Multi-operator remote session path adaptive arrangement method, system, medium and device

Abstract

The invention relates to a multi-operator remote session path adaptive arrangement method, system, medium and device, which are suitable for a network architecture comprising a client, an edge relay node and a control plane. The method comprises the following steps: converging candidate paths of multiple operators, and carrying out low-disturbance quality and cost measurement on the paths; identifying the service type of the current remote session, and calculating a multi-target score in combination with the quality sub-score, the cost sub-score and the risk penalty term; selecting a main path and a backup path according to a scoring result, establishing a preheating connection, and starting parallel mirroring or fragmentation drainage when the score falls to a threshold value or the budget is limited, so as to realize main-backup switching or degradation backspacing; in the scoring process, a budget penalty factor is introduced to limit the use of a high-price path, and a small-flow exploration mechanism is set to carry out steady-state evaluation and strategy adjustment. According to the scheme, path dynamic scheduling, cost controlled selection and rapid non-inductive switching of the remote session in a multi-link and multi-tenant environment can be realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of network communication and path scheduling optimization, and particularly relates to a multi-operator remote session path adaptive arrangement method and system fusing quality and cost constraints, a device and a computer readable storage medium. BACKGROUND

[0002] With the popularity of remote office, remote operation and multimedia interaction scenarios, remote sessions are increasingly used in various distributed applications. To improve the flexibility and availability of network connections, more and more systems support multi-operator and multi-link access, and use dynamic path selection to improve service quality. However, the existing path selection scheme still has the following technical deficiencies: Firstly, traditional path arrangement methods are mostly based on static strategies or single index evaluation, and it is difficult to consider multiple quality indicators such as latency, jitter, packet loss rate and throughput capacity at the same time, resulting in a lag in path switching response and unstable business experience when the network state fluctuates.

[0003] Secondly, most existing schemes lack systematic modeling of cost factors, and fail to introduce cost information such as unit traffic cost, budget quota and cross-region additional cost into the scheduling logic, which may cause resource waste or budget overrun in multi-tenant or cross-region deployment scenarios.

[0004] Thirdly, when the main path is abnormal or mutated, the existing switching strategy mostly uses reconnection or passive retry, which cannot achieve millisecond-level seamless switching and may cause remote interaction interruption, screen freezing or batch task failure.

[0005] In addition, although some systems have a primary and backup path mechanism, they do not establish an association between the path scoring function and the business type, and cannot dynamically adjust the scoring weight and selection logic according to different session scenarios (such as remote control, preview return, file synchronization, etc.), resulting in inaccurate resource allocation and rigid scheduling strategy.

[0006] Finally, the existing path scheduling mechanism generally lacks a safe exploration mechanism and steady-state detection strategy, which may cause link resource overload or scheduling out of control in frequent switching or strategy shock, and is not conducive to the long-term stable operation of large-scale business systems.

[0007] In summary, there is an urgent need for a remote session path adaptive orchestration scheme that supports collaborative modeling of quality and cost, has an adaptive scoring mechanism for session types, enables seamless path-level switching, and introduces budget governance and security exploration strategies, in order to improve session quality assurance capabilities and resource scheduling efficiency in complex multi-carrier network environments. Summary of the Invention

[0008] This invention aims to address the problems of single path scheduling, slow handover response, uncontrollable costs, and lack of steady-state control in existing multi-operator remote session scenarios. It proposes a path adaptive orchestration method that integrates quality and cost constraints. This method can dynamically score path quality according to session type, achieve intelligent scheduling by combining budget status, and achieve seamless handover through primary / backup preheating and mirroring mechanisms, thereby improving the stability, economy, and controllability of remote sessions in multi-link environments.

