Device control method, device control apparatus, device, storage medium, computer program product

Abstract

The application discloses a device control method and device, equipment, storage medium and computer program product, and is used to solve the problem that the user plane device running state and service load lack dynamic matching ability under the existing device scheduling mechanism, resulting in device idling and high energy consumption at low load. The method comprises the following steps: acquiring energy efficiency index data and service traffic data corresponding to each user plane device in a device pool; determining predicted traffic in a future preset time window based on a pre-trained prediction model according to the service traffic data; determining a control strategy for at least one user plane device in the device pool according to the predicted traffic and the energy efficiency index data, wherein the control strategy is used to control the user plane device to enter a sleep state; generating a control instruction according to the control strategy and sending the control instruction to the user plane device to control the user plane device to perform a corresponding sleep operation.

Description

Technical Field

[0001] This application relates to the field of core network technology, and in particular to a device control method, apparatus, equipment, storage medium, and computer program product. Background Technology

[0002] As communication networks develop towards higher density and intelligence, core network user plane equipment, such as SAE-GW in 4G, UPF in 5G, and SBC / UMF in the IMS domain, are usually deployed in a resource pool (POOL) manner, and the control plane network element (Session Management Function, SMF) performs unified session management and load scheduling for the equipment in the pool.

[0003] In actual operation, network traffic exhibits significant periodic variations. For example, during nighttime or holidays, the overall network load is significantly lower than during daytime peak hours. However, under the existing scheduling mechanism, there is a lack of effective dynamic matching between the operating status of user plane devices within the resource pool and real-time traffic volume. This results in devices continuously consuming large amounts of power under low load conditions, leading to high overall network energy consumption. Furthermore, the hardware resource scheduling method is relatively fixed, making it difficult to fine-tune the operating status of internal hardware modules based on real-time changes in traffic volume, thus failing to achieve on-demand resource utilization.

[0004] Therefore, how to achieve dynamic adaptation between the operating status of user plane devices and business load while ensuring the quality of business services has become a technical problem that needs to be solved by existing technologies. Summary of the Invention

[0005] This application provides a device control method to solve the problem that under the existing device scheduling mechanism, the operating status of user plane devices lacks dynamic matching capability with service load, resulting in device idling and high energy consumption under low load.

[0006] This application also provides a device control apparatus to solve the problem that under the existing device scheduling mechanism, the user plane device operating status and service load lack dynamic matching capability, resulting in device idling and high energy consumption under low load.

[0007] This application also provides a device control device to solve the problem that under the existing device scheduling mechanism, the user plane device operating status and service load lack dynamic matching ability, resulting in device idling and high energy consumption under low load.

[0008] This application also provides a computer-readable storage medium to solve the problem that under the existing device scheduling mechanism, the user plane device operating status and service load lack dynamic matching capability, resulting in device idling and high energy consumption under low load.

[0009] A computer program product designed to address the problem that under the existing equipment scheduling mechanism, the lack of dynamic matching between the operating status of user plane equipment and business load leads to equipment idling and high energy consumption during low load periods.

[0010] The embodiments of this application adopt the following technical solutions: A device control method includes: acquiring energy efficiency index data and service traffic data corresponding to each user plane device in a device pool; determining predicted traffic within a preset time window based on the service traffic data and a pre-trained prediction model; determining a control strategy for at least one user plane device in the device pool based on the predicted traffic and the energy efficiency index data, wherein the control strategy is used to control the user plane device to enter a sleep state; generating a control command based on the control strategy and sending the control command to the user plane device to control the user plane device to perform a corresponding sleep operation.

[0011] A device control apparatus includes: a data acquisition unit for acquiring energy efficiency index data and service traffic data corresponding to each user plane device in a device pool; a prediction unit for determining predicted traffic within a preset time window based on the service traffic data and a pre-trained prediction model; a strategy generation unit for determining a control strategy for at least one user plane device in the device pool based on the predicted traffic and the energy efficiency index data, wherein the control strategy is used to control the user plane device to enter a sleep state; and a control unit for generating control commands according to the control strategy and sending the control commands to the user plane device to control the user plane device to perform a corresponding sleep operation.

[0012] A device control device, comprising: The processor; and a memory configured to store computer-executable instructions, which, when executed, cause the processor to perform the following operations: acquire energy efficiency index data and service traffic data corresponding to each user plane device in the device pool; determine the predicted traffic within a preset time window based on the service traffic data and a pre-trained prediction model; determine a control strategy for at least one user plane device in the device pool based on the predicted traffic and the energy efficiency index data, wherein the control strategy is used to control the user plane device to enter a sleep state; generate a control instruction based on the control strategy, and send the control instruction to the user plane device to control the user plane device to perform a corresponding sleep operation.

[0013] A computer-readable storage medium stores one or more programs that, when executed by an electronic device including multiple applications, cause the electronic device to perform the following operations: acquire energy efficiency index data and service traffic data corresponding to each user plane device in a device pool; determine predicted traffic within a future preset time window based on the service traffic data and a pre-trained prediction model; determine a control strategy for at least one user plane device in the device pool based on the predicted traffic and the energy efficiency index data, wherein the control strategy is used to control the user plane device to enter a sleep state; generate a control command according to the control strategy, and send the control command to the user plane device to control the user plane device to perform a corresponding sleep operation.

[0014] A computer program product includes a computer program that, when executed by a processor, performs the following: acquiring energy efficiency index data and service traffic data corresponding to each user plane device in a device pool; determining predicted traffic within a future preset time window based on the service traffic data and a pre-trained prediction model; determining a control strategy for at least one user plane device in the device pool based on the predicted traffic and the energy efficiency index data, wherein the control strategy is used to control the user plane device to enter a sleep state; generating a control command based on the control strategy and sending the control command to the user plane device to control the user plane device to perform a corresponding sleep operation.

