CPE network tuning method and system
By building a business context drift perception model and a cross-layer compensation tuning mechanism, the problem of CPE device tuning lag in a dynamic network environment is solved, high-precision and adaptive network configuration updates are achieved, and the stability and adaptability of network services are improved.
Patent Information
- Application Number
- CN202510942149.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-09
- Publication Date
- 2025-09-30
AI Technical Summary
Existing CPE devices lack the ability to deeply perceive the service context status when facing diverse services and dynamic network environments, resulting in delayed tuning and configuration and difficulty in achieving dynamic adaptation. This is especially true in scenarios with multiple terminals running concurrently, high-frequency online and offline operations, or frequent QoS fluctuations, resulting in low network service stability and tuning accuracy.
By building a business context drift perception model, collecting and generating context signature fingerprints, detecting business drift events in real time, identifying tuning mismatch items, and dynamically updating configuration parameters through a cross-layer compensation tuning mechanism, adaptive evolutionary tuning is achieved.
It significantly improves the tuning response accuracy and adaptability of CPE devices in complex environments, enhances the stability and intelligent operation and maintenance capabilities of network services, and can quickly restore performance and optimize resource allocation strategies.
Smart Images
Figure CN120730341A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of wireless communication networks, and in particular to a CPE network optimization method and system. Background Art
[0002] With the widespread adoption of IoT operating systems (IoT OS) in home gateways, industrial terminals, and edge CPE devices, the number of access terminals continues to grow and service types become more diverse. Network resource scheduling faces a structural contradiction between the dynamic evolution of service states and the static and fixed network protocol configurations. CPE devices not only need to coordinate physical link resources such as WiFi / Ethernet, but also complete service scheduling, resource binding, and QoS assurance across protocol stack layers. This leads to increasing tuning pressure. To improve terminal service stability and user experience, it is urgent to build a tuning mechanism that can perceive service changes and dynamically optimize configurations, supported by the IoT OS.
[0003] Most current CPE tuning methods rely on fixed rule engines or policy templates, such as automatically switching channels or adjusting power based on RSSI / packet loss rate. They lack the ability to deeply perceive the evolution of business context states. Specifically: (1) They are unable to perceive the scheduling behavior and resource binding status of terminal threads in the IoT operating system; (2) They lack structured analysis of protocol stack call traces, resulting in configuration adjustments lagging behind the actual evolution of the business; (3) Existing optimization methods are mainly passive responses, and no mapping mechanism between tuning history and business logic has been established, making it difficult to achieve dynamic adaptive updates. Especially in scenarios with multiple terminals concurrently, high-frequency online and offline, or frequent QoS fluctuations, existing technologies have low tuning accuracy and high mismatch rate, which seriously affects the stability of network services. Summary of the Invention
[0004] The present invention provides a CPE network tuning method and system, which realizes the self-evolutionary update of the tuning model. This method significantly improves the adaptability and tuning efficiency of CPE equipment to complex business environments.
[0005] A CPE network optimization method includes the following steps: S1, based on the IoT operating system, collects the service scheduling path, network protocol stack call trace, and resource mapping information of each access terminal in the CPE device, generates a context signature fingerprint to represent the relationship between the service logic state and the IoT operating system scheduling, and stores the context signature fingerprint in time series as a service context drift perception model; S2 detects service context drift events and identifies tuning mismatch items. It performs fuzzy matching based on the current context signature fingerprint and the historical context sequence in the service context drift perception model to determine whether a service context drift event has occurred. If so, it extracts tuning configuration items affected by drift based on the drift judgment threshold in the service context drift perception model, including channel allocation, power control, interface binding, and QoS level parameters, and marks them as tuning items to be reconfigured. S3 performs cross-layer compensation tuning and updates the business context drift perception model. Based on the tuning items to be reconstructed, configuration parameter compensation and reconstruction operations are performed between the network stack and the IoT operating system. At the same time, based on the current round of context matching results and policy adaptation feedback, the drift judgment threshold in the business context drift model parameters is corrected to form an adaptive and evolving business context drift perception model.