[0009] To address the aforementioned technical problems, the first aspect of the present invention provides a multi-carrier remote session path adaptive orchestration method, applicable to network architectures including clients, edge relay nodes, and control planes. The method includes: Step 1: Aggregate candidate path resources from multiple operators and perform low-disturbance quality and cost measurements on each path. The quality indicators include latency, jitter, packet loss, throughput, and queuing depth. The cost indicators include unit traffic cost, budget utilization rate, and cross-regional additional fees. Step 2: Identify the type of the current session, calculate the multi-objective score for each candidate path, the score integrates the quality sub-score, cost sub-score and risk penalty item, and is dynamically adjusted based on the weight parameters; Step 3: Select the primary path and backup path with the highest scores, establish a warm-up connection and synchronize context information; Step 4: When the primary path score is lower than the switching threshold or the cost is limited, start parallel transmission with the backup path, perform mirroring or sharding traffic redirection, and complete the primary / backup switchover when the hysteresis and hold-up time conditions are met, or fall back to the suboptimal path according to the degrade sequence. Step 5: Introduce a budget penalty factor to control concurrency of high-cost paths, and set up a low-volume exploration mechanism to conduct periodic path evaluations while meeting quality and cost constraints; among which... The scoring function has the following form:

[0010] The scoring items and weight parameters are used to reflect the multi-objective fusion optimization relationship and can be dynamically adjusted according to the business type; in For quality sub-scores, Score the cost sub-item. It is a fluctuation penalty term, and the weight is... , , It dynamically adjusts based on session type to achieve type-driven path optimization, seamless switching, and cost-controlled scheduling.

[0011] Optionally, the quality sub-score Based on the latency, jitter, packet loss rate, throughput, availability and queuing depth of the candidate paths, the results are obtained through normalization and robust aggregation calculation within a sliding window.

[0012] Optionally, the cost sub-score The score includes unit traffic price, current budget usage percentage, cross-regional surcharges, and path quota consumption factor. The score is used to reflect the relative cost burden of path usage.

[0013] Optionally, the risk penalty item This includes the statistical variance of the anomaly count and quality metrics of the path within a recent time window; the weighting parameters. , and Dynamically adjusts based on the current session type; interactive sessions are increased. Weighting, postback-type session increases Weights.

[0014] Optionally, the switching criteria between the primary path and the backup path include a score difference greater than a preset threshold, a score duration exceeding the minimum retention time, and meeting hysteresis control conditions; if the switching fails or the cost exceeds the limit, the path will automatically switch to the second-best low-cost path according to the predefined path degradation sequence and the reason for the rollback will be recorded.

[0015] The second aspect of the technical solution of the present invention provides a multi-carrier remote session path adaptive orchestration system, deployed in a network architecture including clients, edge relay nodes, and a control plane, characterized in that it includes: The Path Measurement and Profiling module is used to perform low-disturbance quality measurements on candidate paths from multiple operators and to build path quality and cost profiles. The type identification and weight configuration module is used to identify the service type of the remote session and configure the weight parameters in the multi-objective scoring according to the service type. The multi-objective scoring and ranking module is used to calculate the score of each candidate path based on the quality sub-score, cost sub-score, and risk penalty term. And sort the paths according to the scoring results; The dual-path preheating and switching module is used to maintain the preheating connection between the primary path and the backup path, and to perform path switching or fragmentation traffic diversion operations when the scoring results meet the preset conditions. The budget governance and quota module is used to monitor the budget usage status of tenants or projects and dynamically adjust the cost weight parameters and high-price path selection permissions in the multi-objective scoring and ranking module. The exploration and steady-state control module is used to perform small-volume exploration on non-main paths and conduct periodic steady-state assessments. The strategy storage and log module is used to record scoring parameters, path switching events and budget status, and supports historical backtracking and strategy playback.

[0016] The scoring results of the multi-objective scoring and ranking module are used to drive the dual-path preheating and switching module to select the current primary path and backup path. Combined with the budget status feedback from the budget governance and quota module, the dynamic adjustment of path selection priority and switching criteria is realized to achieve stable, efficient and cost-controllable path scheduling driven by multi-objective fusion.

[0017] Optionally, the scoring results of the multi-objective scoring and ranking module are used to trigger the dual-path warm-up and switching module to perform primary and backup path selection and switching operations, and to determine the path switching based on the scoring difference, holding time and hysteresis conditions; the budget status of the budget governance and quota module is used to dynamically adjust the cost weight parameters and high-price path selection priority in the scoring module, thereby realizing path scoring correction and concurrency limitation based on budget feedback.

