A dynamic scheduling system and method for spaceborne computing resources for integrated space-ground root services

Abstract

This invention discloses a dynamic scheduling system and method for spaceborne computing resources for integrated space-ground root services. The system includes a multi-dimensional perception module, a computing power modeling and abstraction module, a service grid management module, a dynamic scheduling decision engine, a computing power migration and state synchronization module, and a graceful degradation survival assurance module. The multi-dimensional perception module collects real-time data on spaceborne computing power, network links, and service load status; the computing power modeling and abstraction module constructs a global dynamic computing power graph; the service grid management module encapsulates root service functions into microservices; the dynamic scheduling decision engine dynamically allocates computing power nodes to microservices based on deep reinforcement learning; the computing power migration module enables cross-node migration of service instances; and the graceful degradation survival assurance module achieves millisecond-level thermal control and reconfiguration in extreme space environments. This invention achieves globally optimized utilization of spaceborne computing power while ensuring the survivability of computing power nodes in extreme space environments.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of space information network and computing power network technology, and specifically relates to a dynamic scheduling system and method for spaceborne computing power resources for integrated space-ground root service, providing survivability assurance for computing power nodes in extreme space environments for space-based distributed root service systems. Background Technology

[0002] With the rapid development of low-Earth orbit broadband satellite constellations, building space-based information networks has become a global trend. The patent "A Space-Based Distributed Root Service System Coexisting with the Existing Internet Root System" proposes deploying Internet root services on satellite constellations to solve problems such as high access latency in remote areas and fragile ground infrastructure.

[0003] However, the inventors discovered new technical challenges in practice: First, the space-based root service components have vastly different and dynamically changing on-board computing power requirements, and a fixed computing power allocation model leads to low resource utilization or performance bottlenecks. Second, the high-speed movement of satellite nodes necessitates frequent service switching, and traditional virtual machine or container migration technologies cannot guarantee the continuity of stateful services in scenarios with high latency and intermittent connections in inter-satellite links. Finally, existing technologies generally neglect the survivability of onboard computing nodes in extreme space environments: micrometeoroid impacts, space debris collisions, and extreme thermal stress can cause localized failures of computing nodes. Traditional solutions rely on hardware redundancy, but this approach is costly, heavy, and cannot cope with simultaneous failures at multiple points.

[0004] While existing technologies include research on satellite edge computing and on-board processing, none address how to provide a deeply coupled, dynamically adaptive, task-aware computing power scheduling system with hardware survivability guarantees for critical national network infrastructure such as space-based root services. Therefore, an innovative solution is urgently needed to address these issues. Summary of the Invention

[0005] To address the aforementioned technical challenges, this invention proposes a dynamic scheduling system and method for spaceborne computing resources for integrated space-ground root services. Its core objective is to transform dispersed spaceborne computing power into a flexible, reliable, and intelligent computing resource pool, providing high-quality and highly available computing power guarantees for space-based root services. Simultaneously, by integrating graceful degradation survival assurance technology, it enables computing nodes to achieve millisecond-level self-healing capabilities in extreme space environments.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: This system includes a multi-dimensional perception module, a computing power modeling and abstraction module, a service grid management module, a dynamic scheduling decision engine, a computing power migration and state synchronization module, and a graceful degradation survival guarantee module.

[0007] The multi-dimensional sensing module is used to collect real-time data from satellite platforms, network links, and root service systems, including computing resource utilization, inter-satellite link quality, service load status, and the real-time temperature of each computing node.

[0008] The computing power modeling and abstraction module is used to abstract heterogeneous hardware resources on different satellites into standardized computing power units and construct a global dynamic computing power map, which reflects the distribution of computing power resources, network connection costs, and thermal capacity margin of each node.

[0009] The service mesh management module is used to decompose and encapsulate the core functions of the Tianji root service into lightweight, independently manageable microservices. Each microservice has a resource requirement tag and service level agreement requirements.

[0010] The dynamic scheduling decision engine is used to run a scheduling strategy model based on the global dynamic computing power graph and the queue of microservices to be scheduled, and make a globally optimal scheduling decision.

[0011] The computing power migration and state synchronization module is used to safely and efficiently migrate microservice instances and their complete states from one satellite to another when scheduling decisions require service migration or satellite switching is predicted, ensuring service continuity.

