future Resource Utilization Prediction: Based on the fitted linear regression model, the system predicts the resource utilization values ​​of the node in the next minute. This prediction helps the system understand the node's load in advance, enabling more reasonable scheduling decisions. For example, through linear regression model prediction, the system can determine that a node's CPU utilization may reach 85% in the next minute.

[0067] S23. Preset Threshold Comparison and Node Filtering: A detailed comparison of real-time resource data and load prediction values ​​with preset thresholds is a crucial step in filtering edge nodes that meet health requirements. The preset thresholds are set based on the characteristics and requirements of audio and video services, combined with the performance indicators of edge nodes, and mainly include the following aspects:

[0068] Safety threshold settings: Set safety thresholds for CPU utilization, memory utilization, and network bandwidth utilization. These thresholds are important criteria for determining whether a node is available. For example, the safety threshold for CPU utilization is set to 80%, the safety threshold for memory utilization is set to 70%, and the safety threshold for network bandwidth utilization is set to 75%. Node availability determination: If the predicted value of any resource utilization exceeds its set safety threshold, the node will be determined as unavailable in that scheduling cycle. For example, if the predicted value of a node's CPU utilization is 82%, exceeding the preset 80% safety threshold, even if the node's memory utilization and network bandwidth utilization are within the safe range, the system will still determine the node as unavailable and exclude it from the available node pool. Only when the node's real-time resource data and load prediction value do not exceed the preset threshold will the node be considered to meet the health requirements and included in the available node pool.

[0069] S24. The available node pool is formed by aggregating all edge nodes that meet the health requirements. The available node pool provides a wealth of candidate resources for the final selection of subsequent nodes. The system can further filter the edge nodes that are most suitable for providing services to users based on the user's specific needs and business characteristics. For example, in the available node pool, the system can comprehensively evaluate factors such as the node's geographical location, network quality, and remaining resources to select the optimal node to provide audio and video services to users.

[0070] For example, suppose that in step one, the Shanghai area was selected as the candidate target area. There are three edge nodes in the Shanghai area, namely node A, node B, and node C. First, regarding the real-time resource data collection: the system collects the following real-time data: node A has a CPU utilization of 70%, memory utilization of 60%, and network bandwidth utilization of 65%; node B has a CPU utilization of 85%, memory utilization of 55%, and network bandwidth utilization of 70%; and node C has a CPU utilization of 65%, memory utilization of 72%, and network bandwidth utilization of 60%.

[0071] Secondly, load prediction values ​​are calculated: by performing linear regression fitting and prediction on historical time-series data from the past 5 minutes, the predicted CPU utilization for node A in the next minute is 72%, memory utilization is 62%, and network bandwidth utilization is 67%; the predicted CPU utilization for node B in the next minute is 88%, memory utilization is 58%, and network bandwidth utilization is 72%; the predicted CPU utilization for node C in the next minute is 67%, memory utilization is 75%, and network bandwidth utilization is 62%.

[0072] Finally, a preset threshold comparison and node selection are performed: Assuming the safe threshold for CPU utilization is 80%, the safe threshold for memory utilization is 70%, and the safe threshold for network bandwidth utilization is 75%, the predicted resource utilization values ​​of node A do not exceed the safe thresholds and meet the health requirements; the predicted CPU utilization value of node B exceeds the safe threshold and is judged as unavailable; the predicted memory utilization value of node C exceeds the safe threshold and is also judged as unavailable; available node pool aggregation: Ultimately, only node A is included in the available node pool, and the system will subsequently further select the optimal node from node A to provide audio and video services to users.