[0018] The third aspect of the present invention provides a non-transitory computer-readable storage medium storing program instructions thereon, characterized in that, when the program instructions are executed by a processor, they are used to implement the method as described in claim 1, including steps such as path quality and cost measurement, type identification and scoring, primary and backup path warm-up and switching, budget control and path exploration.

[0019] The fourth aspect of the present invention provides an electronic device, including a processor and a memory, wherein the memory stores program instructions, characterized in that the processor, when executing the program instructions, is used to implement the method as described in claim 1, including operations such as path scoring, path switching, budget governance, and low-flow exploration.

[0020] Optionally, the processor is also used to identify the service type of the remote session in real time during operation, and perform path selection, path switching or degradation control based on the current scoring result, so as to achieve seamless switching and stable scheduling for real-time interactive scenarios.

[0021] The beneficial effects of the technical solution of this invention are: The adaptive orchestration method and system of the present invention achieve more accurate primary and backup path selection by integrating multi-dimensional quality indicators and risk penalty items into a path scoring mechanism and dynamically adjusting the scoring weight based on the session type, thereby significantly reducing the interruption rate and lag rate during remote interaction.

[0022] The adaptive orchestration method and system of this invention introduce cost modeling and budget penalty mechanism in the scoring and scheduling process, so that the use of high-priced paths and budget occupation form dynamic constraints, ensuring priority passage of critical sessions while reducing the unit session cost by 10% to 35%.

[0023] The adaptive orchestration method and system of the present invention achieve path switching within 200 to 800 milliseconds through a primary and backup path preheating mechanism and short-term parallel mirroring or sharding diversion method. Compared with the traditional reconnection mechanism, the switching latency is reduced by more than 80%, and the user-perceived interruption is greatly reduced.

[0024] The adaptive orchestration method and system of this invention introduce a safe exploration strategy and a steady-state judgment mechanism, which effectively suppresses frequent path oscillations, enables the scheduling system to have adaptive capabilities and stable convergence, reduces the switching frequency by more than 30%, and improves the overall reliability of the system.

[0025] The adaptive orchestration method and system of the present invention have the ability to identify service types and adapt scoring strategies, and can be compatible with various session scenarios such as remote interaction, screen preview, and file return, meeting the needs of multi-tenant budget isolation and differentiated services. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating the steps of the adaptive orchestration method in an embodiment of the present invention; Figure 2 This is a block diagram of the adaptive orchestration system in an embodiment of the present invention. Detailed implementation method:

[0027] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but this is not intended to limit the scope of the invention.

[0028] like Figure 1 As shown, the multi-carrier remote session path adaptive orchestration method of the present invention, which integrates quality and cost constraints, is deployed in an architecture consisting of clients, edge relays or forwarding nodes, and the control plane. The specific steps are as follows: Step 1: Path Resource Registration and Baseline Measurement. This step aggregates line resources from multiple operators and of various types, including public BGP lines, enterprise leased lines, WAN acceleration, cross-border optimization, and self-built backbones. A set of candidate paths is assigned to each session based on proximity, and quality metrics such as round-trip latency, jitter, packet loss, effective throughput, availability, and queuing depth are collected in a low-disturbance manner. Cost metrics such as line unit price, quota, and budget consumption are also recorded. Shadow probing and packet sampling can be used to keep measurement overhead within 0.2% to 2% of the total bandwidth. For example, if the candidate path set includes multiple lines from operators A, B, and C, line metrics are obtained through periodic probing and shadow sampling, with the period set to 1 to 5 seconds, and shadow packets occupying no more than 1% of the total bandwidth. Quality metrics include RTT, Jitter, PLR, TP, Availability, and Qlen, employing robust quantiles and sliding windows to smooth out abnormal spikes; cost metrics include Price per unit of traffic, BudgetUse, RegionFee, and Quota.