[0015] The above-described technical solutions adopted in the embodiments of this application can achieve the following beneficial effects: The device control method provided in this application first obtains energy efficiency index data and service traffic data corresponding to each user plane device in the device pool. Then, based on the service traffic data and a pre-trained prediction model, the predicted traffic within a preset time window is determined. Based on the predicted traffic and the energy efficiency index data, a control strategy is determined for at least one user plane device in the device pool. According to the control strategy, a control command is generated and sent to the user plane device to control the user plane device to perform a corresponding hibernation operation. The load control method provided in this application has two advantages. First, by acquiring energy efficiency index data and service traffic data of each user plane device in the device pool, it achieves real-time dynamic perception of device operating status and network load. Furthermore, based on this, a pre-trained prediction model is used to predict service traffic within a preset time window, overcoming the shortcomings of existing technologies that rely solely on static configuration or current load for scheduling and cannot anticipate service changes. Second, the method provided in this application can determine a control strategy for at least one user plane device in the device pool to enter a sleep state based on predicted traffic and energy efficiency index data. This achieves dynamic matching between device operating status and service load. Compared to the shortcomings of existing technologies where all devices still need to remain operational during low-load periods, resulting in significant idling energy consumption, the device control method provided in this application can adaptively put some devices into a sleep state based on predicted service volume trends while ensuring service carrying capacity, effectively reducing overall energy consumption during low-load periods. Attached Figure Description

[0016] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a schematic diagram of the specific structure of a device control system provided in an embodiment of this application; Figure 2 This is a schematic flowchart illustrating a device control method provided in an embodiment of this application. Figure 3 This is a schematic flowchart illustrating a hibernation control method provided in an embodiment of this application. Figure 4 This is a schematic diagram illustrating a specific process of a session migration method provided in an embodiment of this application; Figure 5 This is a schematic diagram illustrating a specific process of a device wake-up method provided in an embodiment of this application; Figure 6 This is a schematic diagram of the specific structure of a device control apparatus provided in an embodiment of this application; Figure 7 This is a schematic diagram of the specific structure of a device control device provided in an embodiment of this application. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0018] This application provides a device control method to solve the problem that under the existing device scheduling mechanism, the user plane device operating status and service load lack dynamic matching ability, resulting in device idling and high energy consumption under low load.

[0019] The execution subject of the device control method provided in this application embodiment may be, but is not limited to, at least one of a load scheduling server, a device management server, and a operation monitoring server; in addition, the execution subject of the method may also be the system or application (APP) itself running on these servers.

[0020] For ease of description, the following description uses the example of an equipment control system as the executing entity of this method. It should be understood that using an equipment control system as the executing entity is merely an illustrative example and should not be construed as a limitation of the method.

[0021] This application describes the User Plane Function (UPF) in the 5G core network as an example. Those skilled in the art will understand that the methods provided in this application are also applicable to the system architecture evolution gateway, session boundary controller and user plane management function in the IP multimedia subsystem domain in the 4G core network, as well as the user plane equipment architecture in the subsequent 6G and later networks.

[0022] In one implementation, the specific system architecture of the device control system is as follows: Figure 1 As shown, the architecture mainly includes: 1. Data Acquisition Layer: In this embodiment, the data acquisition layer can use Telemetry technology to monitor the energy efficiency indicators of the User Plane Function (UPF) cluster in real time. The energy efficiency indicators may include, but are not limited to: service indicators, such as throughput, number of sessions, and QoS level; energy consumption indicators, such as dynamic power consumption of the central processing unit / graphics processor and memory bandwidth utilization; and device status, such as chip temperature, fan speed, and power mode.

[0023] 2. Intelligent Decision Layer: In this embodiment, the intelligent decision layer can use a Long Short-Term Memory (LSTM) predictive model to drive shallow sleep / deep sleep hibernation hierarchical control, thereby achieving hierarchical dynamic power consumption adjustment.

[0024] 3. Execution Control Layer: Specifically, the execution control layer can prioritize scheduling high-energy-efficiency nodes through an energy efficiency weighting algorithm, and combine it with GTP-U hot migration technology to ensure lossless service switching and ensure that service switching latency is controlled within the millisecond range.

[0025] 4. Cross-domain collaboration layer: In this embodiment, the cross-domain collaboration layer can link control plane devices (such as Session Management Function (SMF) and transmission controllers (such as Sliced ​​Packet Network (SPN) controllers) to achieve joint energy-saving control of the core network and transmission network.

[0026] Based on the aforementioned equipment control system, a schematic diagram illustrating the specific implementation flow of the equipment control method provided in this application is shown below. Figure 2 As shown, the main steps include the following: Step 11: Obtain energy efficiency index data and service traffic data corresponding to each user plane device in the device pool; In this embodiment, the device pool may include multiple UPFs, which are deployed in a POOL networking mode. Specifically, the device control system can collect energy efficiency index data and service traffic data corresponding to each user plane device in the device pool through an energy efficiency monitoring agent deployed in the UPF cluster.

[0027] In one implementation, the monitoring indicators collected in real time by the energy efficiency monitoring agent may include, but are not limited to, the following: 1. Business metrics, such as throughput, number of sessions, quality of service (QoS) level (e.g., Ultra-Reliable Low Latency Communication (URLLC), enhanced mobile broadband (eMBB), etc.) and packet loss rate.

[0028] 2. Energy consumption indicators, such as: CPU / GPU dynamic power consumption collected by the RAPL interface through the average power limit, memory bandwidth utilization collected by the performance monitoring counter (PMC), and device-level power consumption collected by the power distribution unit (PDU).

