PON gateway fault self-checking method and self-checking system

Abstract

The invention relates to the technical field of optical access network equipment maintenance, and discloses a PON gateway fault self-checking method and a self-checking system, in particular to a PON gateway fault self-checking method integrating task priority scheduling, hardware abstraction layer adaptation, multi-gateway cooperative detection and intelligent reasoning diagnosis. Intelligent upgrading from single equipment detection to network-level fault positioning is realized by constructing a standardized detection instruction system, a time-space aligned cooperative detection network and a fault positioning model based on a Bayesian network.

Description

Technical Field

[0001] The invention relates to the technical field of optical access network equipment maintenance, in particular to a PON gateway fault self-detection method and a self-detection system. Background Art

[0002] With the popularization of fiber-to-the-home (FTTH) networks, the reliability and stability of PON gateways, as the core equipment of optical access networks, are directly related to the service quality of the entire communication network. Traditional PON gateway fault detection methods have the following technical defects:

[0003] Single detection model: Existing solutions often rely on offline detection or manual inspections, which are unable to detect device operating status in real time. When network topology changes dynamically (such as new users joining or link failure switching), traditional detection mechanisms struggle to adapt to fault location requirements in complex network environments. For example, in scenarios with sudden traffic surges, traditional detection methods may not be able to promptly detect service interruptions caused by bottlenecks in gateway processing capacity.

[0004] Lack of a coordinated detection mechanism: Single-gateway self-testing has blind spots and cannot effectively identify fault propagation across gateways. When cascading faults occur between multiple gateways, traditional methods struggle to model the fault propagation path, resulting in fault location accuracy rates below 60%. For example, in a ring topology, a single gateway's optical module failure could affect multiple adjacent gateways via an optical splitter. However, traditional detection methods are unable to correlate and analyze such spatially correlated faults.

[0005] Poor compatibility of detection commands: Different vendors have different PON gateway hardware abstraction layers, and existing detection solutions lack a standardized command encapsulation mechanism, resulting in incompatible detection tools when deployed across platforms. For example, a vendor's gateway may use a specific I2C register mapping scheme, and traditional detection command sets cannot adapt to this proprietary design, resulting in insufficient detection coverage.

[0006] Insufficient intelligent fault diagnosis: Traditional methods rely on manual experience to analyze root causes, lacking intelligent reasoning based on historical data. When faced with intermittent or complex faults, manual troubleshooting efficiency is less than 0.5 times per hour, and the fault recurrence rate is less than 30%. For example, certain timing errors caused by electromagnetic interference have difficulty capturing their characteristic patterns with traditional detection methods.

[0007] Lack of multi-dimensional data fusion: Existing solutions fail to effectively integrate device status data, network topology information, and environmental perception data, resulting in a single-dimensional fault diagnosis. For example, in high-temperature environments, the laser output power of an optical module may drift, but traditional detection systems lack the ability to correlate environmental parameters with device performance. Summary of the Invention

[0008] The purpose of the present invention is to overcome the deficiencies of the prior art and provide a PON gateway fault self-detection method, comprising the following steps:

[0009] In step 1, the self-test task scheduling module generates self-test task attributes by parsing the task priority parameters and device identifiers. If the task attribute is a single-gateway self-test task, it extracts the local fault feature library to generate a basic detection instruction sequence, and then proceeds to step 2. If the task attribute is a multi-gateway collaborative detection task, it calls the cloud-based fault diagnosis server through the communication module to generate a collaborative detection instruction sequence, and then proceeds to step 4.

[0010] Step 2: logically combine the basic detection instruction sequence with the device status monitoring atomic instruction to generate a standardized detection instruction string, wherein the device status monitoring atomic instruction includes an optical power detection instruction, a link layer discovery protocol detection instruction, and a VLAN configuration verification instruction;

[0011] In step 3, the self-test execution container encapsulates the standardized test instruction string into a device compatibility test process container according to the hardware abstraction layer interface specification of the target gateway, injects the test process container into the control management interface of the target gateway through the I2C bus, performs device status collection, protocol consistency verification, and performance threshold comparison operations according to the preset timing, and generates a self-test result data packet; then proceeds to step 7;

[0012] Step 4: Parse the collaborative detection instruction sequence into multi-dimensional detection subtasks and assign the detection subtasks to designated collaborative detection nodes based on the network topology. The collaborative detection nodes include backup detection channels and edge computing units of adjacent gateways.