[0073] Step 3: Network Matching and Node Determination. From the available node pool obtained in Step 2, nodes are first sorted by network priority, prioritizing nodes consistent with the user's ISP, followed by BGP nodes supporting multiple ISPs. For nodes with the same priority, a composite weighted scoring model is used to calculate their overall score, and the node with the highest score is selected as the service node. The formula for the composite weighted scoring model is:

[0074]

[0075] Here, Ccur represents the current CPU utilization of the node, reflecting the node's current computing resource usage. Excessive CPU utilization may lead to a decrease in the node's processing capacity, affecting service performance. Cthr represents the CPU utilization threshold, a critical value for judging whether the node's CPU resources are sufficient. When Ccur is greater than Cthr, it indicates that the node's CPU resources are strained and may not be able to meet service demands. RTTcur represents the current network round-trip latency between the node and the user, measuring the time required for data transmission between the node and the user. Excessive latency can cause stuttering and delays in audio and video services, affecting user experience. RTTthr represents the latency threshold configured for the service type. α and β are the CPU weight factor and latency weight factor, respectively, where β > α and satisfies α + β = 1. In this solution, the latency threshold RTTthr is dynamically set according to the type of audio and video service. For real-time interactive services, it is set to 90-110ms; for live streaming services, it is set to 250-350ms. The initial values ​​of α and β are 0.4 and 0.6, respectively, and are periodically adjusted by the system's dynamic optimization module based on the correlation analysis results of service quality data over the past 8-12 minutes.

[0076] For example, suppose in an audio / video service scenario, after resource filtering and node pre-selection in step two, a pool of available nodes with 5 available nodes is obtained. Among these 5 nodes, 2 nodes have the same ISP as the user, and 3 nodes are BGP nodes that support multiple ISPs. First, according to network priority, the 2 nodes with the same ISP as the user are ranked first, and the 3 BGP nodes are ranked last. For the 2 nodes with the same ISP as the user, suppose node A has a current CPU utilization CcurA = 30%, a CPU utilization threshold Cthr = 80%, a current network round-trip latency RTTcurA = 80ms, a service type of real-time interactive service, a latency threshold RTTthr = 100ms, and α and β are 0.4 and 0.6 respectively. According to the composite weight scoring model, the comprehensive score of node A is calculated as: SA node = 0.25 + 0.12 = 0.37;

[0077] Assuming node B's current CPU utilization CcurB = 40%, CPU utilization threshold Cthr = 80%, current network round-trip latency RTTcurB = 90ms, service type is real-time interactive service, latency threshold RTTthr = 100ms, and α and β are 0.4 and 0.6 respectively, the comprehensive score of node B calculated using the composite weighted scoring model is: SB_node = 0.2 + 0.06 = 0.26. Since node A's comprehensive score is higher than node B's, node A is preferentially selected as the service node among nodes consistent with the user's ISP. If node A becomes unavailable for some reason later, node B and BGP nodes will be considered. For BGP nodes, the comprehensive score is calculated and sorted according to the above method to determine the final service node.

[0078] Step 4: Cross-regional expansion and degraded scheduling. During the service node scheduling process for audio and video services, if Step 2 fails to form an available node pool in the primary candidate target area, it means that the edge nodes in the current area cannot meet the service's requirements for resource health and network performance. At this time, the system will initiate a cross-regional expansion mechanism. By utilizing the leased line topology table, it will find the neighboring area with a leased line connection to the current primary area and the lowest leased line latency as the new scheduling target, and re-execute the resource filtering and network matching process. If a suitable service node still cannot be selected after cross-regional expansion, the system will trigger a global degraded strategy and finally schedule to a global BGP node to ensure service continuity and basic service quality.

[0079] Specifically, it includes:

[0080] S41. Cross-region expansion start condition judgment: After step two is completed, the system will check the formation of available node pools in all candidate target areas. If it is found that no available node pools have been formed in all candidate target areas, that is, there are no edge nodes that meet the health requirements in the current area, the system will immediately start the cross-region expansion process. This judgment mechanism ensures that when local resources are insufficient, the search range can be expanded in time to find available resources in other areas.