[0029] 3. Device status, which may include chip temperature, fan speed, and power mode.

[0030] It should be noted that, in this embodiment of the application, the collection frequency can be dynamically adjusted in real time according to the type of collected indicators and actual business needs. The collection granularity for key indicators can be set to the second level, while the collection granularity for auxiliary indicators can be set to the minute level. Furthermore, the collection frequency can be adjusted to a higher level during periods of rapid changes in business volume, while the collection frequency can be appropriately reduced during periods of stable business to save resources.

[0031] In this embodiment of the application, the energy efficiency monitoring agent can also use telemetry technology to send the collected data to the energy efficiency analysis engine and the LSTM traffic prediction model module in real time and continuously in a streaming manner.

[0032] Step 12: Based on the business traffic data obtained by executing Step 11, determine the predicted traffic within the future preset time window based on the pre-trained prediction model.

[0033] After acquiring the service traffic data corresponding to each user plane device, the device control system can determine the predicted traffic within a preset time window based on this service traffic data and a pre-trained prediction model. This preset time window can be flexibly set according to specific service needs; for example, the time window can be set to the next hour, with a time resolution of 5 minutes, thus allowing for the output of 12 predicted time points within a single prediction time window.

[0034] In one implementation, the prediction model can be a traffic prediction model built on a Long Short-Term Memory (LSTM) network. Specifically, in the embodiments of this application, the prediction model can be trained according to the following sub-steps, including: Sub-step 1201: Generate a training dataset based on the acquired historical business traffic data, business flow characteristics, network status indicators, and external event data; In this embodiment of the application, in order to accurately predict the traffic flow within a preset time window, it is necessary to construct multi-dimensional training data as input data for model training.

[0035] In one implementation, the training dataset may specifically include data in the following dimensions: 1. Historical business traffic data: For example, it may include UPF throughput (Gbps) over 72 hours, peak / average number of sessions per hour, and traffic time distribution. In the application embodiment, the above-mentioned historical service traffic data can be collected at a collection frequency of minutes. At the same time, the collected historical service traffic data can be preprocessed, such as standardization or missing value imputation.

[0036] 2. Real-time business flow characteristics: In one embodiment, the device control system can collect service flow characteristics at a collection frequency of seconds. The collected service flow characteristics may include, but are not limited to, protocol type distribution (i.e., TCP / UDP / ICMP ratio), average packet size, QoS level distribution, etc.

[0037] 3. Network status indicators: In this embodiment of the application, the device control system can collect network status indicators at a sampling frequency of seconds. The collected network status indicators may include, but are not limited to, UPF node load rate, transmission network latency, and average RSRP / SNR values ​​on the wireless side.

[0038] 4. External event data: In this embodiment of the application, the equipment control system can collect external event data when an event is triggered. The collected external event data may include, but is not limited to, large-scale event schedules (such as sports events / live broadcasts), weather data (rainfall / wind speed), and holidays.

[0039] During the feature engineering phase, the multi-dimensional data collected above can be processed in the following ways: 1. Multi-scale time alignment: For example, data collected at a second-level sampling period can be downsampled to a 5-minute sampling period, and event data can be converted into 0 / 1 flags, such as "1 during major events" and "0 during non-major events".

[0040] 2. Spatiotemporal correlation construction: For example, a lag feature (lag=12, i.e., data from the previous hour) can be added, and a periodic feature can be generated to capture the periodicity and trend of traffic data.

[0041] 3. Dynamic feature filtering: In this embodiment, feature importance can be evaluated based on mutual information, and top-ranked features can be retained to reduce model complexity and the risk of overfitting.

[0042] Sub-step 1202: Train the initial prediction model based on the training dataset; After completing the collection and preprocessing of training data by executing sub-step 1201, the initial Long Short-Term Memory (LSTM) network model is trained using the training dataset.

[0043] In this embodiment of the application, the mathematical construction and training process of the prediction model is as follows, including: Process 1: Mathematical representation of input data; In this embodiment, assuming the input data contains D features, such as traffic, protocol type, QoS level, etc., and the historical window length is T (assuming 72 hours, time resolution of 5 minutes, and a total of 864 time steps), then the input tensor is as shown in the following formula [1]: [1] in, This represents the feature vector at time step t, which has been standardized, encoded, and processed for missing values.

[0044] Step 2: LSTM time series modeling; In this embodiment of the application, for each time step t, the calculation formulas for the input gate, forget gate, candidate cell state, cell state, output gate and hidden state of the initial LSTM model are as follows: Formulas [2] to [7]: Input gate i t : [2] Forgotten Gate f t : [3] Candidate cell state : [4] Cell state c t : [5] Output gate o t : [6] Hidden state h t : [7] in, It is a sigmoid function, and the tanh function is the hyperbolic tangent function. This is element-wise multiplication. i w f w c w o Here, H is the weight matrix, and b is the hidden layer dimension. i b fb c b o This is a bias term.

[0045] Process 3: Hybrid attention mechanism; It should be noted that, in order to improve the predictive model's ability to focus on key time points and key features, a hybrid attention mechanism, including spatiotemporal attention and feature attention, is introduced in this embodiment.

[0046] Among them, the spatiotemporal attention mechanism is used to process the output sequence of the LSTM. The attention weights for each time step are calculated according to the following formulas [8] and [9]: [8] [9] in, , These are learnable parameters.