[0006] Optionally, the S1 includes: S11, calling the scheduling monitoring interface, protocol stack tracking interface, and resource mapping query interface of the IoT operating system to collect service scheduling paths, network protocol stack call traces, and resource mapping information of each CPE access terminal; S12: Based on the unique identifier of the terminal, the collected service scheduling path, network protocol stack call trace, and resource mapping information are aggregated, and the resource mapping relationship of the terminal within the specified time window is extracted in combination with the status information of the IoT operating system to form a resource usage structure including thread priority, cache binding, and channel occupancy; S13, the business scheduling path, network protocol call trace and resource usage structure corresponding to each terminal are fused and encoded to generate a context signature fingerprint, which is written into the business context drift perception model in chronological order to describe the dynamic relationship between the business logic state and the scheduling behavior of the IoT operating system.
[0007] Optionally, the S11 includes: S111, calling the scheduling monitoring interface of the IoT operating system to collect task scheduling paths for threads bound to each access terminal, including thread identifiers, task scheduling time, priority, and scheduling frequency; S112, calling the protocol stack tracking interface to collect the activity track of the access terminal in the IoT protocol stack, including the protocol call sequence, transmission status code and port resource occupancy time; S113, calling a resource mapping query interface to collect the binding relationship between the terminal thread and the system resources, including the binding relationship between the terminal thread and the processor core, the cache usage status and the IO channel occupancy status.
[0008] Optionally, the S12 includes: S121, based on the unique identifier of each access terminal, the collected service scheduling path, network protocol stack call trace and resource mapping information are aggregated to form a structured data set corresponding to the terminal ; S122, from the generated structured data set corresponding to the terminal Extract resource binding information and set the time window based on the scheduling status (scheduling queue length, system load index) in the IoT operating system , filter the effective resource usage records of the terminal in the window to form a resource occupancy filter set ; S123, filter the set based on the filtered resource occupancy The attributes of each item in the ,thread priority structure, cache binding structure, channel occupancy structure are extracted, and the resource usage structure of the current terminal is generated.
[0009] Optionally, the S122 includes: S1221, from the structured data set corresponding to the terminal Extract the resource binding information part and construct a resource record set; S1222, extract the current scheduling queue length from the IoT operating system scheduling status information and the system load index , and according to and Dynamically set resource active time window length; S1223, in the setting Filter resource record entries Start time Is it within the window range? Get the resource occupancy filter set ,in, The current system time.
[0010] Optionally, the S123 includes: S1231, filter set from resource occupancy Filter out resource record items whose resource type is processor core, extract the scheduling priority and occupancy time of the corresponding thread, and build the thread priority structure ; S1232: Filter out the record items whose resource types belong to cache unit resources from the resource occupation filter set, collect statistics on the binding status and cache hit status of each thread on different levels of cache units, and build a cache binding structure. ; S1233: Filter out the record items of resource types belonging to input and output channel resources from the resource occupancy screening set, count the cumulative usage time of each thread on different I / O channels, and build a channel occupancy structure. , while the thread priority structure , cache binding structure , channel occupancy structure Summarize and form a resource usage structure .
[0011] Optionally, the S13 includes: S131, the terminal In the current time slice Business scheduling path , Network protocol call trace and resource usage structure Encoded as a joint context feature vector ; S132, context feature vector Input fingerprint mapping function , generate context signature fingerprint ; S133, the generated context signature fingerprint Time tag For index, write the business context drift perception model corresponding to the terminal ,constructing contextual trajectories in the form of time series.
[0012] Optionally, the S2 includes: S21, extract the context signature fingerprint generated by the terminal at the current moment and business context drift-aware model Historical fingerprint sequence in One-to-one Hamming distance Calculate, if , it is determined that a business context drift event is currently occurring, where is the drift judgment threshold; S22: If drift is determined to have occurred, extract the tuning parameter configuration set corresponding to the historical fingerprint from the business context drift perception model. , and get the currently effective tuning configuration set , configure the tuning parameter set corresponding to the historical fingerprint The currently effective tuning configuration set Compare each item one by one, extract inconsistent or invalid tuning items, and form a set of tuning items to be reconstructed ,in, Respectively The first in the history Tuning parameters, The currently effective Tuning parameters, The currently effective Tuning parameters, For the history The corresponding tuning parameter items in the records; S23, the set to be reconstructed and tuned Each tuning item in Mark as items to be restructured and optimized, including channel allocation parameters (access frequency band, bandwidth index), transmit power control factor, interface binding priority (WiFi / Ethernet interface weight), QoS level parameters (DSCP value, transmission rate level), and output a structured optimization update requirement table .
