A wireless sensor node monitoring system, method, and medium based on rack management controller scheduling.
By employing a wireless sensor node monitoring system in the liquid-cooled cabinet, utilizing Bluetooth anchor nodes and short-range BLE communication, the risks of coolant leakage and signal attenuation in the liquid-cooled cabinet are solved, achieving low-power, highly reliable sensor data acquisition and rapid response.
Patent Information
- Application Number
- CN202611440605.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-16
- Publication Date
- 2026-11-17
AI Technical Summary
Liquid-cooled cabinets pose a high risk of coolant leakage, sensor cables are prone to corrosion, installation is difficult and there is a risk of wear, and wireless signals are severely attenuated in metal casings, making effective monitoring difficult.
A wireless sensor node monitoring system is adopted, which is scheduled by the cabinet management controller and uses a two-hop wake-up path to bypass the metal cabinet wall. Combined with Bluetooth anchor nodes and short-range BLE communication, it can achieve low power consumption and high reliability of sensor data acquisition and leakage detection.
It achieves low power consumption, anti-interference, real-time monitoring and rapid response of liquid-cooled cabinet, reduces sensor installation complexity and leakage risk, and improves monitoring reliability and efficiency.
Smart Images

Figure CN123113571A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of data center liquid cooling monitoring technology, and in particular to a wireless sensor node monitoring system, method and medium based on rack management controller scheduling. Background Technology
[0002] In high-density liquid-cooled rack systems for AI applications, it is necessary to monitor parameters such as coolant temperature, piping pressure, flow rate, and leakage in real time. Existing sensing and monitoring solutions mainly include the following: (1) Wired sensor network: Each sensor is connected to the BMC or CDU controller via a cable (I2C / SMBus / analog signal line). The sensor is continuously powered, continuously collects data, and continuously reports data. This is the mainstream solution for liquid-cooled cabinets.
[0003] (2) Wireless sensor + continuous power supply: The sensor node has a built-in wireless module (WiFi / Zigbee / BLE) and is powered by a battery or energy harvester. It periodically wakes up to collect data and report it. Usually, a fixed period is used (such as collecting data once every 1 minute), regardless of whether there is a person patrolling.
[0004] (3) Active polling by BMC: The BMC periodically polls the status of each sensor through a wired bus. The sensor can enter a low-power mode during non-polling periods. However, this method still requires a wired connection, and the BMC polling frequency is limited by the bus bandwidth.
[0005] Existing patents disclose a method and system for reducing power consumption in IoT wireless sensor systems. The method includes: a system initialization phase, a daily inspection phase, an intelligent decision-making phase, an instantaneous response phase, a main control processing phase, and a task completion phase. The coprocessor used employs subthreshold design technology, with a standby current of less than 100nA. It runs continuously during the wireless SoC's sleep period, autonomously managing sensor polling, threshold comparison, and wake-up decisions. The coprocessor only wakes up the wireless SoC via a built-in power switch to perform data processing and wireless transmission when sensor data exceeds a preset threshold, reaches a scheduled reporting time, or receives an external interrupt. This invention achieves an architectural shift from "timed polling" to "event response," reducing overall power consumption by up to 90% and extending battery life by 5-10 times. It is suitable for IoT applications such as smart agriculture, industrial monitoring, cold chain logistics, and smart buildings.
[0006] The existing technical solutions described above have the following drawbacks: 1. Liquid-cooled cabinets pose a risk of coolant leakage, and wired connectors are prone to corrosion in humid environments. Sealing of sensor cables through walls increases the complexity of the cabinet structure and adds to the risk of leakage. 2. The liquid-cooled cabinet has a dense network of pipes (supply pipes, return pipes, cold plate piping), making the space extremely limited. Sensor cables must run along the pipes, crossing cooling lines, which makes installation difficult and poses a risk of wear. Leaks can also cause short circuits in the cables. 3. The liquid-cooled cabinet has a metal casing, which provides strong shielding against 2.4GHz radio frequency signals (Faraday cage effect). If the wireless signal needs to penetrate the metal wall from the outside of the cabinet to reach the internal sensor, the signal attenuation can reach 40-60dB or more, making it essentially infeasible. Furthermore, the liquid cooling medium (PG25 with 75% water content) strongly absorbs 2.4GHz signals. Therefore, a purely external wireless wake-up solution is physically impractical. Summary of the Invention
[0007] To address the shortcomings of existing technologies, the purpose of this application is to provide a wireless sensor node monitoring system, method, and medium based on rack management controller scheduling. Through a two-hop wake-up path, the wake-up signal bypasses the physical barrier of the metal rack wall: the maintenance terminal communicates with the BMC via WiFi (without penetrating the rack wall), the BMC triggers the Bluetooth anchor node inside the rack via I2C (wired reliable), and the Bluetooth anchor node inside the rack wakes up the wireless sensor node via short-range BLE inside the rack (without penetrating the rack wall).