[0047] By using a feature attention mechanism, the input feature x is... t Dynamic weighting is performed as shown in the following formulas

[10] and

[11] :

[10]

[11] The weighted eigenvectors are shown in the following formula

[12] :

[12] By combining spatiotemporal attention with feature attention, a context vector is generated, as shown in the following formulas

[13] and

[14] :

[13]

[14] The final context vector is shown in the following formula

[15] :

[15] Step 4: Set the quantile loss function: In this embodiment, the prediction model is trained using a quantile loss function to output a confidence interval for the predicted flow. For each prediction time point t (a total of 12 points) and quantile τ... For {0.1, 0.5, 0.9}, the loss function is defined as shown in the following formula

[16] :

[16] The total loss is the weighted sum of all quantiles and prediction points, as shown in the following formula

[17] :

[17] Step 5: Online Learning To enable the model to adapt to dynamic changes in the network, this embodiment uses an online learning mechanism to incrementally update the model parameters. The FTRL (Follow-the-Regularized-Leader) optimizer is used, and the update rule is shown in the following formula

[18] :

[18] Where η is the learning rate, α [0, 1] controls the weight of new and old data; for example, α can be set to 0.7. and These are the gradients for the new data and the old data, respectively.

[0048] Step 6: Output prediction; In this embodiment of the application, the output of the prediction model is shown in the following formula

[19] :

[19] This output represents the predicted traffic flow for the next hour (12 time points) and its confidence interval. This is the median prediction. and These are the P10 and P90 quantiles, respectively. Based on the above prediction results, the prediction results shown in Table 1 below can be generated: Table 1 Prediction Results

[0049] Step 13: Based on the predicted flow rate and energy efficiency index data obtained by performing Step 12, determine the control strategy for at least one of the user plane devices in the device pool. After obtaining the predicted flow rate and energy efficiency data, the equipment control system can determine a control strategy for at least one user plane device within the equipment pool. This control strategy is used to control the user plane device to enter a sleep state, including tiered sleep states such as standby-lite or deep sleep.

[0050] In one implementation, the device control system may specifically determine a control strategy for at least one user plane device within the device pool according to the following sub-steps: Sub-step 1301: Based on the predicted traffic, determine the first number of user plane devices that the device pool needs to run within the future preset time window; Specifically, based on the predicted traffic output by the prediction model, the total amount of traffic that needs to be carried within a future preset time window (such as the next hour) can be estimated. Combined with the carrying capacity of a single User Plane Function (UPF) within the equipment pool (e.g., the rated throughput capacity of a single UPF), the initial number of user plane devices required to maintain normal service operation can be determined.

[0051] For example, assuming there are 6 UPFs in the device pool, the total traffic volume is divided according to the UPF carrying capacity. When the traffic volume corresponding to the predicted traffic reaches 100% of the total capacity, 6 UPFs need to run; when the traffic volume is 80% of the total capacity, 5 UPFs need to run; when the traffic volume is 60% of the total capacity, 4 UPFs need to run; when the traffic volume is 40% of the total capacity, 3 UPFs need to run; and when the traffic volume is 20% or less, 2 UPFs need to run.

[0052] Sub-step 1302: Determine the devices to be hibernated based on the second number and the first number of user plane devices currently running in the device pool.

[0053] Specifically, the second number of user plane devices currently running in the device pool can be compared with the first number required to run. If the second number is greater than the first number, it indicates that there are redundant running devices, and the excess devices can be identified as devices to be put into hibernation.

[0054] For example, assuming there are currently 6 UPFs in operation in the device pool, and the initial number required to operate in the future is determined to be 4 based on predicted traffic, then 2 UPFs can be identified as devices to be put into hibernation. When specifically selecting devices to be put into hibernation, the selection can be based on the energy efficiency ratio (η) of each UPF, prioritizing the UPFs with lower energy efficiency ratios for hibernation, so that business can be concentrated on high-energy-efficiency UPFs.

[0055] Sub-step 1303: Based on the operating status data of the device to be hibernated, determine the hierarchical control strategy used to indicate whether the device to be hibernated enters the first hibernation state or the second hibernation state.

[0056] After identifying the device to be hibernated, it is also necessary to determine which hibernation state to adopt based on the current operating status of the device, such as whether it is carrying sessions and the overall traffic volume.

[0057] In one implementation, this application provides the following hierarchical hibernation strategy, specifically including: First sleep state (shallow sleep, Standby-Lite): This state is suitable for scenarios where the device to be sleeped is currently carrying a session or is expected to need to be woken up relatively quickly. In shallow sleep mode, hardware operations include: shutting down 50% of the CPU cores and putting the Dynamic Random Access Memory (DRAM) into partial self-refresh mode. Regarding session persistence: maintaining the N4 interface heartbeat with control plane devices (such as the Session Management Function (SMF), keeping the session state resident in memory, and ensuring a wake-up time of less than or equal to 500 milliseconds.

[0058] The second sleep state (deep sleep): This state is suitable for scenarios where the device to be hibernated has no session carrying capacity and is expected to not need to be woken up for a considerable period of time. In deep sleep mode, hardware operations include: shutting down the service plane power domain, retaining only the Baseboard Management Controller (BMC) management module, and ensuring that the overall power consumption is less than or equal to 5% of the rated value. Regarding state persistence: session data is encrypted using the advanced encryption standard AES-256 and stored in non-volatile memory or a solid-state drive (SSD), with a loading time of less than or equal to 8 seconds.

[0059] Step 14: Based on the control strategy obtained by executing step 13, generate control instructions and send the control instructions to the user plane device to control the user plane device to perform the corresponding hibernation operation.

[0060] After determining the control strategy, the equipment control system can generate corresponding control commands based on the strategy and send the control commands to the user plane equipment to be hibernated, so as to control it to perform the corresponding hibernation operation.