[0012] The graceful degradation survival assurance module is used to collect real-time temperature data of computing nodes through the status monitoring subunit, replan the computing task path for overheated components through the thermal control path planning subunit, adjust the heat dissipation power through the local thermal control subunit, predict failure risks in advance through the pre-reconstruction subunit, and form a closed-loop optimization through the effect evaluation subunit to ensure the survivability of computing nodes in extreme space environments.

[0013] Compared with the prior art, the present invention has the following beneficial effects: (1) It realizes the on-demand allocation of computing resources and improves the utilization rate of onboard computing power through dynamic scheduling; (2) It ensures the continuity of critical services and ensures the high availability of Tianjigen services through predictive computing power migration and state synchronization; (3) It endows the system with task awareness, and the scheduling strategy can distinguish task priorities and prioritize the protection of critical tasks; (4) A spaceborne service grid architecture was constructed, laying the foundation for deploying more network applications on satellite constellations; (5) It achieves the survivability guarantee of computing nodes and copes with hardware failure in extreme space environments through graceful degradation and reconstruction technology; (6) A closed loop of thermal-computation collaborative optimization was formed, achieving optimal system-level energy efficiency. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is the overall architecture diagram of the system of the present invention in a space-ground integrated network; Figure 2 This is a flowchart illustrating the internal module interactions and workflow of the system of this invention; Figure 3 This is a schematic diagram of the global dynamic computing power graph in this invention; Figure 4 This is a schematic diagram of the prediction-based computing power migration process in this invention; Figure 5 This is a diagram illustrating the structure and workflow of the graceful degradation survival assurance module in this invention. Figure 6 This is a schematic diagram of the closed-loop thermal-computational collaborative optimization in this invention; Figure 7 This is a schematic diagram of the three-dimensional temperature field map and pre-reconstruction trigger of the computing node in this invention. Detailed Implementation

[0016] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0017] like Figure 1 As shown, this system operates within a space-ground integrated network consisting of a low-Earth orbit satellite constellation, high-Earth orbit satellites, ground gateway stations, and user terminals. In the system, each satellite node deploys the computing power scheduling agent of this invention, which coexists with components of the space-based distributed root service system. Each computing power node also integrates a graceful degradation survival assurance module.

[0018] like Figure 2 As shown, the workflow of this system is as follows: Step S201: The multi-dimensional sensing module continuously collects the computing power resource indicators, network indicators, and service requests from the root service system of each satellite node; Step S202: The computing power modeling and abstraction module updates the global dynamic computing power graph based on the data from S201; Step S203: The service mesh management module receives a microservice deployment or invocation request initiated by the root service system; Step S204: The dynamic scheduling decision engine takes the current computing power graph status and microservice requirements as input, feeds them into the scheduling strategy model, and outputs the optimal scheduling scheme. Step S205: The computing power migration and status synchronization module executes the scheduling scheme. If migration is required, the microservice and its running status are synchronized to the target satellite through the inter-satellite link. Step S206: The service mesh management module on the target satellite starts microservice instances and begins processing service requests; Step S207: The graceful degradation survival protection module runs in parallel, continuously monitoring the temperature of the computing nodes. When an abnormal temperature is detected, the reconstruction process is triggered.

[0019] Example 1: Computing power migration and thermal control coordination in root zone key rolling update

[0020] Scenario: The root zone key signing key is updated on a rolling basis. This task is considered the highest priority consensus microservice.

[0021] Implementation process: 1. Prediction: The system predicts that the current consensus group master node responsible for the East Asia region will fly away from the coverage area in a few minutes; 2. Scheduling and Migration: The dynamic scheduling decision engine makes decisions, and the instruction computing power migration module synchronizes the consensus microservice and its current state to the satellite to be taken over; 3. Temperature sensing: The multi-dimensional sensing module detects that the source node's temperature is rising due to high-load computing. The dynamic scheduling decision engine adjusts its strategy to migrate some non-critical computing tasks to other nodes. 4. Graceful Degradation Assurance: When the source node temperature continues to rise to a dangerous threshold, the graceful degradation survival assurance module is triggered to plan alternative computing paths in advance; 5. Seamless takeover: Once the target satellite is ready, seamless takeover of consensus leadership ensures uninterrupted key rolling update process.