[0008] This was achieved using the following technical solutions: In a first aspect, this application provides a wireless sensor node monitoring system based on rack management controller scheduling, comprising: Wireless sensor nodes are integrated sensor modules deployed in key parts of liquid-cooled cabinets to acquire and report corresponding sensor data according to a preset multi-level power management state machine. The wireless sensing node includes a temperature sensor, a pressure sensor, a flow sensor, a leakage sensor, a Bluetooth chip, an antenna, and a power management circuit; the key components include the CDU inlet and outlet, the cold plate quick connector, the surface of the GPU heat sink, and the water distribution pipes; the sensing data of these components include pumping pressure, supply temperature, return temperature, return flow rate, coolant temperature, and leakage sensing resistance. The wireless sensor node is set with a periodic self-test timer. It automatically wakes up the corresponding sensor at a preset fixed time interval, collects and caches the sensor data of the part, and retransmits it during the next inspection. The Bluetooth anchor node inside the cabinet is a Bluetooth module deployed inside the liquid-cooled cabinet. It is used to receive Bluetooth wake-up signals via serial communication protocol, and at the same time receive part sensor data and transmit it to the cabinet management controller. The Bluetooth anchor node inside the cabinet wakes up the corresponding wireless sensor node through short-range BLE communication and collects sensor data from the corresponding part; and the Bluetooth anchor node inside the cabinet is equipped with a safety mechanism to prevent accidental wake-up. The cabinet management controller is used to receive and decompose cabinet inspection instructions, extract Bluetooth wake-up signals, and dispatch them to the Bluetooth anchor nodes in the cabinet according to the wake-up mechanism. At the same time, it receives and forwards the sensor data of the parts to the operation and maintenance terminal. The operation and maintenance terminal connects to the cabinet management controller, initiates inspection requests, converts them into cabinet inspection instructions, and receives location sensor data. The liquid leakage detection terminal connects to the wireless sensor node and is used to monitor the liquid cooling cabinet in real time through a general interface asynchronous interrupt mechanism, and generate a liquid leakage alarm notification. The multi-level power management state machine includes a deep sleep state, an authentication connection state, a data acquisition state, and a data reporting state; the leakage sensor monitors the liquid-cooled cabinet's cold plate quick connectors, water distribution pipes, and the bottom of the cabinet in real time during the deep sleep state through a general input / output port asynchronous interrupt mechanism.
[0009] By adopting the above technical solution, integrating multiple sensors and multi-level power state machines, and combining BLE wake-up and anti-false wake-up mechanisms, the Bluetooth anchor node in the Bluetooth cabinet wakes up the wireless sensor node as needed to collect data such as pump pressure, temperature, flow rate and leakage resistance. The data is then forwarded to the maintenance terminal by the cabinet management controller. Leakage detection is detected and alarmed in real time via asynchronous interruption. This achieves low-power, high-reliability, anti-interference wireless monitoring and rapid response for key parts of the liquid-cooled cabinet.
[0010] Furthermore, the false wake-up prevention security mechanism includes: Based on a pre-defined whitelist of Bluetooth anchor node addresses within the cabinet, the wireless sensor node performs address matching for Bluetooth connection requests. If the Bluetooth address of the current Bluetooth anchor node in the cabinet is within the Bluetooth anchor node address whitelist in the cabinet, the wireless sensor node extracts and verifies the wake-up authentication code in the Bluetooth connection request. If the wake-up authentication code is successfully verified, the wireless sensor node is woken up according to the minimum wake-up interval, and the data acquisition mode is triggered to collect data on the key parts of the liquid-cooled cabinet to obtain part sensor data. According to the preset node timer, the wireless sensor node performs periodic self-tests and caches the sensor self-test data.
[0011] By adopting the above technical solution, using whitelist address filtering and dynamic wake-up authentication code verification algorithm, the wireless sensor node is triggered to collect sensor data of the liquid-cooled cabinet only when the address of the Bluetooth anchor node in the cabinet matches the preset whitelist and the wake-up authentication code is verified. At the same time, the node caches self-test data through a periodic self-test timer, which effectively prevents illegal or false wake-ups, significantly reduces power consumption and improves monitoring security and reliability.
[0012] Secondly, this application also provides a wireless sensor node monitoring method based on cabinet management controller scheduling, which adopts the following technical solution; A method for monitoring wireless sensor nodes based on rack management controller scheduling includes: According to the preset multi-level power management state machine, acquire and report the corresponding component sensor data; The system receives Bluetooth wake-up signals via serial communication protocol, and simultaneously receives and transmits part sensor data to the cabinet management controller. Initiate inspection requests according to the preset inspection cycle, convert them into rack inspection instructions, and receive sensor data from the relevant parts. Receive and decompose cabinet inspection instructions, extract Bluetooth wake-up signals, and dispatch them to the Bluetooth anchor node in the cabinet according to the wake-up mechanism, while forwarding the sensor data of the parts to the operation and maintenance terminal. The system uses a general interface asynchronous interrupt mechanism to monitor key components of the liquid cooling cabinet in real time and generate leakage alarm notifications.
[0013] By adopting the above technical solution, through multi-level power state machine, BLE wake-up and serial communication protocol coordinated scheduling, the inspection command initiated by the operation and maintenance terminal is decomposed and scheduled to the Bluetooth anchor node in the Bluetooth cabinet by the cabinet management controller. The wireless sensor node is woken up according to the anti-false wake-up mechanism to collect sensor data of key parts of the liquid-cooled cabinet and transmit it back. At the same time, the general interface asynchronous interrupt is used to monitor the liquid leakage in real time and alarm, realizing low power consumption, high reliability and real-time response of liquid-cooled cabinet wireless monitoring and fault early warning.
[0014] Furthermore, the wake-up mechanism includes: According to the preset inspection cycle, the maintenance terminal sends an inspection request through the universal standard interface of the cabinet management controller via wireless / wired network connection; The inspection request is parsed by the cabinet management controller to extract the cabinet number to be inspected and the Bluetooth anchor node in the target area of the cabinet. The cabinet inspection command is triggered based on the inspection request and transmitted to the Bluetooth anchor node in the cabinet through the cabinet management controller. The cabinet inspection command is parsed through the Bluetooth anchor node inside the cabinet, and the Bluetooth wake-up signal is extracted. Match the corresponding wireless sensor node based on the Bluetooth wake-up signal, and send a Bluetooth connection request and wake-up authentication code; The wireless sensor node is triggered to acquire data based on the Bluetooth connection request. After the wake-up authentication code is verified, the sensor data is acquired and sent sequentially to the Bluetooth anchor node in the cabinet, the cabinet management controller, and the maintenance terminal.