[0013] Step 5: After each collaborative detection node executes the assigned detection subtask, it generates a local detection report with a timestamp through the time synchronization module. The self-test execution container performs spatiotemporal alignment on the local detection report to construct a global fault correlation matrix.

[0014] Step 6: Based on the device compatibility test results and the global fault correlation matrix, a fault location report is generated using the built-in Bayesian inference engine and uploaded to the network management system via an encrypted channel.

[0015] Step 7: Complete the PON gateway fault self-test.

[0016] Furthermore, the self-check task scheduling module generates self-check task attributes by parsing task priority parameters and device identifiers, including:

[0017] Parse the QoS parameter field in the task information. When the priority is detected as emergency maintenance, it is a multi-gateway collaborative detection task; when it is detected that the device identifier corresponds to the core node gateway, the dual-channel redundant detection mode is forced.

[0018] Furthermore, the generation of the basic detection instruction sequence includes: retrieving a detection template that matches the target gateway model from a local fault feature library, the detection template containing pre-configured detection instruction timing and threshold parameters; dynamically correcting the detection template based on the real-time collected environmental parameters, and adding an optical module heat dissipation efficiency detection instruction when the ambient temperature exceeds the preset temperature threshold.

[0019] Furthermore, the generation of the collaborative detection instruction sequence includes: obtaining a network topology snapshot through the SDN controller to identify gateway devices on the critical path; calculating the optimal detection path based on the Dijkstra algorithm, and generating a collaborative detection instruction sequence including relay detection nodes and hop count limits.

[0020] Furthermore, the generation of a standardized detection instruction string includes: decomposing the optical power detection instruction into a transmitter calibration sub-instruction and a receiver sensitivity test sub-instruction; embedding a dynamic port mapping table in the VLAN configuration verification instruction to support real-time parsing of 802.1Q tags.

[0021] Furthermore, the construction of the global fault association matrix includes: using a weighted directed graph model to describe the detection results, where nodes represent gateway devices and edge weights represent the fault propagation probability; applying the PageRank algorithm to calculate the device fault influence index, and triggering an early warning mechanism when the index exceeds a threshold.

[0022] Furthermore, generating a fault location report includes:

[0023] Perform time series pattern mining on historical fault data and establish an LSTM prediction model. Combine real-time detection data with model prediction results to generate an enhanced diagnostic report that includes fault development trends.

[0024] A PON gateway fault self-detection system, applying the PON gateway fault self-detection method, comprises: a self-detection task scheduling module, a collaborative detection network module, and a cloud-based diagnostic platform; the self-detection task scheduling module is connected to the collaborative detection network module and the cloud-based diagnostic platform respectively;

[0025] Wherein, the self-check task scheduling module is configured with a task priority parsing engine and a hardware abstraction layer interface library;

[0026] The collaborative detection network module is composed of multiple intelligent gateways with edge computing capabilities, each of which has a built-in detection agent module and a time-space synchronization unit;

[0027] The cloud-based diagnostic platform includes a fault feature knowledge graph and an adaptive reasoning engine, and communicates with the PON network through an SDN controller.

[0028] The beneficial effects of the present invention are as follows: by constructing a task priority-driven self-inspection task scheduling mechanism, a standardized detection instruction system, a time-space aligned collaborative detection network, and an intelligent reasoning diagnosis model, the present invention achieves improved PON gateway fault detection efficiency and improved fault location accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 A flowchart of a PON gateway fault self-detection method is provided;

[0030] Figure 2 The figure is a schematic diagram of the principle of a PON gateway fault self-detection system. DETAILED DESCRIPTION

[0031] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings, but the protection scope of the present invention is not limited to the following.

[0032] The features and performance of the present invention are further described in detail below with reference to the embodiments.

[0033] like Figure 1 As shown, in step 1, the self-test task scheduling module generates self-test task attributes by parsing the task priority parameters and device identifiers. If the task attribute is a single-gateway self-test task, the local fault feature library is extracted to generate a basic detection instruction sequence, and the process proceeds to step 2. If the task attribute is a multi-gateway collaborative detection task, the cloud fault diagnosis server is called through the communication module to generate a collaborative detection instruction sequence, and the process proceeds to step 4.