[0022] Example 2: Resilient Response to Sudden Regional Traffic Surges

[0023] Scenario: The opening of a global sporting event caused a surge in DNS query requests in a certain region.

[0024] Implementation process: 1. Perception: The multi-dimensional perception module detected a sharp increase in DNS query load on satellite nodes covering the area, indicating that computing resources were about to be exhausted and response latency was increasing; 2. Decision-making: Based on the global dynamic computing power map, the dynamic scheduling decision engine discovers that the satellites in the nearby airspace have a light load and a low temperature, and decides to elastically expand the intelligent DNS resolution microservice instances on the nearby satellites. 3. Execution: The service grid management module quickly starts a new parsing instance on the target satellite according to the scheduling instructions, and the computing power migration module configures a new service route; 4. Results: The newly added computing instances were put into service to share the surge in query traffic, and the overall DNS resolution latency in the region returned to normal levels.

[0025] Example 3: Preemptive scheduling and thermal control priority protection of computing power for high-security tasks

[0026] Scenario: To perform the root zone key signature key rolling update task, it is necessary to ensure that the satellite node group performing the task has sufficient computing power resources.

[0027] Implementation process: 1. Mission Issuance and Marking: The ground control center issues missions of the highest security level, with clear resource requirements; 2. Global Resource Awareness and Pre-clearing: The multi-dimensional awareness module responds to priority tags and scans the current network computing resources; 3. Preemptive scheduling decision: The engine preempts computing power for non-critical tasks, migrating them to other nodes or suspending them; 4. Temperature Reserve: The scheduling engine checks the temperature status of the target node to ensure that there is sufficient thermal capacity margin to cope with high-load computing. 5. Resource reservation and task deployment: After the target node's computing power resources are cleared, the service mesh management module deploys consensus microservice instances, and the scheduling engine sets resource occupancy locks; 6. Efficient execution of the consensus process: With exclusive high-quality computing power, the consensus group satellites operate at full speed, and the mission is successfully completed; 7. Post-event recovery: After the task is successful, the scheduling system releases the resource lock, and the suspended task is automatically resumed.

[0028] Example 4: Intelligent Sensing and Thermal Adaptation of Heterogeneous Computing Power on Spaceborne Hardware

[0029] Scenario Background: The security enhancement module of the Tianji Root Service System requires a large number of post-quantum cryptographic operations, which puts a heavy burden on the general-purpose CPU.

[0030] Implementation process: 1. Sensing and Registration: The multi-dimensional sensing module accurately identifies and registers the heterogeneous computing units of each satellite; 2. Requirement Labeling: Security-enhanced microservices are labeled with their preference for heterogeneous computing units during deployment; 3. Temperature-aware scheduling: When a query requiring post-quantum cryptography arrives, the dynamic scheduling decision engine queries the global dynamic computing power graph and, in conjunction with the temperature status of each node, schedules the task to a node with a moderate load and low temperature in the heterogeneous computing unit. 4. Effect: Cryptographic operations are performed efficiently on heterogeneous computing units, while avoiding overheated nodes.

[0031] Example 5: Performance Verification of Graceful Degradation Survival Assurance

[0032] This embodiment verifies the performance of the graceful degradation survivability assurance module on an onboard computing node. Verification is based on a hardware-in-the-loop simulation test conducted on a space environment simulator using an onboard computer platform. Environmental parameters include: temperature range -40℃ to 85℃, vacuum degree ≤1×10⁻⁶. -5 Pa.

[0033] The condition monitoring subunit collects the temperature of the CPU, GPU, FPGA, and storage unit in real time with a high sampling rate, forming a three-dimensional temperature field map.

[0034] The thermal control path planning subunit employs a GPU-based parallel shortest path algorithm to replan computational task paths for overheated components. Testing has shown that when the number of GPU cores reaches a certain scale, the single path planning time can meet millisecond-level real-time requirements. The new path satisfies the following conditions: connecting the heat sink unit to all healthy components; the computational load on each edge does not exceed a preset percentage of its maximum capacity; and the temperature difference between adjacent components does not exceed a preset threshold.

[0035] The local thermal control subunit adjusts the heat dissipation power of the components surrounding the damaged area through a MEMS variable thermal resistance structure.