[0015] By adopting the above technical solution, based on the inspection task parsing and cabinet Bluetooth anchor node mapping algorithm, the request is converted into a Bluetooth wake-up signal. The wireless sensor node state machine is triggered by whitelist address matching and dynamic wake-up authentication code verification. The data of the location is collected and uploaded step by step along the BLE-serial link, realizing on-demand accurate wake-up and secure communication, significantly reducing power consumption, avoiding false wake-ups and improving inspection efficiency and reliability.
[0016] Furthermore, the general-purpose interface asynchronous interrupt mechanism includes: The resistance value of the leakage sensor is detected and judged. If the resistance value is greater than the preset first resistance threshold, the environment of the liquid cooling cabinet is determined to be normal dry, triggering a high level on the general interface. If the resistance value is less than the preset second resistance threshold, the environmental condition of the liquid-cooled cabinet is determined to be partial leakage, triggering a low level on the general interface; Based on the low level of the general interface, the wireless sensor node establishes a Bluetooth connection with the Bluetooth anchor node inside the cabinet and sends a platform-level leakage alarm message. The platform-level leakage alarm message is forwarded to the cabinet management controller via the Bluetooth anchor node inside the cabinet; Based on the platform-level leakage alarm message, the rack management controller triggered an emergency procedure.
[0017] By adopting the above technical solution, based on resistance threshold comparison and asynchronous interrupt algorithm, the resistance value of the leakage sensor is detected in real time. When the resistance value is lower than the second threshold, a low-level interrupt is triggered on the general interface. The wireless sensing node immediately sends a platform-level leakage alarm to the Bluetooth anchor node in the Bluetooth cabinet through BLE, and triggers an emergency process through the cabinet management controller. This achieves millisecond-level proactive reporting and rapid linkage handling of leakage events, significantly improving the operational safety and fault response efficiency of the liquid-cooled cabinet.
[0018] Thirdly, this application also provides a storage medium storing at least one instruction, at least one program, code set, or instruction set, wherein the at least one instruction, at least one program, code set, or instruction set is loaded and executed by a processor to implement the wireless sensor node monitoring method based on rack management controller scheduling as described above.
[0019] In summary, the beneficial technical effects of this application are as follows: The wake-up signal is transmitted through a two-hop path: "WiFi → BMC → I2C → in-cabinet Bluetooth anchor node → in-cabinet BLE → wireless sensor node". All BLE communication is completed within a short distance of 30-80cm in the cabinet, without penetrating the metal cabinet wall, thus avoiding the physical obstacles of the Faraday cage effect and the absorption of liquid cooling fluid. Employing BLE connection wake-up (binary criterion: connection request received / not received), it does not rely on the absolute value of RSSI to determine distance, resulting in significantly higher reliability in multi-path metal environments compared to RSSI threshold solutions. The wireless sensor nodes are cableless, directly pasted / snap-on mounted to the surface of liquid-cooled piping. With no sealed wall penetrations, leakage risks are eliminated, and data is only collected when maintenance personnel initiate inspections. Attached Figure Description
[0020] Figure 1 This is a block diagram showing the relationship between the modules of the wireless sensor node monitoring system in this application; Figure 2 This is a flowchart illustrating the security mechanism for preventing accidental wake-up in this application; Figure 3 This is a flowchart illustrating the wireless sensor node monitoring method in this application; Figure 4 This is a flowchart illustrating the wake-up mechanism in this application. Detailed Implementation
[0021] The present application will be further described in detail below with reference to the accompanying drawings.
[0022] Reference Figure 1 The present application discloses a wireless sensor node monitoring system based on rack management controller scheduling, comprising: Wireless Sensor Node: An integrated sensing module deployed in key locations within the liquid cooling cabinet (CDU inlet / outlet, quick-connect fittings for cold plates, GPU heatsink surface, water distribution piping). Each node includes a BLE SoC chip, sensors (temperature / pressure / flow / leakage), an antenna, and power management circuitry. The node is normally in deep sleep mode, but is awakened by a Bluetooth anchor node within the cabinet via BLE to collect and report data.
[0023] Cabinet BLE Anchor: A Bluetooth module deployed inside the cabinet frame (it can reuse the slot-based Bluetooth anchor node from Document 1, or be deployed independently as an area-specific Bluetooth anchor node), connected to the BMC via I2C / SPI. Under the BMC's scheduling, the cabinet-based Bluetooth anchor node performs two tasks: ① Sending BLE wake-up commands to the wireless sensor nodes in the designated area; ② Receiving data reported by the wireless sensor nodes and forwarding it to the BMC. The cabinet-based Bluetooth anchor node and the wireless sensor nodes are located within the same cabinet, with a communication distance of 30-80cm, and do not penetrate the cabinet walls.
[0024] Cabinet Management Controller (BMC / RMC): As the system scheduling hub, it receives inspection instructions from maintenance personnel (via WiFi / Redfish API), breaks down the instructions into wake-up tasks for Bluetooth anchor nodes in each cabinet area (via I2C), collects sensor data forwarded by Bluetooth anchor nodes in the cabinet, and returns it to the maintenance terminal via Redfish API.