[0029] Step 2: Session Type Awareness and Multi-Objective Scoring: Identify session types and business intentions, such as remote interaction, screen viewing, screen wall thumbnail preview, file synchronization, or video recording playback. Calculate the multi-objective score for the i-th alternative path. .in The quality sub-score is obtained by normalizing and robustly aggregating factors such as latency, jitter, packet loss, and near-end and far-end CPU or queue utilization. The cost sub-score is determined by combining the unit traffic price, the percentage of budget already used, and cross-regional surcharges. Penalties for risk and volatility, such as quality variance and recent outlier counts. , , Adjust dynamically according to the type of conversation, such as when the interaction scenario increases. In conjunction with jitter weighting, throughput and cost weighting are increased in backhaul scenarios. Threshold thresholds and weighted sorting can be introduced, with the top 1 path as the primary path and the top 2 as a warm-up backup. For example, a weighting example for remote interaction sessions would be... =0.65, =0.20, =0.15, with jitter and RTT having a higher weighting in the quality sub-item; an example of file return session weighting is... =0.45, =0.45, =0.10, with a higher throughput proportion in the quality sub-item. A stability penalty term, Risk, is also introduced, including anomaly counts within the near window and metric variance. The main path must satisfy the condition that its Score is higher than the secondary path by at least ΔS and its duration is greater than... The path can only be switched after it reaches a stable threshold.

[0030] Step 3: Dual-Path Warm-up and Seamless Switchover: Maintain a lightweight warm-up connection or fast handshake state for both the primary and backup paths, ensuring that context such as keys and congestion windows can be quickly activated. When the primary path score falls below the switchover threshold or a sudden packet loss occurs, trigger short-term parallel mirroring or sharding of traffic to both paths, while monitoring the business layer QoE metrics. If the backup path score remains higher than the primary path score by exceeding the hysteresis threshold and minimum hold time, the primary / backup switchover is completed. If the backup path does not meet stability standards or exceeds cost limits, it is downgraded to a suboptimal low-cost path or falls back to fault-tolerant mode according to the fallback sequence. For example, the backup path maintains a warm-up state, typically keeping it in a handshake state or with low-frequency heartbeats, and caching the congestion window and keys. When the primary path quality threshold is triggered, such as RTT exceeding the threshold or PLR exceeding the threshold, or when budget constraints are triggered, enter the mirroring phase, lasting 100 to 400 milliseconds, to perform dual transmission or sharding of critical business packets. After the switchover is completed, the window... The system observes the QoE (Quality of Equivalent) and automatically backtracks or switches to the next candidate in the sequence if the QoE is not met.

[0031] Step 4: Cost Budget Governance and Quota Protection: Set time window budgets and instantaneous quota limits for tenants, projects, or regions. The scheduler introduces a budget penalty factor during scoring, gradually increasing the cost weight of high-priced paths or reducing their maximum concurrency when the budget percentage exceeds a threshold p. Strong constraints are set for high-priced resources such as cross-border or dedicated lines, allowing access only when necessary, such as in alarms or high-priority sessions, and compensating for historical performance losses due to budget limitations during off-peak periods. For example, set periodic budgets and instantaneous concurrency limits for tenants; when BudgetUse exceeds a threshold p percentage (e.g., 90%), increase the cost penalty coefficient or reduce the concurrency limit and maximum occupancy of high-priced paths. Cross-border and dedicated line paths must meet triggering conditions, such as session priority being alarms or interaction level being high, before selection is allowed.

[0032] Step 5: Safe Exploration and Steady-State Convergence: Employ a safe exploration strategy to perform limited business sampling or shadow replication on non-current main paths. Common strategies include... A safe variant of greedy sampling combined with UCB or Bayesian Thompson sampling. During exploration, protective constraints of a lower quality limit and a higher cost limit must be met, and frequent oscillations are avoided through steady-state detection and hysteresis windows. After convergence, exploration is only repeated within anomaly or periodic re-evaluation windows. For example, each scoring period uses probability... Choose a non-primary path for low-volume exploration, or use UCB Thompson sampling. Exploration is constrained by quality and cost protections, such as RTT not exceeding a threshold limit and unit cost not exceeding a set limit. When the system enters steady state, reduce... And extend the reassessment window to avoid frequent switching.

[0033] likeFigure 2 As shown, the multi-carrier remote session path adaptive orchestration method of the present invention, which integrates quality and cost constraints, also has a corresponding system and apparatus: The system of the present invention corresponds to the above method steps and includes the following modules: Path Measurement and Profiling Module: Performs low-disturbance measurement and statistical modeling on candidate paths, obtains path quality and cost indicators, and establishes path profiles.

[0034] Type recognition and weight configuration module: Classifies sessions and configures λ weights and thresholds to adapt to the needs of different session types.