[0036] The pre-reconfiguration sub-unit predicts high-failure-risk areas based on a hybrid neural network model. The model takes historical temperature fluctuations, vibration characteristics, cumulative operating time, computing load rate, and task priority as inputs, and outputs a predicted failure probability. Pre-reconfiguration is triggered when the predicted probability exceeds a preset threshold, allowing for advance planning of alternative computation paths.

[0037] The performance evaluation sub-unit calculates temperature uniformity and heat dissipation efficiency indicators to determine whether a secondary reconstruction is triggered.

[0038] Test results show that the graceful degradation survival guarantee module can achieve millisecond-level thermal control reconstruction in multi-point failure scenarios, effectively maintaining the heat dissipation capacity and service continuity of computing nodes.

[0039] Example 6: Closed-loop optimization of heat and computation

[0040] This embodiment is used to verify the closed loop of heat-computation collaborative optimization.

[0041] Closed-loop process: 1. Scheduling Decision: The dynamic scheduling decision engine allocates tasks based on the global dynamic computing power graph, prioritizing the scheduling of high-load tasks to nodes with lower temperatures and sufficient heat capacity. 2. Temperature change: As the computational load increases, the node temperature gradually rises; 3. Status Monitoring: The status monitoring subunit of the graceful degradation survival module monitors temperature changes in real time. When the node temperature exceeds the warning threshold, it notifies the scheduling engine. 4. Dynamic adjustment: The scheduling engine responds to temperature warnings and prioritizes assigning subsequent tasks to other nodes with lower temperatures; 5. Thermal Control Reconfiguration: If the node temperature continues to rise to the dangerous threshold, the graceful degradation module triggers thermal control reconfiguration, and the scheduling engine suspends the allocation of new tasks to that node; 6. Feedback: After the reconstruction is completed, the graceful degradation module will report the recovery status of the node's thermal capacity margin to the scheduling engine as the basis for the next round of scheduling.

[0042] The thermal-computation collaborative optimization closed loop achieves optimal energy efficiency at the system level, avoids node failure due to overheating, and ensures the continuity of task execution.

[0043] Industrial applicability

[0044] This invention provides a dynamic scheduling system and method for spaceborne computing resources for integrated space-ground root services, which has broad industrial applicability. The system is based on deep reinforcement learning, graph theory, and optimization theory, and can be implemented on FPGAs, GPUs, or dedicated ASICs. The multi-dimensional perception module can be implemented using a high-speed ADC in collaboration with an FPGA; the dynamic scheduling decision engine can deploy a deep reinforcement learning model using an inference engine; the thermal control path planning unit of the graceful degradation survival assurance module can adopt a GPU parallel computing architecture, and the pre-reconstruction unit can be deployed using a hybrid neural network model.

[0045] Through hardware-in-the-loop simulation verification on a space-based satellite platform, the system has complete functions and can be directly applied to the onboard computing nodes of the space-based distributed root service system, providing them with survivability assurance in extreme space environments.

Claims

1. A dynamic scheduling system for spaceborne computing resources for integrated space-ground root services, characterized in that, The system includes: The multi-dimensional sensing module is used to collect information on the computing resources status, network link status, service load status, and service demand of satellite nodes in real time. The computing power modeling and abstraction module is used to abstract heterogeneous onboard computing resources into measurable computing power units and construct a global dynamic computing power graph. The service mesh management module is used to encapsulate the functional components of the space-based distributed root service system into microservices that can be independently deployed, discovered, and communicated. The dynamic scheduling decision engine, based on the information from the multi-dimensional perception module and preset optimization objectives, dynamically allocates appropriate onboard computing units and execution nodes to the microservice instance through a scheduling strategy model; The computing power migration and state synchronization module is used to migrate microservice instances and their runtime contexts between satellite nodes based on scheduling decisions, and to ensure the consistency of service states. The graceful degradation survival module, integrated into the onboard computing node, is used to maintain the continuity of computing services through dynamic reconstruction when local hardware failures occur due to extreme space environments.

2. The system according to claim 1, characterized in that, The graceful degradation survival protection module includes: The status monitoring subunit is used to collect real-time data on the temperature, power consumption, and health status of computing nodes. The thermal control path planning subunit is used to replan the calculation task path for overheated components; The local thermal control subunit is used to adjust the heat dissipation power of components surrounding the damaged area; The pre-reconfiguration sub-unit is used to predict high-failure-risk areas and plan backup computing power paths in advance. The performance evaluation sub-unit is used to evaluate the temperature uniformity and heat dissipation efficiency after reconstruction and to trigger a secondary reconstruction.