[0025] Inspection Terminal: This is the inspection terminal (PC / tablet / phone) used by operations and maintenance personnel. It connects to the BMC's Redfish API via WiFi to initiate inspection requests and receive sensor data. The terminal itself does not require BLE functionality: it communicates with the BMC via WiFi and does not need to communicate directly with the wireless sensor nodes.
[0026] Leakage detection end: The leakage sensor is directly connected to the wireless sensing node MCU via GPIO hardware interrupt. It monitors in real time in deep sleep mode and immediately wakes up the node to send an alarm when leakage is detected, without being restricted by the on-demand wake-up process. Upon receiving a BLE connection request while in deep sleep, the system enters the authentication connection state. After successful authentication, it enters the data acquisition state. Once data acquisition is complete, it enters the data reporting state. After reporting is complete, it returns to deep sleep.
[0027] Reference Figure 2 Preferred, accidental wake-up prevention security mechanisms include: Based on a pre-defined whitelist of Bluetooth anchor node addresses within the cabinet, the wireless sensor node performs address matching for Bluetooth connection requests. If the Bluetooth address of the current Bluetooth anchor node in the cabinet is within the Bluetooth anchor node address whitelist in the cabinet, the wireless sensor node extracts and verifies the wake-up authentication code in the Bluetooth connection request. If the wake-up authentication code is successfully verified, the wireless sensor node is woken up according to the minimum wake-up interval, and the data acquisition mode is triggered to collect data on the key parts of the liquid-cooled cabinet to obtain part sensor data. According to the preset node timer, the wireless sensor node performs periodic self-tests and caches the sensor self-test data.
[0028] In this embodiment, the following table shows the details.
[0029] Security Mechanism illustrate BLE address matching The wireless sensor node only responds to the pre-configured BLE addresses of the cabinet's Bluetooth anchor nodes (whitelist). Connection requests from non-whitelisted devices are directly rejected. The whitelist is distributed by the BMC to the cabinet's Bluetooth anchor nodes via I2C, and the cabinet's Bluetooth anchor nodes include a wake-up authentication code in their connection requests. Wake-up authentication code verification When the Bluetooth anchor node inside the cabinet initiates a BLE connection, it carries a 4-byte wake-up authentication code (updated and synchronized to the Bluetooth anchor node inside the cabinet by the BMC every 24 hours). The wireless sensor node only enters the data acquisition mode after successful verification. This prevents unauthorized BLE devices from imitating the Bluetooth anchor node inside the cabinet to trigger wake-up. Wake-up rate limit The node limits the minimum wake-up interval to 30 seconds (configurable) to prevent frequent wake-ups and power consumption caused by being "bombarded" by high-frequency connection requests. Periodic self-check Every 6 hours (configurable), the node timer automatically wakes up to perform a sensor self-check and data acquisition, caching the data. This data is then reported during the next inspection. This ensures that even if there is no inspection for an extended period, basic data records are still available. Reference Figure 3 The present application discloses a wireless sensor node monitoring method based on rack management controller scheduling, comprising: S1: According to the preset multi-level power management state machine, acquire and report the corresponding component sensor data; S2: Receives Bluetooth wake-up signal via serial communication protocol, and simultaneously receives part sensor data and transmits it to the cabinet management controller; S3: Initiates an inspection request according to the preset inspection cycle, converts it into a cabinet inspection command, and receives part sensor data; S4: Receive and decompose cabinet inspection instructions, extract Bluetooth wake-up signals, and dispatch them to the Bluetooth anchor node in the cabinet according to the wake-up mechanism, while forwarding the sensor data of the parts to the operation and maintenance terminal. S5: Real-time monitoring of key components of the liquid-cooled cabinet via a general interface asynchronous interrupt mechanism, generating leakage alarm notifications.
[0030] In this embodiment, the maintenance personnel open the Redfish management interface of BMC on the maintenance terminal (PC / tablet), select the target cabinet and inspection area (such as "RACK-001, CDU + cold plate area"), and click "Start Inspection".
[0031] After receiving the Redfish request, the BMC parses the EID list of the Bluetooth anchor nodes in the cabinet corresponding to the target area, and sends a wake-up command (including the target wireless sensor node BLE address list and wake-up authentication code) to these Bluetooth anchor nodes in the cabinet via I2C. The I2C transmission latency is <1ms.
[0032] After receiving the I2C command, the Bluetooth anchor node inside the cabinet sequentially initiates a BLE connection request (carrying a wake-up authentication code) to each target wireless sensor node. The distance between the Bluetooth anchor node and the wireless sensor node is 30-80cm, and the BLE connection establishment success rate is >99%.
[0033] The BLE SoC of the wireless sensor node scans for a connection request from the Bluetooth anchor node inside the cabinet, triggering a hardware interrupt to wake up the MCU. After the MCU verifies the wake-up authentication code, it establishes a BLE connection and enters data acquisition mode.
[0034] The MCU connects the sensor power supply via a load switch. Temperature (TMP117, 0~15ms), pressure (BMP390, 0~5ms), flow rate (GPIO count, 0~10ms), and leakage status (GPIO read, <1ms) are read sequentially. The total acquisition time is approximately 50~100ms.
[0035] The MCU encapsulates the sensor readings into a PLDM GetSensorReading message and sends it to the Bluetooth anchor node inside the cabinet via a BLE connection. Upon receiving the message, the Bluetooth anchor node replies with an ACK and forwards it to the BMC via I2C using an MCTP message. If no ACK is received after three retransmissions, the node buffers the data. The PLDM GetSensorReading message includes a PLDM header (3 bytes), sensor ID (2 bytes), sensor reading (4 bytes), sensor status (1 byte), event status (1 byte), timestamp (8 bytes), battery level (1 byte), and node ID (2 bytes).