[0013] Optionally, the S3 includes: S31, update the requirements table for the marked tuning , respectively, calling the unified configuration interface between the network stack protocol module and the IoT operating system resource management module to reset parameters; S32, construct the currently tuned context state into a new signature-configuration pair Writing a business context drift-aware model , as a new reference history, where The latest tuning configuration set after this round of execution; S33, based on the current fingerprint The closest to the business context drift awareness model The average distance between historical fingerprints and the tuning success feedback rate, and the dynamic correction of the drift judgment threshold .
[0014] A CPE network optimization system, used to implement the above-mentioned CPE network optimization method, includes the following modules: Contextual data collection module: Based on the IoT operating system, it collects the service scheduling path, network protocol stack call trace, and resource mapping information of each access terminal in the CPE device, and constructs a contextual signature fingerprint; Context drift perception module: This module stores context signature fingerprints in time series, builds a business context drift perception model, and performs fuzzy matching between the current fingerprint and the historical fingerprint to detect whether there is a business context drift event. Tuning mismatch identification module: After detecting a drift event, it extracts the affected tuning configuration items from the business context drift awareness model and generates the tuning items to be reconstructed. Cross-layer tuning execution module: Based on the tuning items to be reconfigured, it performs compensation and reconstruction of configuration parameters between the network protocol stack and the IoT operating system; Model adaptive update module: Based on the context matching results and policy adaptation feedback, it modifies the drift judgment threshold and updates the business context drift perception model.
[0015] Beneficial effects of the present invention: The present invention establishes a dynamic scheduling behavior mapping relationship between the terminal service status and the IoT operating system by constructing a context signature fingerprint mechanism that integrates service scheduling paths, protocol call traces, and resource mapping information. This can depict the operating characteristics of each access terminal in the CPE device in real time, effectively avoiding the limitations of the existing technology of fuzzy tuning basis and reliance on static configuration, and significantly improving the pertinence and accuracy of the tuning response.
[0016] The present invention, by introducing a service context drift perception model and combining it with a fingerprint fuzzy matching mechanism, can accurately detect tiny service behavior drifts caused by system load fluctuations, protocol state changes, or multi-terminal competition. It can locate potential tuning mismatch items at the perception layer, avoiding link fluctuations and service quality degradation caused by parameter failure or historical configuration deviation, thereby enhancing the stability and adaptability of the CPE network in a multi-service access environment.
[0017] The present invention realizes the coordinated optimization between network stack parameters and operating system resource configuration by constructing a cross-layer compensation tuning mechanism. It automatically generates a tuning update table by combining historical fingerprint-configuration mapping, and corrects the judgment threshold in real time according to the tuning results, forming an adaptive evolutionary closed-loop tuning model. It not only has the ability to quickly restore performance, but also can sustainably optimize resource allocation strategies in a dynamically changing environment, greatly improving the network service consistency and intelligent operation and maintenance capabilities of CPE equipment. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0019] Figure 1 A flow chart of a tuning method according to an embodiment of the present invention; Figure 2 Schematic diagram of system function modules according to an embodiment of the present invention. DETAILED DESCRIPTION
[0020] The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. Those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments and are not intended to specifically limit the present invention.
[0021] like Figure 1 As shown, a CPE network tuning method includes the following steps: S1, based on the IoT operating system, collects the service scheduling path, network protocol stack call trace, and resource mapping information of each access terminal in the CPE device, generates a context signature fingerprint to represent the relationship between the service logic state and the IoT operating system scheduling, and stores the context signature fingerprint in time series as a service context drift perception model; S2 detects service context drift events and identifies tuning mismatch items. It performs fuzzy matching based on the current context signature fingerprint and the historical context sequence in the service context drift perception model to determine whether a service context drift event has occurred. If so, it extracts tuning configuration items affected by drift based on the drift judgment threshold in the service context drift perception model, including channel allocation, power control, interface binding, and QoS level parameters, and marks them as tuning items to be reconfigured. S3 performs cross-layer compensation tuning and updates the business context drift perception model. Based on the tuning items to be reconstructed, configuration parameter compensation and reconstruction operations are performed between the network stack and the IoT operating system. At the same time, based on the current round of context matching results and policy adaptation feedback, the drift judgment threshold in the business context drift model parameters is corrected to form an adaptive and evolving business context drift perception model.