3. The system according to claim 1, characterized in that, The global dynamic computing power graph constructed by the computing power modeling and abstraction module has node attributes including at least CPU core available cycles, remaining memory capacity, storage I / O bandwidth, available heterogeneous computing units, current temperature and temperature change rate. Its edge attributes include at least the propagation delay, available bandwidth, and packet loss rate of inter-satellite links and satellite-to-ground links.

4. The system according to claim 1, characterized in that, The scheduling strategy model of the dynamic scheduling decision engine is a deep reinforcement learning model. Its state space includes the real-time state of the global dynamic computing power graph, the resource requirement description of the microservice to be scheduled, and the real-time temperature and heat capacity margin of each computing power node. Its action space is the mapping relationship between microservice instances and onboard computing power units. Its reward function is constructed based on the overall system service latency, computing power resource utilization, overall system energy consumption, task completion success rate, and temperature uniformity.

5. The system according to claim 4, characterized in that, When making decisions, the scheduling strategy model assigns higher scheduling priority to the root zone data update consensus task and the high-security-level DNS query task from the space-based distributed root service system, and prioritizes scheduling high-priority tasks to nodes with temperatures below a preset threshold based on the current temperature of the computing power nodes.

6. The system according to claim 1, characterized in that, The computing power migration and status synchronization module, when predicting that the current service satellite will soon be unable to provide services, migrates the microservice instance and its running context to subsequent satellite nodes within the predicted coverage area in advance based on satellite ephemeris prediction; at the same time, when predicting that the temperature of the current node exceeds the warning threshold, it migrates the computing tasks to adjacent nodes with temperatures below the preset threshold in advance.

7. The system according to claim 1, characterized in that, The service mesh management module has built-in service circuit breaking, degradation and load balancing mechanisms. When it detects that the microservice instance of a satellite node has a timeout response or an error rate exceeding the threshold, it automatically switches traffic to other healthy instances and marks the faulty node. When the fault is caused by overheating, it notifies the graceful degradation survival protection module to handle the heat dissipation reconstruction of the node.

8. The system according to claim 2, characterized in that, The pre-reconstruction subunit predicts high failure risk areas based on a hybrid neural network model. The input features include at least one of historical temperature fluctuations, vibration characteristics, cumulative working time, computing power load rate, and task priority. The output is a predicted failure probability value. Pre-reconstruction is triggered when the predicted probability exceeds a preset threshold.

9. The system according to claim 2, characterized in that, The thermal control path planning subunit uses a GPU-based parallel shortest path algorithm for path planning, and the computational task path generated by the path planning satisfies: Path integrity constraint: The new path connects the heat sink unit to all healthy computing power components; Traffic capacity constraint: The computational load on each edge shall not exceed a preset percentage of its maximum capacity. Temperature gradient constraint: The temperature difference between adjacent components after reconstruction shall not exceed a preset threshold.

10. The system according to claim 2, characterized in that, The graceful degradation survival assurance module receives priority instructions from the dynamic scheduling decision engine. When a computing node fails, it prioritizes the hot-control reconstruction of the node carrying the root zone data update consensus task and the high-security-level DNS query task.

11. The system according to claim 2, characterized in that, The state monitoring subunit is also configured to: collect the temperature and MEMS thermal control structure status of each computing unit in real time, generate a three-dimensional temperature field map of the computing node based on the collected data, and use the three-dimensional temperature field map as the input feature of the pre-reconstruction subunit.

12. The system according to claim 1, characterized in that, The graceful degradation survival guarantee module and the dynamic scheduling decision engine form a collaborative optimization closed loop: the dynamic scheduling decision engine adjusts task allocation according to the temperature of the computing node; the graceful degradation survival guarantee module triggers reconstruction when the node temperature is abnormal, and feeds back the reconstructed heat capacity margin to the dynamic scheduling decision engine as the basis for the next round of scheduling.

13. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the system as described in any one of claims 1 to 12.

14. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the system as described in any one of claims 1 to 12.

Images