[0081] S42. Candidate Second-Best Region Determination: The system uses a pre-built leased line topology table, which details the leased line connections and latency information between various regions, to identify all neighboring regions with leased line connections to the current primary region. These regions are then designated as candidate second-best regions. Leased line connections provide relatively stable and low-latency network channels, ensuring better data transmission quality for audio and video services compared to public network connections. The candidate second-best regions are sorted in ascending order of leased line latency and descending order of the number of available nodes within the region. Ascending order of leased line latency prioritizes neighboring regions with the lowest network latency to the current primary region, reducing the number of potential second-best regions. Based on transmission time cost, the descending order of available nodes within a region considers selecting regions with more available nodes when latency is similar, increasing the probability of finding a suitable service node. For example, if two neighboring regions A and B are connected to the primary region by dedicated lines, region A has a dedicated line latency of 20ms and 5 available nodes, while region B has a dedicated line latency of 30ms and 8 available nodes. According to the sorting rules, region A is ranked first because of its lower dedicated line latency. However, if region A does not have available nodes after resource filtering and network matching, the system will continue to consider region B, which has more available nodes and may find a suitable node there.

[0082] S43. According to the sorting order, the resource filtering and network matching stages are repeatedly executed in the second-best areas with the highest sorting. The resource filtering stage collects the real-time resource data of each edge node in the area and the load prediction value based on historical data, and compares it with the preset threshold to filter out the edge nodes that meet the health requirements and form an available node pool. The network matching stage selects the most suitable node to provide services to users from the available node pool according to the network priority and composite weight scoring model.

[0083] S44. Global degradation strategy triggered, specific situations include:

[0084] Handling situations where no available nodes are found: If no available nodes are found in all candidate suboptimal regions, that is, if a suitable service node cannot be selected even after cross-region expansion, the system will trigger a global degradation strategy and schedule the service to a global BGP node. Global BGP nodes have extensive network coverage capabilities and can provide services to users through multi-ISP connections. Although their network performance may not be as good as local or leased line connected nodes, they can serve as a fallback solution to ensure the basic operation of the business when no better node can be found.

[0085] Global BGP Node Scheduling Process: After the global BGP node scheduling process is triggered, the system selects a node from the global BGP node pool as the service node. During the selection process, factors such as the node's geographical location and current load may be considered to optimize service performance. For example, a global BGP node that is relatively close to the user and has a low load may be selected to reduce network latency and ensure service stability.

[0086] Specifically, in this solution, assuming a large audio / video service platform, the primary candidate target area is Region X. During step two, it is found that all edge nodes within Region X cannot form a usable node pool due to excessive CPU utilization or network latency exceeding a threshold. The system immediately initiates a cross-region expansion process. Based on the dedicated line topology table, three neighboring regions connected to Region X via dedicated lines are identified: Region A, Region B, and Region C. The dedicated line topology table shows that the dedicated line latency between Region A and Region X is 15ms, with 3 usable nodes within that region; the dedicated line latency between Region B and Region X is 25ms, with [missing information - likely a number of nodes]. The system has 6 nodes; the leased line latency between Region C and Region X is 35ms, and there are 4 available nodes in each region. These nodes are sorted in ascending order of leased line latency and descending order of available nodes, resulting in Region A, Region B, and Region C. The system first repeats steps two and three in Region A to collect real-time resource data and load prediction values ​​for each edge node. It finds that the CPU utilization prediction value of one node exceeds the threshold, and the network round-trip latency of another node exceeds the latency threshold configured for the audio / video service. After filtering, no available node pool is formed in Region A. The system then performs the same operation in Region B. During the resource filtering phase, four edge nodes meeting the health requirements are selected, forming an available node pool. In the network matching phase, based on network priority and a composite weight scoring model, one node is selected as the service node. This node's ISP is consistent with some users, and it has the highest combined score for CPU utilization and network round-trip latency. If no available node is found in Region B, the system will continue to try in Region C. If region C also cannot provide available nodes, the system will trigger a global degradation strategy, selecting a suitable node from the global BGP node pool to provide services to users, ensuring the continuity of audio and video services.

[0087] Step 5: Connection Establishment. The final service node information is returned to the user's end, guiding the user to establish an audio and video service connection.

[0088] Additionally, the method includes the following steps between steps one and two:

[0089] The system queries the global data compliance region mapping table to determine the compliant region range in which the user's data is allowed to be transmitted and processed based on the user's geographic location identifier. The system compares the candidate target region with the compliant region range. If the candidate target region is not within the compliant range, the region is removed from the candidate list, and the region locking operation is re-executed to ensure that the finally selected service node is within the compliant region.