[0061] Before sending the control command to the user plane device, the device control system can migrate the session carried on the device to be hibernating to other user plane devices in the device pool according to the session hot migration technology.

[0062] To ensure uninterrupted service during device hibernation, in one implementation, the device control system may migrate the sessions carried on the device to be hibernated before sending the hibernation command. Figure 3 This is a flowchart of the UPF hibernation process provided in the embodiments of this application. Figure 4 A business migration flowchart provided for embodiments of this application. The following is in conjunction with… Figure 3 and Figure 4 The session migration and hibernation process is explained in detail.

[0063] like Figure 3 As shown, the hibernation process includes the following steps: The A1 and UPF clusters collect data through an energy efficiency monitoring agent, providing a real-time data foundation. The collected monitoring metrics include service metrics (throughput, number of sessions, QoS level, packet loss rate), energy consumption metrics (CPU / GPU dynamic power consumption, memory bandwidth utilization, device-level power consumption), and device status (chip temperature, fan speed, power mode).

[0064] A2-1. The energy efficiency monitoring agent reports (in real-time and continuously) to the energy efficiency analysis engine via Telemetry streaming. The energy efficiency analysis engine evaluates the efficiency based on the UPF energy efficiency ratio (η) model constructed by the multi-dimensional energy efficiency model. The energy efficiency ratio (η) model is shown in the following formula

[20] :

[20] Where λ is the session persistence weight coefficient, for example, the default setting is 0.3.

[0065] A2-2. The Energy Efficiency Monitoring Agent reports the traffic and session data of each UPF to the LSTM traffic prediction model module via Telemetry to provide data for subsequent traffic prediction.

[0066] A3. The Energy Efficiency Monitoring Agent reports the traffic and session data of each UPF to the LSTM traffic prediction model module via Telemetry.

[0067] The A4 LSTM traffic prediction model module generates traffic prediction data by overlaying historical data with special holiday activities, and then sends it to the decision-making module.

[0068] A5. The decision-making module determines the tiered hibernation strategy (shallow hibernation / deep hibernation / wake-up) of the UPF based on energy efficiency assessment results and traffic forecasts. The decision-making module sets the required number of UPFs to be used based on the traffic volume and sends the usage requirements to the energy efficiency-aware scheduler.

[0069] A6. The energy efficiency-aware scheduler generates traffic scheduling strategies (such as session migration rules) and sends them to the control plane SMF to balance load and energy efficiency. According to the dynamic weighted scheduling algorithm, new sessions are allocated according to the weight of the top node to avoid single-point overload; existing sessions are migrated according to a preset time period.

[0070] A7. The decision module sends the UPF sleep command in the UPF POOL to the UPF status controller, which, together with the SMF, puts the UPF into sleep mode.

[0071] In step A8, the UPF state controller notifies the SMF that the UPF has requested to hibernate. Subsequently, the SMF, based on existing specifications, determines whether to no longer select this UPF, using rules including S-NSSAI and DNN, geographical location, and the responsible TAI range. Simultaneously, the SMF receives information from message A6, learning the weights of each node in the UPF, and allocates subsequent session migration increments and the number of newly added or removed sessions according to these weights.

[0072] A9. SMF performs session command migration on UPF, allowing the dormant UPF to be idle.

[0073] A10, SMF notifies the UPF status controller to initiate UPF hierarchical hibernation.

[0074] A11, The UPF state controller controls the UPF to perform hierarchical hibernation.

[0075] In this embodiment of the application, the specific business migration process is as follows: Figure 4 As shown, it includes: 1. The UPF state controller sends PDU session migration information to the SMF, requesting the PDU session to be migrated.

[0076] 2-15. The process is consistent with the existing PDU session modification procedure. The SMF and PCF interoperate to execute the policy association process; the SMF updates user data with the UDM; the SMF decides on PDU session migration and associates it with the PCF policy; the SMF sends N1 and N2 interface session information modification commands to the AMF; the RAN and UE confirm the N / 2 PDU session with the AMF; the SMF sends an N4 session modification request message to the UPF to update the N4 session of the UPF involved in the PDU session modification; the UE confirms the PDU session modification command by sending a NAS message.

[0077] 16. The SMF sends an N4 session migration response to the UPF state controller, updating the N4 session of the UPF involved in the PDU session migration with the (N4 session ID) message.

[0078] During the aforementioned session migration process, GTP-U tunnel hot migration technology can be employed. Specifically, the source UPF (i.e., the device to be dormant) continuously forwards traffic until the target UPF (i.e., the device receiving the session) completes cache synchronization. Simultaneously, seamless bidirectional forwarding detection (Seamless BFD) can be used to detect path status, ensuring that the packet loss rate during migration is less than 0.001% and the migration latency is controlled within the millisecond range (less than 50 milliseconds).

[0079] After the session migration is completed, the UPF state controller sends a hierarchical hibernation control command to the UPF to be hibernating, and the UPF performs the corresponding shallow hibernation or deep hibernation operation according to the command.

[0080] It should be noted that, in this embodiment of the application, the device control system can determine the energy efficiency ratio of each user plane device in the device pool based on the energy efficiency index data corresponding to each user plane device, wherein the energy efficiency ratio is determined based on the ratio of the throughput of the user plane device to the power consumption of the user plane device; the scheduling weight of each user plane device is determined based on the energy efficiency ratio of each user plane device; and the existing sessions are migrated based on the scheduling weight.