[0022] S1 includes: S11, calling the scheduling monitoring interface, protocol stack tracking interface, and resource mapping query interface of the IoT operating system to collect service scheduling paths, network protocol stack call traces, and resource mapping information of each CPE access terminal; S12: Based on the unique identifier of the terminal, the collected service scheduling path, network protocol stack call trace, and resource mapping information are aggregated, and the resource mapping relationship of the terminal within the specified time window is extracted in combination with the status information of the IoT operating system to form a resource usage structure including thread priority, cache binding, and channel occupancy; S13, the business scheduling path, network protocol call trace and resource usage structure corresponding to each terminal are fused and encoded to generate a context signature fingerprint, which is written into the business context drift perception model in chronological order to describe the dynamic relationship between the business logic state and the scheduling behavior of the IoT operating system.
[0023] The S11 includes: S111, calling the scheduling monitoring interface of the IoT operating system to collect task scheduling paths for threads bound to each access terminal, including thread identifiers, task scheduling time, priority, and scheduling frequency, wherein; Thread ID: unique thread number within the IoT operating system; Task scheduling time: the timestamp when the thread is scheduled to enter the running state; Thread priority: the static priority of the thread in the scheduler; Scheduling frequency: The number of times the thread is scheduled per unit time, expressed as: ; in, For the The scheduling frequency of threads, For the time window Neidi The total number of times a thread is scheduled; S112, calling the protocol stack tracking interface to collect the activity track of the access terminal in the IoT protocol stack, including the protocol call sequence, transmission status code and port resource occupancy time, wherein; Protocol calling sequence: The protocol stack layers called in sequence in the current active connection of the terminal, such as ; Transmission status code: The status code returned in each protocol call, including ACK, NACK, and Timeout; Port resource occupancy time: The length of time the terminal maintains a connection to the target port, expressed as: ; in, For the The duration of port resource occupation, The time when the connection to this port was established. The time when the port connection is disconnected; S113, calling the resource mapping query interface to collect the binding relationship between the terminal thread and the system resource, including the binding relationship between the terminal thread and the processor core, the cache usage status and the IO channel occupancy, wherein; Processor core binding relationship: The binding status of each terminal thread to the physical processor core, for example, Thread_ID → Core_2; Cache usage status: L1 / L2 cache hit ratio during thread execution, expressed as: ; in, For the The cache hit rate of each thread, is the number of cache hits, is the total number of cache accesses; IO channel usage: The names and usage duration of the DMA, SPI, and UART channels occupied by the terminal thread during the statistical period.
[0024] S12 includes: S121, based on the unique identifier of each access terminal, the collected service scheduling path, network protocol stack call trace and resource mapping information are aggregated to form a structured data set corresponding to the terminal , expressed as: ; in, is the scheduling path structure data, It is the protocol stack call trace, Bind information to resources; S122, from the generated structured data set corresponding to the terminal Extract resource binding information and set the time window based on the scheduling status (scheduling queue length, system load index) in the IoT operating system , filter the effective resource usage records of the terminal in the window to form a resource occupancy filter set ; S123, filter the set based on the filtered resource occupancy The attributes of each item in the ,thread priority structure, cache binding structure, channel occupancy structure are extracted, and the resource usage structure of the current terminal is generated.
[0025] S122 includes: S1221, from the structured data set corresponding to the terminal Extract the resource binding information part and construct a resource record set, which is expressed as: ; in, For the A subset of resource binding information for each terminal, It is a resource record item, including the binding thread ID, resource type, and start time. A function to determine the record type. Represents the system hardware or low-level communication resource items bound to or used by terminal threads in the IoT operating system. Each resource record represents the occupation or binding behavior of a thread to a certain system resource within a certain period of time. S1222, extract the current scheduling queue length from the IoT operating system scheduling status information and the system load index , and according to and Dynamically set resource active time window The length of is expressed as: ; in, 、 is the adjustment coefficient; S1223, in the setting Filter resource record entries Start time Is it within the window range? Get the resource occupancy filter set ,in, is the current system time, expressed as: .