[0036] After collecting data from all target wireless sensor nodes, the BMC returns the summarized results to the operations and maintenance terminal via the Redfish API. The terminal displays the values of each sensor, highlighting any outliers.
[0037] After data reporting is complete, the BLE connection is disconnected, and the wireless sensor node automatically returns to deep sleep. If there is no new wake-up request within 30 seconds, the node will be forced to return to sleep.
[0038] At any time, if the leak sensor detects liquid continuity, a GPIO interrupt immediately wakes up the wireless sensor node. The node actively broadcasts an alarm packet via BLE, and the Bluetooth anchor node inside the cabinet scans and receives the packet, reporting it to the BMC via I2C. The BMC then triggers an emergency alarm. This process is not limited by the on-demand wake-up mechanism and has a delay of less than 1 second.
[0039] Every 6 hours, the node timer automatically wakes up to perform sensor data acquisition and cache. The data is then reported during the next inspection, ensuring basic data records are maintained even in scenarios with long-term unattended inspections.
[0040] Reference Figure 4 Preferred wake-up mechanisms include: According to the preset inspection cycle, the maintenance terminal sends an inspection request through the universal standard interface of the cabinet management controller via wireless / wired network connection; The inspection request is parsed by the cabinet management controller to extract the cabinet number to be inspected and the Bluetooth anchor node in the target area of the cabinet. The cabinet inspection command is triggered based on the inspection request and transmitted to the Bluetooth anchor node in the cabinet through the cabinet management controller. The cabinet inspection command is parsed through the Bluetooth anchor node inside the cabinet, and the Bluetooth wake-up signal is extracted. Match the corresponding wireless sensor node based on the Bluetooth wake-up signal, and send a Bluetooth connection request and wake-up authentication code; The wireless sensor node is triggered to acquire data based on the Bluetooth connection request. After the wake-up authentication code is verified, the sensor data is acquired and sent sequentially to the Bluetooth anchor node in the cabinet, the cabinet management controller, and the maintenance terminal.
[0041] In this embodiment, the BMC sends a wake-up command to the Bluetooth anchor node in the cabinet via I2C. The command contains the BLE address (6 bytes) and wake-up authentication code (4 bytes) of the target wireless sensor node. The wake-up authentication code is updated by BMC every 24 hours and synchronized to the Bluetooth anchor node in the cabinet via I2C.
[0042] The Bluetooth anchor node inside the cabinet initiates a BLE connection request (connectableadvertising→connection) to the target wireless sensor node.
[0043] The BLE SoC of the wireless sensor node maintains broadcast scanning in deep sleep mode (scan interval=2s, scan window=20ms, power consumption 3-5uA). When a connection request from a Bluetooth anchor node inside the cabinet is detected, the BLE SoC hardware automatically generates an interrupt to wake up the MCU.
[0044] After the MCU wakes up, it verifies the wake-up authentication code in the connection request. Once authentication is successful, a BLE connection is established, and the system enters data acquisition mode.
[0045] After data collection is completed, the data is sent to the Bluetooth anchor node inside the cabinet via BLE connection, and the Bluetooth anchor node inside the cabinet forwards the data to the BMC via I2C.
[0046] After data transmission is complete, the BLE connection is disconnected, and the wireless sensor node returns to deep sleep.
[0047] First hop: Operation and maintenance terminal → BMC (WiFi / wired network); The maintenance terminal connects to the BMC's Redfish API via WiFi (data center-level network, signal does not penetrate cabinet walls) or wired Ethernet to send an inspection request. The request specifies the cabinet ID and area to be inspected (e.g., "RACK-001, CDU area"). After receiving the request, the BMC parses the list of Bluetooth anchor node EIDs in the cabinet corresponding to the target area.
[0048] This jump uses standard WiFi / Ethernet + Redfish API, does not involve BLE, and is not affected by metal rack shielding. The WiFi signal propagates within the server room space without penetrating the rack walls.
[0049] Second jump: BMC → In-cabinet Bluetooth anchor node → Wireless sensor node (I2C + In-cabinet BLE); The BMC sends a wake-up command (MCTP message format) to the Bluetooth anchor node inside the cabinet in the target area via the I2C bus. The command includes the BLE address list and acquisition parameters of the target wireless sensor node. After receiving the command, the Bluetooth anchor node inside the cabinet sends a BLE connection request + wake-up data packet to each target wireless sensor node. Since the Bluetooth anchor node and the wireless sensor node are both inside the cabinet and only 30-80cm apart, BLE communication does not need to penetrate the cabinet wall, resulting in low signal path loss (30-40dB) and reliable reception.
[0050] After the wireless sensor node is woken up, it collects sensor data and sends it back to the Bluetooth anchor node in the cabinet via BLE. The Bluetooth anchor node in the cabinet forwards the data to the BMC via I2C, and the BMC then returns it to the maintenance terminal via the Redfish API.
[0051] Preferred, general-purpose asynchronous interrupt mechanisms include: The resistance value of the leakage sensor is detected and judged. If the resistance value is greater than the preset first resistance threshold, the environment of the liquid cooling cabinet is determined to be normal dry, triggering a high level on the general interface. If the resistance value is less than the preset second resistance threshold, the environmental condition of the liquid-cooled cabinet is determined to be partial leakage, triggering a low level on the general interface; Based on the low level of the general interface, the wireless sensor node establishes a Bluetooth connection with the Bluetooth anchor node inside the cabinet and sends a platform-level leakage alarm message. The platform-level leakage alarm message is forwarded to the cabinet management controller via the Bluetooth anchor node inside the cabinet; Based on the platform-level leakage alarm message, the rack management controller triggered an emergency procedure.