[0081] To achieve energy-efficiency-aware load balancing, the equipment control system in this embodiment can employ a dynamic weighted scheduling algorithm. Specifically, this includes: Specifically, for each UPF node in the equipment pool, the equipment control system can calculate the scheduling weight according to its energy efficiency index using the following formula

[21] : [twenty one] Where, η i The energy efficiency index of the i-th node can be the node's computing power, resource utilization (such as CPU / memory efficiency), or overall performance score; α is an exponential parameter used to amplify the weight difference of high-energy-efficiency nodes. In this embodiment, α can be set to 1.5; Wi is the normalized node weight used for task allocation.

[0082] The weight of high-efficiency nodes can be amplified using formula

[20] . When α is greater than 1, the energy efficiency difference will be amplified exponentially. For example, if the energy efficiency η of node A is greater than 1, the weight of high-efficiency nodes can be amplified exponentially. A The energy efficiency η of node B is equal to 2. B The value is equal to 1 when α equals 1.5: WA=21.5 / (21.5+11.5)≈73.9% WB=11.5 / (21.5+11.5)≈26.1% Therefore, compared to WA being 66.7% when α equals 1, it is evident that α equals 1.5 significantly increases the weight ratio of high-energy-efficiency nodes. Normalization ensures that the total weight sums to 1, facilitating proportional task allocation.

[0083] Scheduling execution strategy: New session allocation: Tasks are assigned to preset energy-efficient nodes (e.g., Top 3 nodes) based on their weights. After calculating the weights of all nodes, the top 3 nodes with the highest weights are selected, and new tasks are allocated proportionally to these Top 3 nodes. For example, if the weights of the Top 3 nodes are 0.5, 0.3, and 0.2, the allocation probabilities are 50%, 30%, and 20%, respectively. This avoids concentrating all traffic on a single node while prioritizing the use of high-efficiency nodes to improve system efficiency.

[0084] Existing session migration: The system periodically (e.g., every minute) reassesses node energy efficiency and dynamically adjusts weights. Each migration involves 5% of the current existing sessions, gradually optimizing load distribution and avoiding disruptions caused by large-scale, sudden migrations. QoS-sensitive sessions are migrated last; that is, tasks insensitive to latency or interruption (such as batch computing) are migrated first, and QoS-sensitive sessions (such as real-time video streaming and online transactions) are processed last to ensure the service quality of critical businesses.

[0085] It should also be noted that when the user plane device enters the second sleep state according to the control command, it can also send a linkage command to the transmission network controller to control the transmission port corresponding to the user plane device to enter the low power mode.

[0086] In this embodiment, to achieve cross-domain collaborative energy saving, when the UPF enters deep sleep mode, this embodiment also establishes a cross-domain linkage mechanism with the transport layer network to dynamically decouple optical transmission energy consumption from service load. Specifically, when the user plane equipment is detected to have entered deep sleep mode, the UPF state controller can issue instructions to the southbound interface through the Network Configuration Protocol (NETCONF) / YANG model, etc., to automatically trigger the corresponding transmission port (such as the SPN port of the sliced ​​packet network) to enter low-power mode.

[0087] For example, by issuing commands via the NETCONF protocol, optical modules can switch from normal operating mode to low-power mode, with optical power adjustment accuracy reaching ±1dBm. The power consumption of a single optical module can be reduced from 10W to 1W, and the port activation latency is less than 10 milliseconds. In this way, when user plane equipment goes into sleep mode, the associated transmission equipment also saves energy, achieving joint energy saving between the core network and the transmission network.

[0088] It should also be noted that the method provided in this application embodiment further includes a wake-up process. When traffic is predicted to increase and more UPFs are needed to carry services, the decision module decides to wake up the dormant UPFs based on the energy efficiency assessment results and traffic prediction. Figure 5 A flowchart illustrating the UPF wake-up process provided in this application embodiment.

[0089] like Figure 5As shown, the UPF wake-up process is similar to the hibernation process, except that the decision module determines the number of UPFs that need to be woken up based on the predicted increase in traffic, and sends a wake-up request to the SMF through the UPF state controller. The SMF then re-includes the UPF in the selectable node pool according to existing specifications (such as S-NSSAI and DNN, geographical location, and the TAI range it is responsible for). The energy efficiency-aware scheduler assigns a higher initial weight to the woken UPF (since its traffic volume is 0) according to the dynamic weight scheduling algorithm, so that new traffic is preferentially allocated to the woken UPF. As the traffic volume increases, the energy efficiency-aware scheduler will recalculate and adjust the allocation weights accordingly.

[0090] The device control method provided in this application first obtains energy efficiency index data and service traffic data corresponding to each user plane device in the device pool. Then, based on the service traffic data and a pre-trained prediction model, the predicted traffic within a preset time window is determined. Based on the predicted traffic and the energy efficiency index data, a control strategy is determined for at least one user plane device in the device pool. According to the control strategy, a control command is generated and sent to the user plane device to control the user plane device to perform a corresponding hibernation operation. The load control method provided in this application has two advantages. First, by acquiring energy efficiency index data and service traffic data of each user plane device in the device pool, it achieves real-time dynamic perception of device operating status and network load. Furthermore, based on this, a pre-trained prediction model is used to predict service traffic within a preset time window, overcoming the shortcomings of existing technologies that rely solely on static configuration or current load for scheduling and cannot anticipate service changes. Second, the method provided in this application can determine a control strategy for at least one user plane device in the device pool to enter a sleep state based on predicted traffic and energy efficiency index data. This achieves dynamic matching between device operating status and service load. Compared to the shortcomings of existing technologies where all devices still need to remain operational during low-load periods, resulting in significant idling energy consumption, the device control method provided in this application can adaptively put some devices into a sleep state based on predicted service volume trends while ensuring service carrying capacity, effectively reducing overall energy consumption during low-load periods.