[0026] S123 includes: S1231, filter set from resource occupancy Filter out resource record items whose resource type is processor core, extract the scheduling priority and occupancy time of the corresponding thread, and build the thread priority structure , expressed as: ; ; in, For the Terminal thread identifier, For the The weighted average priority of the terminal threads, For the The static priority of the thread in the record, For this thread The resource usage time in the record, The number of record entries for this thread; S1232: Filter out the record items whose resource types belong to cache unit resources from the resource occupation filter set, collect statistics on the binding status and cache hit status of each thread on different levels of cache units, and build a cache binding structure. , expressed as: ; ; in, For the threads, For the Layer caching, is the hit rate, is the number of cache hits, is the total number of cache accesses, is a very small constant; S1233: Filter out the record items of input / output channel resources from the resource occupancy filter set, including the occupancy relationship with the DMA controller, serial port module (UART), and SPI interface, count the cumulative usage time of each thread on different I / O channels, and build a channel occupancy structure. , while the thread priority structure , cache binding structure , channel occupancy structure Summarize and form a resource usage structure , expressed as: ; ; in, For the threads, For the Channel names, For threads In the channel The cumulative occupied time on the
[0027] S13 includes: S131, the terminal In the current time slice Business scheduling path , Network protocol call trace and resource usage structure Encoded as a joint context feature vector , expressed as: ; in, 、 、 Feature encoders representing three types of information (such as sequence embedding, hash mapping, or graph structure encoding), It is a vector concatenation operation; S132, context feature vector Input fingerprint mapping function , generate context signature fingerprint , expressed as: ; in, is the Gaussian distribution random projection matrix, is the time position perturbation encoding vector, , is the disturbance amplitude constant, is the frequency of the perturbation in each dimension, It is a sign function, the output is +1 for positive numbers and -1 for negative numbers; S133, the generated context signature fingerprint Time tag For index, write the business context drift perception model corresponding to the terminal , construct context trajectories in the form of time series, expressed as: ; Business context drift awareness model Expressed as: (1) Model structure definition: Business context drift perception model Using time as index, maintain a signature fingerprint sequence, expressed as: ; in, The maximum time length for maintaining the current model; (2) Model update: Generate a new signature fingerprint each time After that, the model update operation is: .
[0028] S2 includes: S21, extract the context signature fingerprint generated by the terminal at the current moment and business context drift-aware model Historical fingerprint sequence in One-to-one Hamming distance Calculate, if , it is determined that a business context drift event is currently occurring, where is the drift judgment threshold, The dimension of the context signature fingerprint is expressed as: ; in, For the history Context fingerprint; S22: If drift is determined to have occurred, extract the tuning parameter configuration set corresponding to the historical fingerprint from the business context drift perception model. , and get the currently effective tuning configuration set , configure the tuning parameter set corresponding to the historical fingerprint The currently effective tuning configuration set Compare each item one by one, extract inconsistent or invalid tuning items, and form a set of tuning items to be reconstructed ,in, Respectively The first in the history Tuning parameters, The currently effective Tuning parameters, The currently effective Tuning parameters, For the history The corresponding tuning parameter items in the records; S23, the set to be reconstructed and tuned Each tuning item in Mark as items to be restructured and optimized, including channel allocation parameters (access frequency band, bandwidth index), transmit power control factor, interface binding priority (WiFi / Ethernet interface weight), QoS level parameters (DSCP value, transmission rate level), and output a structured optimization update requirement table , expressed as: .
[0029] S3 includes: S31, update the requirements table for the marked tuning , respectively call the unified configuration interface between the network stack protocol module and the IoT operating system resource management module to reset the parameters, which is expressed as: ; ; in, is the item to be compensated and tuned, For the configuration action corresponding to the network layer, Configuration actions corresponding to the operating system layer; S32, construct the currently tuned context state into a new signature-configuration pair Writing a business context drift-aware model , as a new reference history, where It is the latest tuning configuration set after this round of execution, expressed as: ; S33, based on the current fingerprint The closest to the business context drift awareness model The average distance between historical fingerprints and the tuning success feedback rate, and the dynamic correction of the drift judgment threshold , expressed as: ; ; in, The most recent one in the current fingerprint and business context drift perception model The average Hamming distance of the fingerprints, is the Hamming distance function, is the smoothing coefficient for threshold update, is the original drift judgment threshold, The new threshold after calculation in this round.