[0052] In this embodiment, the conductive rope-type leakage detection cable is laid along the bottom of the liquid cooling pipeline. During normal drying, the resistance is >10MΩ (GPIO high level), and when in contact with liquid, the resistance is <100kΩ (GPIO low level).
[0053] • The GPIO of the wireless sensor node MCU is configured for falling edge interrupt, and the interrupt remains valid in System OFF (deep sleep) mode (hardware-level interrupt, independent of MCU operation).
[0054] • After the GPIO falling edge is triggered, the MCU hardware automatically wakes up (<1ms) and immediately establishes a BLE connection to send a PLDM leakage alarm message (event code 0x03) to the Bluetooth anchor node in the cabinet.
[0055] • The Bluetooth anchor node inside the cabinet continuously maintains BLE scanning (configured as always-on scanning by the BMC or woken up by a leak alarm broadcast), and immediately reports to the BMC via I2C interrupt upon receiving an alarm.
[0056] Upon receiving a leak alarm, BMC triggers an emergency process (Redfish alarm, email notification, audible and visual alarm) with a delay of less than 1 second.
[0057] Example 1: Standard liquid-cooled AI cabinet (5 wireless sensor nodes + 3 in-cabinet Bluetooth anchor nodes) Taking a liquid-cooled AI server rack as an example, the rack contains 4 GPU computing nodes (each with 1 liquid-cooled heatsink) and 1 CDU cooling distribution unit. The system deployment parameters are as follows: Nodes / Parameters Location Sensor configuration powered by illustrate Node N1 CDU outlet Temperature (TMP117) + Pressure (BMP390) CR2032 Monitor supply temperature and pressure Node N2 CDU return water outlet Temperature (TMP117) + Flow Rate (Hall) CR2032 Monitor return water temperature and flow rate Node N3 GPU Cold Plate Quick Connector Temperature (TMP117) + Leakage (conductive rope) CR2032 Monitor cold plate temperature and joint leakage. Node N4 Water distributor inlet Temperature (TMP117) + Pressure (BMP390) CR2032 Monitor the inlet condition of the water distributor Node N5 Drainage tray at the bottom of the cabinet Leakage (conductive rope) + Temperature (TMP117) CR2032 Monitoring cabinet overall leakage Bluetooth anchor node 1 inside the cabinet Next to CDU (within the rack frame) BLE covers N1, N2 Wired 3.3V Wake-up + Relay Cabinet Bluetooth Anchor Node 2 GPU area (within the rack frame) BLE covers N3, N4 Wired 3.3V Wake-up + Relay Cabinet Bluetooth Anchor Node 3 Bottom of the rack (inside the rack frame) BLE covers N5 Wired 3.3V Wake-up + Relay BMC Top of rack I2C connection to Bluetooth anchor nodes 1-3 inside the cabinet Wired power supply Scheduling + Data Aggregation wake-up threshold BLE connection wake-up (non-RSSI) Wake-up authentication code (4 bytes) - Low error wake-up Self-inspection cycle Every 6 hours Timed wake-up for data collection + caching - Long-term lack of inspection and support Leakage alarm GPIO interrupt, real-time Unrestricted by wake-up - Delay < 1 second The expected performance metrics for this embodiment are: Node standby power consumption: 3~5uA / node, total standby power consumption of 5 nodes <25uA; Time taken for a single inspection: from click to data display, <1 second (5 nodes serial wake-up + data collection + reporting); Battery life: Approximately 1.5 to 2 years for CR2032 (10 wake-ups per day); Leakage alarm delay: <1 second from the occurrence of a leak to the BMC alarm; Data reporting success rate: >99% (3 retransmissions + offline cache retransmission).
[0058] In large liquid-cooled server rooms, a single rack may deploy more than 20 wireless sensor nodes. Multiple regional rack-mounted Bluetooth anchor nodes are deployed (one anchor node for every 3-5 wireless sensor nodes), and these anchor nodes are connected to the BMC via an I2C bus. Up to 127 devices (7-bit address space) can be connected to the I2C bus, meeting the needs of large-scale deployments. If the I2C bus length exceeds 1 meter, an I2C bus extender (such as the PCA9515) can be used, or multiple sub-buses can be created.
[0059] The hardware design of the wireless sensor node is shown in the table below.
[0060] Parameters Design Specifications BLE SoC Nordic nRF52832 or TI CC2640R2F (Bluetooth 5.0, price approximately 3-5 yuan) illustrate Bluetooth 6.0 CS functionality is not required; Bluetooth 5.0 is sufficient for short-range communication within the cabinet, reducing costs. Temperature sensor TI TMP117 (I2C interface, accuracy ±0.1°C, -55~+150°C), or NTC thermistor acquired via ADC. pressure sensor Bosch BMP390 or MEAS MS5837 (I2C interface, range 0~10 bar, accuracy ±0.2 bar) Flow sensor Hall effect flow sensor (G1 / 2 water flow meter), pulse output connected to GPIO Leakage sensor Leakage detection cable (conductive cord), connected to GPIO interrupt input antenna PCB inverted-F antenna, 0dBi gain, 2.4GHz omnidirectional Power Management TPS62740 DC-DC step-down converter (90% efficiency) + TPS22916 load switch Battery CR2032 button cell battery (220mAh) or small LiPo battery (400mAh) Power consumption metrics Deep sleep: 3~5uA; Wake-up acquisition: 0.5~2mA (50-100ms); BLE transmission: 4~8mA (20-50ms) Protection level IP67 sealed housing (PC+ABS injection molding), PCB conformal coating, resistant to coolant immersion. Operating temperature -10°C ~ +85°C Physical dimensions 30mm in diameter x 12mm in thickness (round button type), can be glued or snap-on installed. The design of the Bluetooth anchor nodes inside the cabinet is shown in the table below.