[0034] Step 2: logically combine the basic detection instruction sequence with the device status monitoring atomic instruction to generate a standardized detection instruction string, wherein the device status monitoring atomic instruction includes an optical power detection instruction, a link layer discovery protocol detection instruction, and a VLAN configuration verification instruction;

[0035] In step 3, the self-test execution container encapsulates the standardized test instruction string into a device compatibility test process container according to the hardware abstraction layer interface specification of the target gateway, injects the test process container into the control management interface of the target gateway through the I2C bus, performs device status collection, protocol consistency verification, and performance threshold comparison operations according to the preset timing, and generates a self-test result data packet; then proceeds to step 7;

[0036] Step 4: Parse the collaborative detection instruction sequence into multi-dimensional detection subtasks and assign the detection subtasks to designated collaborative detection nodes based on the network topology. The collaborative detection nodes include backup detection channels and edge computing units of adjacent gateways.

[0037] Step 5: After each collaborative detection node executes the assigned detection subtask, it generates a local detection report with a timestamp through the time synchronization module. The self-test execution container performs spatiotemporal alignment on the local detection report to construct a global fault correlation matrix.

[0038] Step 6: Based on the device compatibility test results and the global fault correlation matrix, a fault location report is generated using the built-in Bayesian inference engine and uploaded to the network management system via an encrypted channel.

[0039] Step 7: Complete the PON gateway fault self-test.

[0040] Specifically include:

[0041] Self-check task scheduling and instruction sequence generation

[0042] The self-test task scheduling module generates self-test task attributes by parsing task priority parameters and device identifiers, including:

[0043] Priority resolution mechanism: parses the QoS parameter field in the task information. When the priority mark "urgent maintenance" is detected, the multi-gateway collaborative detection task is automatically triggered. When the device identifier corresponds to the core node gateway, the dual-channel redundant detection mode is forcibly enabled to ensure the detection reliability of key nodes.

[0044] Dynamic instruction generation:

[0045] Single-gateway self-test scenario: A detection template matching the target gateway model is retrieved from the local fault signature database. This template contains preconfigured detection instruction timing and threshold parameters. The detection template is dynamically modified based on real-time collected environmental parameters (temperature, humidity). For example, when the ambient temperature exceeds a preset temperature threshold (such as 55°C), an instruction to test the heat dissipation efficiency of the optical module is automatically added.

[0046] Multi-gateway collaboration scenarios: The SDN controller obtains a network topology snapshot and identifies gateways on critical paths. The optimal detection path is calculated using the Dijkstra algorithm, generating a collaborative detection instruction sequence that includes relay detection nodes and hop limits. For example, in a ring topology, gateways on the shortest path are prioritized as relay nodes.

[0047] Standardized detection instruction string generation

[0048] Logically combine the basic detection instruction sequence with the equipment status monitoring atomic instructions to generate a standardized detection instruction string:

[0049] Instruction decomposition and expansion:

[0050] The optical power detection command is decomposed into the transmitter calibration sub-command (including laser drive current and modulation depth parameters) and the receiver sensitivity test sub-command (supporting a dynamic range of -30dBm to 0dBm)

[0051] Link Layer Discovery Protocol (LLDP) detection commands integrate topology discovery and neighbor device information analysis functions

[0052] VLAN configuration verification instructions are embedded in the dynamic port mapping table, supporting real-time parsing of 802.1Q tags (parsing delay <50ms)

[0053] Hardware Abstraction Layer (HAL) adaptation: By parsing the target gateway's Hardware Abstraction Layer (HAL) interface specification, standardized commands are mapped to specific register operation sequences. For example, for a Broadcom chipset gateway, optical power detection commands are converted into read and write operations on the 0xC0 / 0xC1 registers on the I2C bus.

[0054] Device compatibility test process execution

[0055] The self-test execution container encapsulates the standardized test instruction string into a device compatibility test process container according to the HAL interface specification, specifically including:

[0056] Containerized encapsulation: Build a detection process container that includes an instruction execution sequence, timing constraints (such as optical power detection must be executed within 500ms after the link is stable), and an exception handling mechanism.

[0057] Bus protocol adaptation: Inject the detection process container into the control and management interface of the target gateway through the I2C bus, and support the SMBus protocol extended instruction set (such as the 0x80-0x8F private command space).