[0091] In one embodiment, this application also provides a device control apparatus to address the problem that under existing device scheduling mechanisms, the lack of dynamic matching between the operating status of user plane devices and service load leads to device idling and high energy consumption during low load periods. A schematic diagram of the specific structure of this device control apparatus is shown below. Figure 6 As shown, it includes: a data acquisition unit 61, a prediction unit 62, a strategy generation unit 63, and a control unit 64.

[0092] Among them, the data acquisition unit 61 is used to acquire energy efficiency index data and business traffic data corresponding to each user plane device in the device pool; The prediction unit 62 is used to determine the predicted traffic within a future preset time window based on the business traffic data and a pre-trained prediction model. The strategy generation unit 63 is used to determine a control strategy for at least one user plane device in the device pool based on the predicted traffic and the energy efficiency index data, wherein the control strategy is used to control the user plane device to enter a sleep state. The control unit 64 is configured to generate control commands according to the control strategy and send the control commands to the user plane device to control the user plane device to perform corresponding hibernation operations.

[0093] In one implementation, the redundant architecture includes at least two sets of second UPS systems capable of independently handling the entire load, each set of second UPS systems containing multiple UPS units.

[0094] In one embodiment, the system further includes a model training unit, specifically configured to: generate a training dataset based on the acquired historical traffic data, traffic flow characteristics, network status indicators, and external event data; and train an initial prediction model based on the training dataset to obtain the prediction model, wherein the traffic prediction model is used to output predicted traffic and its confidence interval.

[0095] In one implementation, the policy generation unit 63 is specifically configured to: determine, based on the predicted traffic, a first number of user plane devices required to run in the device pool within a future preset time window; determine devices to be hibernated based on a second number of user plane devices currently running in the device pool and the first number; and determine a hierarchical control policy for instructing the devices to enter a first hibernation state or a second hibernation state based on the running status data of the devices to be hibernated. The first hibernation state includes: shutting down some computing units of the devices to be hibernated and maintaining the session state corresponding to the devices to be hibernated in resident memory; the second hibernation state includes: shutting down the service plane power domain of the devices to be hibernated and persistently storing the session data corresponding to the devices to be hibernated.

[0096] In one embodiment, a data migration unit is further included, specifically used to: migrate the sessions carried on the device to be hibernated to other user plane devices in the device pool according to session hot migration technology; wherein, the session hot migration technology includes GTP-U tunnel hot migration, and during the session migration process, the device to be hibernated is controlled to continuously forward traffic until the target user plane device completes cache synchronization.

[0097] In one implementation, the data migration unit is specifically configured to: determine the energy efficiency ratio of each user plane device in the device pool based on the energy efficiency index data corresponding to each user plane device, wherein the energy efficiency ratio is determined based on the ratio of the throughput of the user plane device to the power consumption of the user plane device; determine the scheduling weight of each user plane device based on the energy efficiency ratio of each user plane device; and migrate existing sessions based on the scheduling weight.

[0098] In one embodiment, the control unit 64 is specifically configured to: when the user plane device enters a second sleep state according to the control command, send a linkage command to the transmission network controller to control the transmission port corresponding to the user plane device to enter a low-power mode.

[0099] Using the device control apparatus provided in this application embodiment, the energy efficiency index data and service traffic data corresponding to each user plane device in the device pool can be obtained first. Then, based on the service traffic data and a pre-trained prediction model, the predicted traffic within a preset time window is determined. Based on the predicted traffic and the energy efficiency index data, a control strategy is determined for at least one user plane device in the device pool. Based on the control strategy, a control command is generated and sent to the user plane device to control the user plane device to perform a corresponding sleep operation. The load control method provided in this application has two advantages. First, by acquiring energy efficiency index data and service traffic data of each user plane device in the device pool, it achieves real-time dynamic perception of device operating status and network load. Furthermore, based on this, a pre-trained prediction model is used to predict service traffic within a preset time window, overcoming the shortcomings of existing technologies that rely solely on static configuration or current load for scheduling and cannot anticipate service changes. Second, the method provided in this application can determine a control strategy for at least one user plane device in the device pool to enter a sleep state based on predicted traffic and energy efficiency index data. This achieves dynamic matching between device operating status and service load. Compared to the shortcomings of existing technologies where all devices still need to remain operational during low-load periods, resulting in significant idling energy consumption, the device control method provided in this application can adaptively put some devices into a sleep state based on predicted service volume trends while ensuring service carrying capacity, effectively reducing overall energy consumption during low-load periods.

[0100] Figure 7 This is a schematic diagram of the structure of an electronic device according to an embodiment of this application. Please refer to it. Figure 7At the hardware level, the electronic device includes a processor, and optionally also includes an internal bus, a network interface, and memory. The memory may include main memory, such as high-speed random-access memory (RAM), or non-volatile memory, such as at least one disk drive. Of course, the electronic device may also include other hardware required for other business operations.

[0101] The processor, network interface, and memory can be interconnected via an internal bus, which can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. This bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 7 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0102] Memory is used to store programs. Specifically, programs may include program code, which includes computer operation instructions. Memory may include main memory and non-volatile memory, and provides instructions and data to the processor.

[0103] The processor reads the corresponding computer program from non-volatile memory into main memory and then executes it, forming a device control unit at the logical level. The processor executes the program stored in memory and specifically performs the following operations: The system acquires energy efficiency index data and service traffic data corresponding to each user plane device in the device pool; based on the service traffic data and a pre-trained prediction model, it determines the predicted traffic within a preset time window; based on the predicted traffic and the energy efficiency index data, it determines a control strategy for at least one user plane device in the device pool, wherein the control strategy is used to control the user plane device to enter a sleep state; based on the control strategy, it generates a control command and sends the control command to the user plane device to control the user plane device to perform a corresponding sleep operation.