[0061] Parameters Design Specifications BLE SoC Nordic nRF52832 (reduces BOM complexity) Function ① Receive the I2C wake-up command from the BMC; ② Initiate a BLE connection and wake-up to the wireless sensor node; ③ Receive data from the wireless sensor node and forward it to the BMC via I2C. antenna The PCB inverted F antenna is installed inside the cabinet frame, facing the wireless sensor node area. powered by The rack back panel is powered by a 3.3V wired connection (the Bluetooth anchor node inside the rack frame is embedded, so no battery is required). I2C interface Connect to the BMC, address 0x10~0x3F (address space or partition). BLE coverage Wireless sensor nodes (usually 3-5) covering a 30-80cm range inside the cabinet, covering the area where the Bluetooth anchor node is located. Deployment location One each next to the CDU, next to the water distributor, and in each row of GPU cold plate areas. The power consumption and duration budget for each state are shown in the table below.
[0062] state Current (uA) Typical duration Activity Details Deep sleep 3~5 23h55m (99.7%) BLE scan monitoring + leakage GPIO interrupt + RTC Authentication connection 5~15 100~200ms BLE connection establishment + Wake-up authentication code verification Data collection 500~2000 50~100ms Sensor reading + MCU processing Data reporting 4000~8000 20~50ms BLE data transmission + ACK waiting Single wake-up cycle - ~200~350ms State 2+3+4 total 10 wake-ups per day Average 0.22 24h 0.12mAh (sleep) + 0.10mAh (wake-up) Based on the table above, the CR2032 battery (220mAh), operating in a mode where it is woken up 10 times a day for inspection, has an average daily power consumption of approximately 0.22mAh, theoretically providing a battery life of 1000 days (approximately 2.7 years). Considering battery self-discharge and temperature effects, a conservative estimate is 1.5 to 2 years. If a 400mAh LiPo battery is used, the battery life can reach over 3 years.
[0063] The data reported by the wireless sensor nodes is encapsulated in the PLDM for Platform Monitoring and Control (DMTF DSP0248) standard format and transmitted to the Bluetooth anchor node in the cabinet via MCTP over BLE. The Bluetooth anchor node in the cabinet then forwards the data to the BMC via MCTP over I2C, as shown in the table below.
[0064] Fields length illustrate PLDM Header 3 bytes Version (1B) + Message Type (1B, 0x01=PLDM) + Command Code (1B, 0x11=GetSensorReading) Sensor ID 2 bytes Sensor number Sensor Reading 4 bytes Sensor readings (floating-point or fixed-point) Sensor State 1 byte bit0 = Normal, bit1 = Alarm, bit2 = Critical, bit3 = Invalid Event State 1 byte 0x00 = No event, 0x01 = Temperature exceeded limit, 0x02 = Pressure abnormal, 0x03 = Leakage alarm Timestamp 8 bytes UTC timestamp (generated by node RTC) Battery Level 1 byte Battery percentage (0~100%) Node ID 2 bytes Node unique identifier (1) Multiple sensor reuse: A wireless sensor node can integrate multiple sensors (temperature + pressure + flow) and collect all sensor data in one wake-up cycle, which is then encapsulated into multiple PLDM GetSensorReading messages. Each message is about 20 bytes, and the data from 5 sensors (100 bytes) can be transmitted in one BLE connection.
[0065] (2) Leakage alarm priority: When the leakage sensor detects liquid, it generates a GPIO hardware interrupt, directly waking up the node (skipping the BLE wake-up process of the Bluetooth anchor node in the cabinet), and sending the PLDM alarm message with the highest priority. This interrupt remains effective in deep sleep mode, ensuring that the leakage alarm is not limited by the on-demand wake-up mechanism.
[0066] This application discloses a storage medium storing at least one instruction, at least one program, code set, or instruction set. The at least one instruction, at least one program, code set, or instruction set is loaded and executed by a processor to implement the wireless sensor node monitoring method based on rack management controller scheduling as described above.
[0067] The embodiments described in this specific implementation are preferred embodiments of this application and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.
Claims
1. A wireless sensor node monitoring system based on cabinet management controller scheduling, characterized by, include: Wireless sensor nodes are integrated sensor modules deployed in key parts of liquid-cooled cabinets to acquire and report corresponding part sensor data according to a preset multi-level power management state machine. The Bluetooth anchor node inside the cabinet is a Bluetooth module deployed inside the liquid-cooled cabinet. It is used to receive Bluetooth wake-up signals via serial communication protocol, and at the same time receive part sensor data and transmit it to the cabinet management controller. The cabinet management controller is used to receive and decompose cabinet inspection instructions, extract Bluetooth wake-up signals, and dispatch them to the Bluetooth anchor nodes in the cabinet according to the wake-up mechanism. At the same time, it receives and forwards the sensor data of the parts to the operation and maintenance terminal. The maintenance terminal connects to the cabinet management controller, initiates inspection requests, converts them into cabinet inspection instructions, and receives location sensor data. The leakage detection unit connects to the wireless sensor node and is used to monitor the liquid-cooled cabinet in real time through a general interface asynchronous interrupt mechanism, generating leakage alarm notifications.