[0090] like Figure 2 As shown, this solution also discloses an intelligent edge scheduling system for global audio and video services, used to implement an intelligent edge scheduling method for global audio and video services, including:

[0091] A unified signaling center is used to process user requests, execute scheduling decisions, and perform global state management. A unified signaling center includes at least the following:

[0092] The user identifier resolution unit is used to extract and resolve the IP address from the user request to obtain the geographical location and ISP information;

[0093] The room management unit is used to maintain the lifecycle of audio and video rooms and the status of users within the rooms;

[0094] The scheduling decision unit is used to generate node scheduling instructions;

[0095] The compliance verification unit is used to call the data storage module during the scheduling process to constrain the data compliance of scheduling decisions;

[0096] A distributed edge network consists of clusters of edge nodes deployed in multiple regions around the world, used to carry media streaming data for audio and video services;

[0097] The data storage module is used to store the leased line topology table, the global data compliance region mapping table, edge node resource data, and historical link quality data;

[0098] The dynamic optimization module includes a load prediction unit, a weight adjustment unit, and a leased line status monitoring unit, which are used to perform load prediction, dynamic adjustment of scheduling weights, and leased line status monitoring.

[0099] A single edge node in a distributed edge network must have at least the following deployed:

[0100] The resource acquisition unit is used to periodically collect and report the CPU, memory, bandwidth utilization, and GPU resource usage of this node.

[0101] The link detection unit is used to actively send probe packets to the user end or a specific target to detect and measure the RTT, packet loss rate and jitter indicators of the network link.

[0102] The cascaded forwarding unit is used to select the optimal path between edge nodes in different regions based on the leased line topology table, and to relay media streams across regions using a proprietary encryption protocol. All of the above content is integrated into a single disclosure document, with added content and data.

[0103] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended embodiments and their equivalents.

Claims

1. An intelligent edge scheduling method for global audio and video services, characterized in that, include: Step 1: Request resolution and area locking. The unified signaling center receives the audio and video service request initiated by the user terminal and resolves the user's IP address in the audio and video service request to obtain its geographical location and network service provider (ISP) information. Based on the geographical location and ISP information, it queries the pre-built leased line topology relationship table to lock one or more candidate target areas. When determining the candidate target areas, the area that is geographically closest to the user and has a direct leased line connection is selected first. If no such area exists, the area that is on the same continent as the user and has deployed BGP nodes is selected. Step 2: Resource filtering and node pre-selection. For each candidate target area locked in Step 1, collect real-time resource data and load prediction values ​​based on historical data of each edge node within it. Compare the real-time resource data and load prediction values ​​with preset thresholds to filter out all edge nodes that meet the health requirements and aggregate them to form an available node pool. Step 3: Network matching and node determination. In the available node pool obtained in Step 2, nodes are first sorted by network priority, and nodes that are consistent with the user's ISP are selected first. Then, BGP nodes that support multiple ISPs are selected. For nodes with the same priority, their comprehensive score is calculated by a composite weight scoring model, and the node with the highest score is selected as the service node. Step 4: Cross-region expansion and degraded scheduling. If, in step 2, no available node pool can be formed in all candidate target areas, cross-region expansion is initiated. That is, according to the leased line topology table, the neighboring area with a leased line connection to the current primary area and the lowest leased line latency is selected as the new scheduling target, and steps 2 and 3 are repeated. If a service node still cannot be selected through cross-region expansion, a global degraded strategy is triggered, and scheduling is performed to a global BGP node. Step 5: Connection Establishment. The final service node information is returned to the user's end, guiding the user to establish an audio and video service connection.

2. The intelligent edge scheduling method for global audio and video services according to claim 1, characterized in that, The leased line topology table contains at least the following fields: area identifier, interconnection area identifier, leased line type, leased line bandwidth, real-time latency data, and congestion status identifier; The leased line topology table is periodically updated by the system backend from the software-defined networking (SDN) controller, with an update frequency of no less than once every 5 minutes.