[0104] The above is as stated in this application. Figure 7The method for controlling electronic devices disclosed in the illustrated embodiments can be applied to or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method can be completed by integrated logic circuits in the processor's hardware or by instructions in software form. The processor can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in the embodiments of this application can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software module can reside in a mature storage medium in the field, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method.

[0105] Of course, in addition to software implementation, the electronic device of this application does not exclude other implementation methods, such as logic devices or a combination of hardware and software, etc. In other words, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0106] This application also proposes a computer-readable storage medium that stores one or more programs, the programs including instructions that, when executed by a portable electronic device including multiple applications, enable the portable electronic device to perform... Figure 2 The device control method of the illustrated embodiment is specifically used to perform the following operations: The system acquires energy efficiency index data and service traffic data corresponding to each user plane device in the device pool; based on the service traffic data and a pre-trained prediction model, it determines the predicted traffic within a preset time window; based on the predicted traffic and the energy efficiency index data, it determines a control strategy for at least one user plane device in the device pool, wherein the control strategy is used to control the user plane device to enter a sleep state; based on the control strategy, it generates a control command and sends the control command to the user plane device to control the user plane device to perform a corresponding sleep operation.

[0107] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0108] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0109] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0110] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1The steps of the function specified in one or more boxes.

[0111] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0112] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0113] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0114] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0115] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0116] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A device control method, characterized in that, include: Obtain energy efficiency index data and service traffic data for each user plane device in the device pool; Based on the business traffic data and a pre-trained prediction model, the predicted traffic within a future preset time window is determined. Based on the predicted traffic and the energy efficiency index data, a control strategy is determined for at least one user plane device in the device pool, wherein the control strategy is used to control the user plane device to enter a sleep state. According to the control strategy, a control command is generated and sent to the user plane device to control the user plane device to perform a corresponding hibernation operation.

2. The method according to claim 1, characterized in that, The prediction model is a traffic prediction model built on a Long Short-Term Memory (LSTM) network. The prediction model is then pre-trained, specifically including: A training dataset is generated based on the acquired historical business traffic data, business flow characteristics, network status indicators, and external event data. The initial prediction model is trained based on the training dataset to obtain the prediction model, wherein the traffic prediction model is used to output the predicted traffic and its confidence interval.

3. The method according to claim 1, characterized in that, The step of determining a control strategy for at least one user plane device within the device pool based on the predicted traffic flow and the energy efficiency index data specifically includes: Based on the predicted traffic, determine the first number of user plane devices that the device pool needs to run within a future preset time window; Based on the second number of user plane devices currently running in the device pool and the first number, determine the devices to be put into hibernation; Based on the operating status data of the device to be hibernated, a hierarchical control strategy is determined to instruct the device to enter a first hibernation state or a second hibernation state. The first hibernation state includes: shutting down some computing units of the device to be hibernated, and maintaining the session state corresponding to the device to be hibernated in memory; The second hibernation state includes: shutting down the service plane power domain of the device to be hibernated, and persistently storing the session data corresponding to the device to be hibernated.

4. The method according to claim 1, characterized in that, Before sending the control command to the user plane device, the method further includes: The sessions carried on the device to be hibernated are migrated to other user plane devices in the device pool using session hot migration technology. The session hot migration technology includes GTP-U tunnel hot migration, in which the device to be dormant is controlled to continuously forward traffic until the target user plane device completes cache synchronization.

5. The method according to claim 1, characterized in that, The method further includes: Based on the energy efficiency index data corresponding to each user plane device, the energy efficiency ratio of each user plane device in the device pool is determined respectively, wherein the energy efficiency ratio is determined based on the ratio of the throughput of the user plane device to the power consumption of the user plane device; The scheduling weight of each user plane device is determined based on its energy efficiency ratio. Existing sessions are migrated based on the aforementioned scheduling weights.

6. The method according to any one of claims 1 or 3, characterized in that, The method further includes: When the user plane device enters the second sleep state according to the control command, it sends a linkage command to the transmission network controller to control the transmission port corresponding to the user plane device to enter the low power mode.

7. A device control apparatus, characterized in that, include: The data acquisition unit is used to acquire energy efficiency index data and business traffic data corresponding to each user plane device in the device pool. The prediction unit is used to determine the predicted traffic within a preset time window based on the business traffic data and a pre-trained prediction model. The strategy generation unit is used to determine a control strategy for at least one user plane device in the device pool based on the predicted traffic and the energy efficiency index data, wherein the control strategy is used to control the user plane device to enter a sleep state. The control unit is configured to generate control commands according to the control strategy and send the control commands to the user plane device to control the user plane device to perform corresponding hibernation operations.

8. A device control device, comprising: processor; as well as A memory configured to store computer-executable instructions, which, when executed, cause the processor to perform the following operations: Obtain energy efficiency index data and service traffic data for each user plane device in the device pool; Based on the business traffic data and a pre-trained prediction model, the predicted traffic within a future preset time window is determined. Based on the predicted traffic and the energy efficiency index data, a control strategy is determined for at least one user plane device in the device pool, wherein the control strategy is used to control the user plane device to enter a sleep state. According to the control strategy, a control command is generated and sent to the user plane device to control the user plane device to perform a corresponding hibernation operation.

9. A computer-readable storage medium storing one or more programs, which, when executed by an electronic device including a plurality of applications, cause the electronic device to perform the device control method as described in any one of claims 1-6.

10. A computer program product, characterized in that, It includes a computer program that, when executed by a processor, implements the device control method as described in any one of claims 1-6.

Images