[0035] Multi-objective scoring and ranking module: Calculates the score and outputs the primary and backup paths and hysteresis parameters. By integrating multiple objective factors such as quality, cost and risk, it provides a basis for path selection.

[0036] Dual-path warm-up and switching module: Maintains the warm-up state, implements mirroring or sharding traffic redirection, completes seamless switching and rollback, and ensures that it can quickly switch to the backup path when the main path has problems, thus ensuring business continuity.

[0037] Budget governance and quota module: Maintains budget status and flow restriction rules, injects cost penalties into scoring, and achieves effective governance of cost budgets and quota protection.

[0038] The exploration and steady-state control module performs safe exploration, policy convergence, and jitter suppression to explore better paths while ensuring system stability.

[0039] Policy storage and log module: Stores parameters, models, permissions and execution logs, and provides auditing and playback functions to facilitate monitoring and analysis of system operation.

[0040] The computer-readable storage medium corresponding to the multi-carrier remote session path adaptive orchestration method that integrates quality and cost constraints plays a crucial role in this invention. It is a carrier of stored program instructions that enable the processor to perform operations related to implementing multi-carrier remote session path adaptive orchestration that integrates quality and cost constraints, namely, the steps from step 1 to step 5 described above.

[0041] Non-transitory computer-readable storage media means that the media has permanent or semi-permanent storage capabilities and will not lose the stored information due to power outages or other transient system changes. Common non-transitory computer-readable storage media include, but are not limited to, hard disk drives (HDDs), solid-state drives (SSDs), optical discs (such as CD-ROMs and DVD-ROMs), flash drives, and various other forms of non-volatile memory.

[0042] When the processor reads and executes program instructions from the storage medium, it sequentially executes steps 1 through 5 according to predetermined logic and algorithms. Specifically, in step 1, the processor, based on the code instructions in the storage medium, controls relevant modules to aggregate candidate path resources from multiple operators and collects quality indicators such as round-trip latency, jitter, packet loss, effective throughput, availability, and queuing depth in a low-perturbation manner, while simultaneously recording cost indicators such as line unit price, quota, and budget consumption. For step 2, the processor identifies the type of the current session by executing instructions and calculates the multi-objective score for each candidate path according to the stored algorithm, realizing the fusion of quality sub-scores, cost sub-scores, and risk penalty items, as well as the dynamic adjustment of weight parameters. In step 3, the processor executes instructions to select the primary path and backup path with the highest score, establishes a pre-warm-up connection, and synchronizes context information. When step 4 is executed, the processor monitors the primary path score and cost. If the primary path score falls below the switching threshold or is cost-constrained, parallel transmission with the backup path is initiated, mirroring or sharding is performed, and a primary / backup switchover is completed based on hysteresis and hold time conditions, or a fallback to the suboptimal path is performed according to the degradation sequence. Finally, in step 5, the processor introduces a budget penalty factor to control the concurrency of high-cost paths according to program instructions, and sets up a low-volume exploration mechanism to conduct periodic path evaluations while meeting quality and cost constraints.

[0043] The electronic device corresponding to the multi-operator remote session path adaptive orchestration method that integrates quality and cost constraints of the present invention serves as the physical device for implementing the method of the present invention, and its core components are a processor and a memory.

[0044] The processor is the core of an electronic device's computation and control. It is responsible for executing program instructions stored in memory, thereby enabling a series of complex operations, from path resource registration and baseline measurement to safety exploration and steady-state convergence. The processor possesses powerful computing and logic processing capabilities, enabling it to rapidly process large amounts of path-related data and session information. For example, during path measurement, the processor needs to quickly process and analyze the collected quality and cost metrics data; when calculating multi-objective scores, the processor must accurately perform calculations and weight adjustments on various scoring factors based on different session types.

[0045] The memory stores the instructions and related data required by the processor to execute the program. It stores not only the program code implementing steps 1 through 5, but also various metric data obtained from path measurement, session type information, scoring parameters, and other configuration information related to system operation. For example, the memory records the quality sub-score, cost sub-score, risk penalty, and corresponding weight parameters for each candidate path, so that the processor can access them at any time when calculating multi-objective scores. Simultaneously, the memory also stores information related to the primary and backup paths, including pre-warm-up connection status and context information, supporting accurate operation by the processor during path switching.