[0030] like Figure 2 As shown, a CPE network optimization system is used to implement the above-mentioned CPE network optimization method, including the following modules: Contextual data collection module: Based on the IoT operating system, it collects the service scheduling path, network protocol stack call trace, and resource mapping information of each access terminal in the CPE device, and constructs a contextual signature fingerprint; Context drift perception module: This module stores context signature fingerprints in time series, builds a business context drift perception model, and performs fuzzy matching between the current fingerprint and the historical fingerprint to detect whether there is a business context drift event. Tuning mismatch identification module: After detecting a drift event, it extracts the affected tuning configuration items from the business context drift awareness model and generates the tuning items to be reconstructed. Cross-layer tuning execution module: Based on the tuning items to be reconfigured, it performs compensation and reconstruction of configuration parameters between the network protocol stack and the IoT operating system; Model adaptive update module: Based on the context matching results and policy adaptation feedback, it modifies the drift judgment threshold and updates the business context drift perception model.
[0031] The present invention encompasses any alternatives, modifications, equivalents, and solutions that fall within the spirit and scope of the present invention. To provide a thorough understanding of the present invention, specific details are described in detail below in connection with the preferred embodiments of the present invention, but those skilled in the art will be able to fully understand the present invention without these detailed descriptions. Furthermore, to avoid unnecessary confusion regarding the essence of the present invention, well-known methods, processes, procedures, components, and circuits have not been described in detail.
[0032] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.
Claims
1. A CPE network tuning method, characterized in that: The following steps are involved: S1, based on the IoT operating system, collects the service scheduling path, network protocol stack call trace, and resource mapping information of each access terminal in the CPE device, generates a context signature fingerprint to represent the relationship between the service logic state and the IoT operating system scheduling, and stores the context signature fingerprint in time series as a service context drift perception model; S2 detects service context drift events and identifies tuning mismatch items. It performs fuzzy matching based on the current context signature fingerprint and the historical context sequence in the service context drift perception model to determine whether a service context drift event has occurred. If so, it extracts tuning configuration items affected by drift based on the drift judgment threshold in the service context drift perception model, including channel allocation, power control, interface binding, and QoS level parameters, and marks them as tuning items to be reconfigured. S3 performs cross-layer compensation tuning and updates the business context drift perception model. Based on the tuning items to be reconstructed, configuration parameter compensation and reconstruction operations are performed between the network stack and the IoT operating system. At the same time, based on the current round of context matching results and policy adaptation feedback, the drift judgment threshold in the business context drift model parameters is corrected to form an adaptive and evolving business context drift perception model.
2. A CPE network tuning method according to claim 1, characterized in that: Said S1 comprises: S11, calling the scheduling monitoring interface, protocol stack tracking interface, and resource mapping query interface of the IoT operating system to collect service scheduling paths, network protocol stack call traces, and resource mapping information of each CPE access terminal; S12: Based on the unique identifier of the terminal, the collected service scheduling path, network protocol stack call trace, and resource mapping information are aggregated, and the resource mapping relationship of the terminal within the specified time window is extracted in combination with the status information of the IoT operating system to form a resource usage structure including thread priority, cache binding, and channel occupancy; S13, the business scheduling path, network protocol call trace and resource usage structure corresponding to each terminal are fused and encoded to generate a context signature fingerprint, which is written into the business context drift perception model in chronological order to describe the dynamic relationship between the business logic state and the scheduling behavior of the IoT operating system.
3. A CPE network tuning method according to claim 2, characterized in that: The S11 includes: S111, calling the scheduling monitoring interface of the IoT operating system to collect task scheduling paths for threads bound to each access terminal, including thread identifiers, task scheduling time, priority, and scheduling frequency; S112, calling the protocol stack tracking interface to collect the activity track of the access terminal in the IoT protocol stack, including the protocol call sequence, transmission status code and port resource occupancy time; S113, calling a resource mapping query interface to collect the binding relationship between the terminal thread and the system resources, including the binding relationship between the terminal thread and the processor core, the cache usage status and the IO channel occupancy status.
4. A CPE network tuning method according to claim 3, characterized in that: The S12 includes: S121, based on the unique identifier of each access terminal, the collected service scheduling path, network protocol stack call trace and resource mapping information are aggregated to form a structured data set corresponding to the terminal ; S122, from the generated structured data set corresponding to the terminal Extract resource binding information and set the time window based on the scheduling status in the IoT operating system , filter the effective resource usage records of the terminal in the window to form a resource occupancy filter set ; S123, filter the set based on the filtered resource occupancy The attributes of each item in the ,thread priority structure, cache binding structure, channel occupancy structure are extracted, and the resource usage structure of the current terminal is generated.