3. The intelligent edge scheduling method for global audio and video services according to claim 1, characterized in that, The method for calculating the load prediction value is as follows: Collect historical time-series data on CPU utilization, memory utilization, and network bandwidth utilization of edge nodes within the candidate target area over the past 5-minute time window; A linear regression algorithm is used to fit the historical time series data to predict the resource utilization values ​​of the node in the next minute. If the predicted value of any resource utilization exceeds its set safety threshold, the node will be deemed unavailable for that scheduling cycle.

4. The intelligent edge scheduling method for global audio and video services according to claim 1, characterized in that, The calculation formula for the composite weighted scoring model is as follows: Where Ccur represents the current CPU utilization of the node, Cthr represents the CPU utilization threshold, RTTcur represents the current network round-trip delay between the node and the user, RTTthr represents the delay threshold configured for the service type, α and β are the CPU weight factor and delay weight factor, respectively, β>α, and α+β=1.

5. The intelligent edge scheduling method for global audio and video services according to claim 4, characterized in that, The latency threshold RTTthr is dynamically set according to the type of audio and video service. For real-time interactive services, it is set to 90-110ms, and for live streaming services, it is set to 250-350ms. The initial values ​​of α and β are 0.4 and 0.6, respectively, and are periodically adjusted by the system's dynamic optimization module based on the correlation analysis results of the service quality data over the past 8-12 minutes.

6. The intelligent edge scheduling method for global audio and video services according to claim 1, characterized in that, The cross-regional expansion specifically includes: When there are no available nodes in the target area, all neighboring areas connected to the target area by dedicated lines are identified as candidate suboptimal areas based on the dedicated line topology table. The candidate suboptimal regions are sorted in ascending order of dedicated line delay and in descending order of the number of available nodes within the region. In the sorted order, the resource filtering and network matching phases are repeated in the top-ranked suboptimal regions until a usable node is found. If no available nodes are found in any of the candidate suboptimal regions, the global BGP node scheduling process is triggered.

7. The intelligent edge scheduling method for global audio and video services according to claim 1, characterized in that, Between step one and step two of the method, the following is also included: Query the global data compliance region mapping table to determine the compliant regions in which a user's data is allowed to be transmitted and processed, based on the user's geographic location identifier; The candidate target area is compared with the compliant area range. If the candidate target area is not within the compliant range, the area is removed from the candidate list and the area locking operation is re-executed to ensure that the finally selected service node is within the compliant area.

8. An intelligent edge scheduling system for global audio and video services, used to implement the intelligent edge scheduling method for global audio and video services as described in any one of claims 1-7, characterized in that, include: The unified signaling center is used to process user requests, execute scheduling decisions, and perform global state management. A distributed edge network consists of clusters of edge nodes deployed in multiple regions around the world, used to carry media streaming data for audio and video services; The data storage module is used to store the leased line topology table, the global data compliance region mapping table, edge node resource data, and historical link quality data; The dynamic optimization module includes a load prediction unit, a weight adjustment unit, and a leased line status monitoring unit, which are used to perform load prediction, dynamic adjustment of scheduling weights, and leased line status monitoring.

9. The intelligent edge scheduling system for global audio and video services according to claim 8, characterized in that, The unified signaling center includes at least: The user identifier resolution unit is used to extract and resolve the IP address from the user request to obtain the geographical location and ISP information; The room management unit is used to maintain the lifecycle of audio and video rooms and the status of users within the rooms; The scheduling decision unit is used to generate node scheduling instructions; The compliance verification unit is used to call the data storage module during the scheduling process to constrain the data compliance of scheduling decisions.

10. The intelligent edge scheduling system for global audio and video services according to claim 8, characterized in that, A single edge node in the distributed edge network is deployed with at least: The resource acquisition unit is used to periodically collect and report the CPU, memory, bandwidth utilization, and GPU resource usage of this node. The link detection unit is used to actively send probe packets to the user end or a specific target to detect and measure the RTT, packet loss rate and jitter indicators of the network link. The cascaded forwarding unit is used to select the optimal path between edge nodes in different regions based on the leased line topology table and to relay media streams across regions using a private encryption protocol.

Images