[0046] Once the electronic device is powered on, the processor reads program instructions from memory and executes steps 1 through 5 sequentially according to the logical order of the instructions. During operation, the processor interacts with the memory in real time, reading the necessary data and storing the processing results back into memory. For example, during path exploration, the processor selects a non-primary path for low-volume exploration based on the quality and cost constraints stored in memory, and stores the exploration results and related statistics in memory for subsequent steady-state evaluation and path selection adjustments. Through close collaboration between the processor and memory, the electronic device can efficiently and stably achieve adaptive orchestration of multi-carrier remote session paths, meeting the needs of different session types while balancing quality and cost.

[0047] The following are specific embodiments of the multi-carrier remote session path adaptive orchestration method that integrates quality and cost constraints according to the present invention: Example 1, Path Measurement and Profiling: In practical application scenarios, the candidate path set includes multiple lines from operators A, B, and C. Line metrics are obtained through periodic probing and shadow sampling, with the period set to 3 seconds (selectable within the range of 1 to 5 seconds). Shadow packets occupy 0.8% of the total bandwidth (not exceeding 1% of the total bandwidth). Regarding quality metrics, robust quantiles and sliding window methods are used to process latency (RTT), jitter, packet loss rate (PLR), throughput (TP), availability (Availability), and queuing depth (Qlen) to smooth out abnormal spikes and accurately reflect path quality. For cost metrics, detailed records are kept of unit traffic price (Price), used budget (BudgetUse), inter-regional surcharge (RegionFee), and quota (Quota), providing comprehensive cost information for subsequent path scoring.

[0048] Example 2, Multi-objective scoring and type weighting: For remote interactive sessions, weights are set. =0.65, =0.20, =0.15, within the quality sub-items, jitter and RTT have relatively high weights to meet the real-time and stability requirements of interactive sessions. For file return sessions, the weights are set to... =0.45, =0.45, =0.10, with throughput receiving a higher weight in the quality sub-item because file return prioritizes data transfer speed. A stability penalty term, Risk, is also introduced, including anomaly counts within the near window and metric variance, to assess path stability. The primary path must satisfy the following conditions: Score is at least ΔS higher than the secondary path (e.g., 5 points, adjustable based on actual conditions) and duration is greater than [missing value]. (e.g., 2 seconds, which can be adjusted according to the actual situation) The path can only be switched after it reaches a stable threshold to ensure the rationality and stability of the path switching.

[0049] Example 3, Dual-Path Warm-up and Seamless Switchover: During system operation, the backup path maintains a warm-up state, keeping in handshake mode or low-frequency heartbeat, and caches congestion windows and keys for rapid activation when needed. When the primary path experiences quality issues, such as RTT exceeding a preset threshold (e.g., 100ms, adjustable according to actual conditions) or PLR exceeding a preset threshold (e.g., 1%, adjustable according to actual conditions), or when switching is triggered due to budget constraints, a mirroring phase is initiated, with a duration set to 300 milliseconds (within the range of 100 to 400 milliseconds), during which critical service packets are dual-transmitted or fragmented for diversion. After the switchover is completed, in the window... (e.g., 5 seconds, which can be adjusted according to the actual situation) Observe the QoE within a certain period. If the QoE does not meet the standard, automatically backtrack or switch to the next candidate path in the sequence to ensure the stability of the business and the user experience.

[0050] Example 4, Budget Governance and Quota Protection: A monthly budget of 1000 yuan is set for each tenant (adjustable based on actual needs), and the instantaneous concurrent connection limit is 50 connections (adjustable based on actual needs). When BudgetUse exceeds a threshold percentage (e.g., 90%), the cost penalty coefficient is increased, for example, by adjusting the cost weighting. Increase the concurrency limit and maximum resource usage of high-priced paths by 20% (adjustable based on actual conditions). For cross-border and dedicated line paths, set the trigger condition to allow selection only when the session priority is alarm or the interaction level is high, ensuring the reasonable use of high-priced resources. During off-peak periods, such as 2 AM to 6 AM (adjustable based on actual conditions), compensate for historical performance losses caused by budget constraints and optimize resource allocation.