5. A CPE network tuning method according to claim 4, characterized in that: The S122 includes: S1221, from the structured data set corresponding to the terminal Extract the resource binding information part and construct a resource record set; S1222, extract the current scheduling queue length from the IoT operating system scheduling status information and the system load index , and according to and Dynamically set resource active time window length; S1223, in the setting Filter resource record entries Start time Is it within the window range? Get the resource occupancy filter set ,in, The current system time.
6. A CPE network tuning method according to claim 5, characterized in that: The S123 includes: S1231, filter set from resource occupancy Filter out resource record items whose resource type is processor core, extract the scheduling priority and occupancy time of the corresponding thread, and build the thread priority structure ; S1232: Filter out the record items whose resource types belong to cache unit resources from the resource occupation filter set, collect statistics on the binding status and cache hit status of each thread on different levels of cache units, and build a cache binding structure. ; S1233: Filter out the record items of resource types belonging to input and output channel resources from the resource occupancy screening set, count the cumulative usage time of each thread on different I / O channels, and build a channel occupancy structure. , while the thread priority structure , cache binding structure , channel occupancy structure Summarize and form a resource usage structure .
7. A CPE network tuning method according to claim 6, characterized in that: The S13 includes: S131, the terminal In the current time slice Business scheduling path , Network protocol call trace and resource usage structure Encoded as a joint context feature vector ; S132, context feature vector Input fingerprint mapping function , generate context signature fingerprint ; S133, the generated context signature fingerprint Time tag As the index, write the business context drift perception model corresponding to the terminal ,constructing contextual trajectories in the form of time series.
8. A CPE network tuning method according to claim 7, characterized in that: The S2 includes: S21, extract the context signature fingerprint generated by the terminal at the current moment and business context drift-aware model Historical fingerprint sequence in One-to-one Hamming distance Calculate, if , it is determined that a business context drift event is currently occurring, where is the drift judgment threshold; S22: If drift is determined to have occurred, extract the tuning parameter configuration set corresponding to the historical fingerprint from the business context drift perception model. , and get the currently effective tuning configuration set , configure the tuning parameter set corresponding to the historical fingerprint The currently effective tuning configuration set Compare each item one by one, extract inconsistent or invalid tuning items, and form a set of tuning items to be reconstructed ,in, Respectively The first in the history Tuning parameters, The currently effective Tuning parameters, The currently effective Tuning parameters, For the history The corresponding tuning parameter items in the records; S23, the set to be reconstructed and tuned Each tuning item in Mark as tuning items to be reconstructed, including channel allocation parameters, transmit power control factors, interface binding priorities, QoS level parameters, and output a structured tuning update requirement table .
9. A CPE network tuning method according to claim 8, characterized in that: The S3 includes: S31, update the requirements table for the marked tuning , respectively, calling the unified configuration interface between the network stack protocol module and the IoT operating system resource management module to reset parameters; S32, construct the currently tuned context state into a new signature-configuration pair Writing a business context drift-aware model , as a new reference history, where The latest tuning configuration set after this round of execution; S33, based on the current fingerprint The closest to the business context drift awareness model The average distance between historical fingerprints and the tuning success feedback rate, and the dynamic correction of the drift judgment threshold .
10. A CPE network tuning system, used to implement a CPE network tuning method according to any one of claims 1 to 9, characterized in that: Includes the following modules: Contextual data collection module: Based on the IoT operating system, it collects the service scheduling path, network protocol stack call trace, and resource mapping information of each access terminal in the CPE device, and constructs a contextual signature fingerprint; Context drift perception module: This module stores context signature fingerprints in time series, builds a business context drift perception model, and performs fuzzy matching between the current fingerprint and the historical fingerprint to detect whether there is a business context drift event. Tuning mismatch identification module: After detecting a drift event, it extracts the affected tuning configuration items from the business context drift awareness model and generates the tuning items to be reconstructed. Cross-layer tuning execution module: Based on the tuning items to be reconfigured, it performs compensation and reconstruction of configuration parameters between the network protocol stack and the IoT operating system; Model adaptive update module: Based on the context matching results and policy adaptation feedback, it modifies the drift judgment threshold and updates the business context drift perception model.