[0058] Multi-dimensional detection execution:

[0059] Device status collection: periodically collects optical module parameters (temperature, bias current, transmit power) and MAC layer statistics (CRC error count, packet loss rate);

[0060] Protocol consistency verification: perform LLDP neighbor discovery test and STP protocol state machine verification;

[0061] Performance threshold comparison: compares the real-time collected parameters with the pre-configured threshold table (e.g., optical receiver sensitivity threshold error <±1dBm);

[0062] Collaborative testing task allocation and execution

[0063] When the task attribute is a multi-gateway collaborative detection task:

[0064] Task decomposition and allocation: The collaborative detection instruction sequence is parsed into multi-dimensional detection subtasks (such as link quality detection and cross-gateway protocol interaction testing). These subtasks are then assigned to designated collaborative detection nodes based on the network topology. Collaborative detection nodes contain backup detection channels from adjacent gateways and edge computing units (such as NXP i.MX8M Mini processors).

[0065] Time and space synchronization mechanism: Each collaborative detection node achieves sub-microsecond time synchronization through the IEEE 1588v2 protocol and generates a local detection report with a timestamp (timestamp accuracy <100ns).

[0066] Global fault correlation matrix construction

[0067] The self-test execution container performs spatiotemporal alignment on local test reports to construct a global fault correlation matrix:

[0068] Weighted directed graph modeling: A weighted directed graph model is used to describe the detection results. The nodes represent gateway devices, and the edge weights represent the fault propagation probability (obtained through training of historical fault data).

[0069] Impact calculation: The PageRank algorithm is used to calculate the impact index of device failures. When the index exceeds a preset threshold (such as 0.7), an early warning mechanism is triggered. For example, the impact index of a core node gateway failure is generally higher than that of an edge node.

[0070] Fault location report generation

[0071] The built-in Bayesian inference engine generates a fault location report based on the device compatibility test results and the global fault correlation matrix:

[0072] Time series pattern mining: Perform time series pattern mining on historical fault data and establish an LSTM prediction model (training data set size > 100,000).

[0073] Enhanced diagnostic reports: Combine real-time detection data with model prediction results to generate enhanced diagnostic reports that include fault development trends (such as remaining life prediction and fault propagation rate). For example, if a continuous rise in optical module temperature is detected, the report will include a warning message that "optical power attenuation fault may occur within 3 days."

[0074] Reporting of self-inspection results

[0075] The fault location report is uploaded to the network management system via an AES-256 encrypted channel, supporting two-way data interaction between northbound interfaces (such as RESTful API) and southbound interfaces (such as SNMP Trap).

[0076] A PON gateway fault self-detection system

[0077] Includes the following core modules:

[0078] Self-check task scheduling module

[0079] Task priority parsing engine: supports IEEE 802.1Qat priority tag parsing and has a built-in priority-detection mode mapping table (e.g., priority 7 corresponds to dual-channel redundant detection).

[0080] Hardware Abstraction Layer Interface Library: Contains HAL interface specifications for gateways from mainstream manufacturers (such as Huawei MA5683T and ZTE C300), and supports dynamic loading of extended interface drivers.

[0081] Detection template database: Stores detection templates for more than 200 gateway models and supports template version management based on JSON Schema.

[0082] 2. Collaborative detection network module

[0083] Intelligent gateway node: Each gateway has a built-in detection agent module (based on the Containerd container runtime) and a time and space synchronization unit (supporting the PTPv2 protocol).

[0084] Edge computing unit: uses the NVIDIA Jetson AGX Xavier platform, provides 16GB of memory and 32TOPS of computing power, and supports localized fault feature extraction.

[0085] Detection channel redundancy: Each gateway is configured with two primary and backup detection channels (the primary channel is I2C and the backup channel is SPI), and the channel switching time is <50ms.

[0086] 3. Cloud diagnostic platform

[0087] Bayesian reasoning engine: implemented using the PyMC3 framework, supports fault propagation probability calculation for 100,000 nodes.

[0088] LSTM prediction model: deployed on a TensorFlow Serving cluster, supporting millisecond-level real-time inference (QPS > 1000).

[0089] Visualization interface: Built based on Grafana, it provides functions such as fault heat map and 3D display of propagation path.