[0051] Example 5, Safety Exploration and Steady-State Control: Each scoring cycle (e.g., 1 minute, which can be adjusted according to actual conditions) is based on probability. (e.g., 0.1, adjustable according to actual conditions) Select non-main paths for low-volume exploration, or use UCB Thompson sampling. The exploration process is subject to quality and cost protection constraints, such as RTT not exceeding the threshold limit (e.g., 150ms, adjustable according to actual conditions), and the unit cost not exceeding the set limit (e.g., 0.1 yuan / MB, adjustable according to actual conditions). When the system enters steady state, reduce... Set the value to 0.05 (adjustable according to actual conditions) and extend the reassessment window to 5 minutes (adjustable according to actual conditions) to avoid frequent switching and ensure stable system operation.

[0052] Example 6, Rollback and Degradation: The rollback sequence is defined as high-level leased lines, cross-border optimization, and public network preferred paths. When the backup path score is insufficient, or when a rollback is triggered due to budget constraints, the next low-cost path in the sequence is selected. For example, if the backup path is a cross-border optimized line, but its score is insufficient to meet business needs, the system switches to the public network preferred path, and records the rollback reason (e.g., "backup path score insufficient") and recovery conditions (e.g., "cross-border optimized line score meets requirements") to facilitate automatic recovery and ensure continuous operation of the business under different conditions.

[0053] In summary, this invention provides a multi-operator remote session path adaptive orchestration scheme that integrates quality assessment, cost constraints, and service type-driven mechanisms. Addressing the technical pain points of current multi-link networks, such as limited path selection, untimely handover, and uncontrollable costs, it constructs a complete scheduling system covering path measurement, scoring calculation, primary / backup coordination, budget governance, and steady-state exploration. By designing a multi-objective scoring function, it comprehensively considers the path's quality sub-score, cost sub-score, and risk penalty factor, and dynamically adjusts the scoring weights based on session characteristics to achieve intelligent sorting and optimal combination selection of primary and backup paths. Combined with hysteresis control and a warm-up mechanism, it ensures a smooth and seamless path handover process.

[0054] Furthermore, this invention introduces a budget penalty and quota control mechanism to prevent the overuse of high-priced paths, ensuring that the overall scheduling strategy balances service quality with cost control. Combined with a low-traffic exploration mechanism and periodic evaluation logic, it enhances the real-time perception of quality changes in non-primary paths, strengthening the system's adaptability to dynamic network environments. The overall solution boasts advantages such as clear structure, flexible deployment, intelligent scheduling, and stable performance. It is suitable for various typical edge computing and multi-carrier collaboration scenarios, including real-time interaction with end users, high-definition video transmission, and remote work, demonstrating broad engineering application prospects and significant promotional value.

[0055] The above are merely preferred embodiments of the present invention and are not intended to limit the implementation methods and protection scope of the present invention. Those skilled in the art should recognize that any equivalent substitutions and obvious changes made based on the description and illustrations of the present invention should be included within the protection scope of the present invention.

Claims

1. A multi-carrier remote session path adaptive orchestration method, applicable to a network architecture including clients, edge relay nodes, and a control plane, the method comprising: Step 1: Aggregate candidate path resources from multiple operators and perform low-disturbance quality and cost measurements on each path. The quality indicators include latency, jitter, packet loss, throughput, and queuing depth. The cost indicators include unit traffic cost, budget utilization rate, and cross-regional additional fees. Step 2: Identify the type of the current session, calculate the multi-objective score for each candidate path, the score integrates the quality sub-score, cost sub-score and risk penalty item, and is dynamically adjusted based on the weight parameters; Step 3: Select the primary path and backup path with the highest scores, establish a warm-up connection and synchronize context information; Step 4: When the primary path score is lower than the switching threshold or the cost is limited, start parallel transmission with the backup path, perform mirroring or sharding traffic redirection, and complete the primary / backup switchover when the hysteresis and hold-up time conditions are met, or fall back to the suboptimal path according to the degrade sequence. Step 5: Introduce a budget penalty factor to control concurrency of high-cost paths, and set up a low-volume exploration mechanism to conduct periodic path evaluations while meeting quality and cost constraints; among which... The scoring function has the following form: ； The scoring items and weight parameters are used to reflect the multi-objective fusion optimization relationship and can be dynamically adjusted according to the business type; in For quality sub-scores, Score the cost sub-item. It is a fluctuation penalty term, and its weight is... , , It dynamically adjusts based on session type to achieve type-driven path optimization, seamless switching, and cost-controlled scheduling.