2. The wireless sensor node monitoring system based on cabinet management controller scheduling of claim 1, wherein, The wireless sensing node includes a temperature sensor, a pressure sensor, a flow sensor, a leakage sensor, a Bluetooth chip, an antenna, and a power management circuit; the key components include the CDU inlet and outlet, the cold plate quick connector, the surface of the GPU heat sink, and the water distribution pipes; the sensing data of these components include pumping pressure, supply temperature, return temperature, return flow rate, coolant temperature, and leakage sensing resistance.
3. The cabinet management controller schedule based wireless sensor node monitoring system of claim 1, wherein, The wireless sensor node is equipped with a periodic self-test timer, which automatically wakes up the corresponding sensor at preset fixed intervals, collects and caches the sensor data of the part, and retransmits it during the next inspection.
4. The cabinet management controller schedule based wireless sensor node monitoring system of claim 1, wherein, The Bluetooth anchor node inside the cabinet wakes up the corresponding wireless sensor node through short-range BLE communication and collects sensor data from the corresponding location; and the Bluetooth anchor node inside the cabinet is equipped with a safety mechanism to prevent accidental wake-up.
5. The cabinet management controller schedule based wireless sensor node monitoring system of claim 2, wherein, The multi-level power management state machine includes a deep sleep state, an authentication connection state, a data acquisition state, and a data reporting state; the leakage sensor monitors the liquid-cooled cabinet's cold plate quick connectors, water distribution pipes, and the bottom of the cabinet in real time during the deep sleep state through a general input / output port asynchronous interrupt mechanism.
6. The wireless sensor node monitoring system based on rack management controller scheduling according to claim 4, characterized in that, The aforementioned security mechanism to prevent accidental wake-up includes: Based on a pre-defined whitelist of Bluetooth anchor node addresses within the cabinet, the wireless sensor node performs address matching for Bluetooth connection requests. If the Bluetooth address of the current Bluetooth anchor node in the cabinet is within the Bluetooth anchor node address whitelist in the cabinet, the wireless sensor node extracts and verifies the wake-up authentication code in the Bluetooth connection request. If the wake-up authentication code is successfully verified, the wireless sensor node is woken up according to the minimum wake-up interval, and the data acquisition mode is triggered to collect data on the key parts of the liquid-cooled cabinet to obtain part sensor data. According to the preset node timer, the wireless sensor node performs periodic self-tests and caches the sensor self-test data.
7. A method for monitoring wireless sensor nodes based on rack management controller scheduling, applied to the wireless sensor node monitoring system according to any one of claims 1-6, characterized in that, include: According to the preset multi-level power management state machine, acquire and report the corresponding component sensor data; The system receives Bluetooth wake-up signals via serial communication protocol, and simultaneously receives and transmits part sensor data to the cabinet management controller. Initiate inspection requests according to the preset inspection cycle, convert them into rack inspection instructions, and receive sensor data from the relevant parts. Receive and decompose cabinet inspection instructions, extract Bluetooth wake-up signals, and dispatch them to the Bluetooth anchor node in the cabinet according to the wake-up mechanism, while forwarding the sensor data of the parts to the operation and maintenance terminal. The system uses a general interface asynchronous interrupt mechanism to monitor key components of the liquid cooling cabinet in real time and generate leakage alarm notifications.
8. The wireless sensor node monitoring method based on rack management controller scheduling according to claim 7, characterized in that, The wake-up mechanism includes: According to the preset inspection cycle, the maintenance terminal sends an inspection request through the universal standard interface of the cabinet management controller via wireless / wired network connection; The inspection request is parsed by the cabinet management controller to extract the cabinet number to be inspected and the Bluetooth anchor node in the target area of the cabinet. The cabinet inspection command is triggered based on the inspection request and transmitted to the Bluetooth anchor node in the cabinet through the cabinet management controller. The cabinet inspection command is parsed through the Bluetooth anchor node inside the cabinet, and the Bluetooth wake-up signal is extracted. Match the corresponding wireless sensor node based on the Bluetooth wake-up signal, and send a Bluetooth connection request and wake-up authentication code; The wireless sensor node is triggered to acquire data based on the Bluetooth connection request. After the wake-up authentication code is verified, the sensor data is acquired and sent sequentially to the Bluetooth anchor node in the cabinet, the cabinet management controller, and the maintenance terminal.
9. The wireless sensor node monitoring method based on rack management controller scheduling according to claim 7, characterized in that, The general interface asynchronous interrupt mechanism includes: The resistance value of the leakage sensor is detected and judged. If the resistance value is greater than the preset first resistance threshold, the environment of the liquid cooling cabinet is determined to be normal dry, triggering a high level on the general interface. If the resistance value is less than the preset second resistance threshold, the environmental condition of the liquid-cooled cabinet is determined to be partial leakage, triggering a low level on the general interface; Based on the low level of the general interface, the wireless sensor node establishes a Bluetooth connection with the Bluetooth anchor node inside the cabinet and sends a platform-level leakage alarm message. The platform-level leakage alarm message is forwarded to the cabinet management controller via the Bluetooth anchor node inside the cabinet; Based on the platform-level leakage alarm message, the rack management controller triggered an emergency procedure.
10. A storage medium storing at least one instruction, at least one program, a code set, or an instruction set, wherein the at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by a processor to implement the wireless sensor node monitoring method based on rack management controller scheduling as described in any one of claims 7-9.