[0090] Example 1: Single Gateway Optical Module Fault Self-Test

[0091] In a carrier's network, an MA5683T OLT gateway experienced service interruption in a high-temperature environment, requiring rapid locating of the fault's root cause.

[0092] Implementation steps:

[0093] Task scheduling: The self-test task scheduling module resolves the device identifier to the core node, and the ambient temperature sensor reports 58°C (>55°C threshold); it generates an enhanced basic test instruction sequence that includes the optical module heat dissipation efficiency test instruction.

[0094] Instruction encapsulation: Decomposes the optical power detection instruction into a transmitter calibration sub-instruction (0xC0 register operation) and a receiver sensitivity test sub-instruction (0xC1 register operation); adds a temperature compensation factor (for every 1°C increase in temperature, the optical power detection threshold increases by 0.05dBm);

[0095] Detection execution: Inject the detection process container into the gateway control and management interface through the I2C bus.

[0096] Execution timing:

[0097] T0: Start optical module temperature monitoring (sampling period 100ms)

[0098] T+500ms: Perform transmitter calibration (drive current adjustment range 20-80mA)

[0099] T+1000ms: Perform receiver sensitivity test (-28dBm→0dBm step scan)

[0100] T+1500ms: Perform cooling performance test (fan speed, heat sink temperature)

[0101] Result analysis: The optical received power was detected to have drifted from -25dBm to -28dBm (exceeding the threshold of -26dBm); the heat dissipation efficiency test showed that the fan speed was 30% lower than the rated value;

[0102] Generate a fault location report: "The performance of the optical module LD chip has degraded. It is recommended to replace the module and check the cooling system."

[0103] Implementation results: Fault location time was shortened from 4 hours with the traditional solution to 18 minutes. Maintenance personnel arrived at the site with the correct spare parts, and the first-time repair rate increased to 100%.

[0104] Example 2: Collaborative detection of multi-gateway cross-device protocol faults

[0105] In a campus network, ONU gateway A and ONU gateway B experience service interruptions simultaneously. It is necessary to troubleshoot cross-device protocol interaction issues.

[0106] Implementation steps:

[0107] Task scheduling: Detecting that the QoS priority is marked as "urgent maintenance" triggers a multi-gateway collaborative detection task. The SDN controller obtains the network topology and identifies that gateways A and B are located in different branches under the same PON port.

[0108] Task Assignment:

[0109] Divided into 3 subtasks:

[0110] Subtask 1: OLT-side protocol consistency detection (assigned to the OLT gateway);

[0111] Subtask 2: A→OLT link quality detection (assigned to gateway A);

[0112] Subtask 3: B→OLT link quality detection (assigned to B gateway);

[0113] The optimal detection paths are determined based on the Dijkstra algorithm: OLT→A (hop number 1) and OLT→B (hop number 1).

[0114] Collaborative testing:

[0115] Each node performs IEEE 1588v2 time synchronization and generates a timestamp-containing test report. Gateway A detects an LLDP neighbor discovery timeout (response delay > 200ms). Gateway B detects a VLAN tag stripping anomaly (802.1Q tag processing error).

[0116] Correlation analysis: A global fault correlation matrix was constructed, revealing that a VLAN configuration error on the OLT side caused cross-device protocol interaction failure. The fault impact index was calculated as follows: OLT = 0.85, A = 0.42, B = 0.38.

[0117] Report Generation:

[0118] Based on historical data, it was found that this OLT model had a VLAN configuration software defect;

[0119] Generate an enhanced diagnostic report: "OLT software version V3.2.1 has a VLAN processing bug. Upgrading to V3.2.3 is recommended."

[0120] Implementation Results: While traditional solutions require device-by-device troubleshooting (taking approximately six hours), this solution located the software defect within 45 minutes through collaborative testing, avoiding widespread network outages.