2. The method according to claim 1, characterized in that, The quality sub-score Based on the latency, jitter, packet loss rate, throughput, availability and queuing depth of the candidate paths, the results are obtained through normalization and robust aggregation calculation within a sliding window.

3. The method according to claim 1, characterized in that, The cost sub-score The score includes unit traffic price, current budget usage percentage, cross-regional additional fees, and path quota consumption factor. The score is used to reflect the relative cost burden of path usage.

4. The method according to claim 1, characterized in that, The aforementioned risk penalty items This includes the statistical variance of the anomaly count and quality metrics of the path within a recent time window; the weighting parameters. , and Dynamically adjusts based on the current session type; interactive sessions are increased. Weighting, postback-type session increases Weights.

5. The method according to claim 1, characterized in that, The switching criteria between the primary path and the backup path include a score difference greater than a preset threshold, a score duration exceeding the minimum retention time, and meeting hysteresis control conditions. If the switching fails or the cost exceeds the limit, the path will automatically switch to the second-best low-cost path according to the predefined path degradation sequence and the reason for the rollback will be recorded.

6. A multi-carrier remote session path adaptive orchestration system, deployed in a network architecture including clients, edge relay nodes, and a control plane, characterized in that, include: The Path Measurement and Profiling module is used to perform low-disturbance quality measurements on candidate paths from multiple operators and to build path quality and cost profiles. The type identification and weight configuration module is used to identify the service type of the remote session and configure the weight parameters in the multi-objective scoring according to the service type. The multi-objective scoring and ranking module is used to calculate the score of each candidate path based on the quality sub-score, cost sub-score, and risk penalty term. And sort the paths according to the scoring results; The dual-path preheating and switching module is used to maintain the preheating connection between the primary path and the backup path, and to perform path switching or fragmentation traffic diversion operations when the scoring results meet the preset conditions. The budget governance and quota module is used to monitor the budget usage status of tenants or projects and dynamically adjust the cost weight parameters and high-price path selection permissions in the multi-objective scoring and ranking module. The exploration and steady-state control module is used to perform small-volume exploration on non-main paths and conduct periodic steady-state assessments. The strategy storage and log module is used to record scoring parameters, path switching events and budget status, and supports historical backtracking and strategy replay; The scoring results of the multi-objective scoring and ranking module are used to drive the dual-path preheating and switching module to select the current primary path and backup path. Combined with the budget status feedback from the budget governance and quota module, the dynamic adjustment of path selection priority and switching criteria is realized to achieve stable, efficient and cost-controllable path scheduling driven by multi-objective fusion.

7. The system according to claim 6, characterized in that, The scoring results of the multi-objective scoring and ranking module are used to trigger the dual-path warm-up and switching module to perform primary and backup path selection and switching operations, and to determine the path switching based on the score difference, holding time and hysteresis conditions; the budget status of the budget governance and quota module is used to dynamically adjust the cost weight parameters and high-price path selection priority in the scoring module, thereby realizing path score correction and concurrency limitation based on budget feedback.

8. A non-transitory computer-readable storage medium having program instructions stored thereon, characterized in that, When the program instructions are executed by the processor, they are used to implement the method as described in claim 1, including steps such as path quality and cost measurement, type identification and scoring, primary and backup path warm-up and switching, budget control and path exploration.

9. An electronic device comprising a processor and a memory, wherein the memory stores program instructions, characterized in that, When executing the program instructions, the processor is used to implement the method as described in claim 1, including operations such as path scoring, path switching, budget governance, and low-traffic exploration.

10. The electronic device according to claim 9, characterized in that, The processor is also used to identify the service type of remote sessions in real time during operation, and to perform path selection, path switching or degradation control based on the current scoring results, so as to achieve seamless switching and stable scheduling for real-time interactive scenarios.

Images