Claims

1. A PON gateway fault self-detection method, characterized in that: The steps include: In step 1, the self-test task scheduling module generates self-test task attributes by parsing the task priority parameters and device identifiers. If the task attribute is a single-gateway self-test task, it extracts the local fault feature library to generate a basic detection instruction sequence, and then proceeds to step 2. If the task attribute is a multi-gateway collaborative detection task, it calls the cloud-based fault diagnosis server through the communication module to generate a collaborative detection instruction sequence, and then proceeds to step 4. Step 2: logically combine the basic detection instruction sequence with the device status monitoring atomic instruction to generate a standardized detection instruction string, wherein the device status monitoring atomic instruction includes an optical power detection instruction, a link layer discovery protocol detection instruction, and a VLAN configuration verification instruction; Step 3: The self-test execution container encapsulates the standardized test instruction string into a device compatibility test process container according to the hardware abstraction layer interface specification of the target gateway. The test process container is injected into the control and management interface of the target gateway through the I2C bus. The device status collection, protocol consistency verification, and performance threshold comparison operations are performed according to the preset timing to generate a self-test result data packet. Go to step seven; Step 4: Parse the collaborative detection instruction sequence into multi-dimensional detection subtasks and assign the detection subtasks to designated collaborative detection nodes based on the network topology. The collaborative detection nodes include backup detection channels and edge computing units of adjacent gateways. Step 5: After each collaborative detection node executes the assigned detection subtask, it generates a local detection report with a timestamp through the time synchronization module. The self-test execution container performs spatiotemporal alignment on the local detection report to construct a global fault correlation matrix. Step 6: Based on the device compatibility test results and the global fault correlation matrix, a fault location report is generated using the built-in Bayesian inference engine and uploaded to the network management system via an encrypted channel. Step 7: Complete the PON gateway fault self-test.

2. A PON gateway fault self-detection method according to claim 1, characterized in that: The self-check task scheduling module generates self-check task attributes by parsing task priority parameters and device identifiers, including: Parse the QoS parameter field in the task information. When the priority is detected as emergency maintenance, it is a multi-gateway collaborative detection task; when it is detected that the device identifier corresponds to the core node gateway, the dual-channel redundant detection mode is forced.

3. The PON gateway fault self-detection method according to claim 2, characterized in that: The generation of the basic detection instruction sequence includes: retrieving a detection template that matches the target gateway model from a local fault feature library, wherein the detection template includes pre-configured detection instruction timing and threshold parameters; dynamically correcting the detection template based on real-time collected environmental parameters, and adding an optical module heat dissipation efficiency detection instruction when the ambient temperature exceeds a preset temperature threshold.

4. A PON gateway fault self-detection method according to claim 3, characterized in that: The method of generating a collaborative detection instruction sequence includes: obtaining a network topology snapshot through an SDN controller to identify gateway devices on a critical path; calculating an optimal detection path based on a Dijkstra algorithm to generate a collaborative detection instruction sequence including relay detection nodes and a hop limit.

5. A PON gateway fault self-detection method according to claim 3, characterized in that: The generation of the standardized detection instruction string includes: decomposing the optical power detection instruction into a transmitter calibration sub-instruction and a receiver sensitivity test sub-instruction; embedding a dynamic port mapping table in the VLAN configuration verification instruction to support real-time analysis of 802.1Q tags.

6. A PON gateway fault self-detection method according to claim 5, characterized in that: The global fault association matrix is ​​constructed, including: using a weighted directed graph model to describe detection results, where nodes represent gateway devices and edge weights represent fault propagation probabilities; applying a PageRank algorithm to calculate a device fault influence index, and triggering an early warning mechanism when the index exceeds a threshold.

7. A PON gateway fault self-detection method according to claim 6, characterized in that: Generating a fault location report includes: Perform time series pattern mining on historical fault data and establish an LSTM prediction model. Combine real-time detection data with model prediction results to generate an enhanced diagnostic report that includes fault development trends.

8. A PON gateway fault self-diagnosis system, characterized in that: A PON gateway fault self-detection method according to any one of claims 1 to 7 is applied, comprising: a self-detection task scheduling module, a collaborative detection network module, and a cloud-based diagnostic platform; the self-detection task scheduling module is connected to the collaborative detection network module and the cloud-based diagnostic platform respectively; Wherein, the self-check task scheduling module is configured with a task priority parsing engine and a hardware abstraction layer interface library; The collaborative detection network module is composed of multiple intelligent gateways with edge computing capabilities, each of which has a built-in detection agent module and a time-space synchronization unit; The cloud-based diagnostic platform includes a fault feature knowledge graph and an adaptive reasoning engine, and communicates with the PON network through an SDN controller.

Images