Service-level-perceptible computing power network dynamic scheduling system for broadcast and television services

Abstract

The invention provides a radio and television service-oriented computing power network dynamic scheduling system with a perceptible service level, and the system comprises a service portrait management module which is internally provided with static computing rules and parameters; the multi-resource monitoring agent module is used for continuously reporting the collected real-time system data; the comprehensive priority calculation engine is used for calculating a comprehensive priority score for each task based on a configured weighted priority formula and generating a task priority queue; the intelligent preemption decision module selects the task combination with the lowest total preemption cost as a target and issues a scheduling and preemption instruction; the multi-resource collaborative allocation module issues an allocation instruction to the infrastructure; and the dynamic feedback control module is used for aggregating the data of all the modules, comparing the monitoring data with a target threshold value, judging whether the running state is optimal or not, and dynamically generating an adjusting and optimizing instruction. According to the method, the multi-dimensional scheduling factors are fused into a quantifiable priority score through weighted summation calculation, so that subsequent intelligent decisions are driven.

Description

Technical Field

[0001] This application relates to the field of cable television network technology, and in particular to a service level-aware dynamic scheduling system for computing power networks for broadcast services. Background Technology

[0002] As media convergence deepens, the business model of the broadcasting industry is undergoing profound changes, shifting from traditional linear audio and video broadcasting to a comprehensive transformation encompassing diversified and intelligent scenarios such as ultra-high-definition live streaming, interactive VR / AR, AIGC content production, and cloud-based virtual production and broadcasting. The demands of these services on underlying information infrastructure have evolved from simple network connectivity to a comprehensive need for computing power and network collaborative scheduling—a concept known as "computing power network." Simultaneously, these services exhibit significantly different Quality of Service (QoS) requirements for the underlying computing power network infrastructure: live streaming services demand millisecond-level low latency and extremely high reliability, while services such as AIGC rendering prioritize high throughput and cost-effectiveness.

[0003] When these heterogeneous services run concurrently on a shared computing network infrastructure, they will compete fiercely for limited computing, network, and storage resources. Existing solutions such as static priority scheduling and simple preemptive scheduling are unable to achieve an intelligent balance between ensuring the SLA of critical services, improving the overall resource utilization of the system, and ensuring fairness to different services. They have exposed many limitations when dealing with complex mixed service scenarios. Summary of the Invention

[0004] In view of this, the present invention provides a service level-aware dynamic scheduling system for computing power networks for broadcasting services.

[0005] The technical solution of this invention proposes a service level-aware dynamic scheduling system for computing power networks for broadcasting services, comprising: The business profile management module has built-in static calculation rules and parameters; The multi-resource monitoring agent module continuously reports the collected real-time system data to the comprehensive priority calculation engine; The comprehensive priority calculation engine, based on the configured weighted priority formula and the calculation rules and parameters, calculates a real-time, quantitative comprehensive priority score for each task, and generates and maintains a global task priority queue. The intelligent preemption decision module receives the priority queue, uses an algorithm to select the task combination with the lowest total preemption cost as the target, and issues scheduling and preemption instructions to the multi-resource collaborative allocation module. The multi-resource collaborative allocation module issues resource allocation instructions to the underlying infrastructure and reports the execution results and other status information to the dynamic feedback control module. The dynamic feedback control module aggregates monitoring data from all modules, compares the monitoring data with preset target thresholds, determines whether the system's operating state is optimal, and dynamically generates optimization instructions based on the analysis results.

[0006] In one implementation, the calculation rules and parameters include: The service level rule L for all services is used to define the static priority of different service types and their weight coefficient α; Maintain the dynamic urgency model D for various business operations, and its weight coefficient β; Construct a cost factor library C to quantify the resource costs or business value of task execution, and its weighting coefficient γ.

[0007] In one implementation, the weighted priority formula is specifically as follows: Where P: Overall priority score, the higher the value, the higher the scheduling priority; For time-limited tasks: Where K is a preset amplification constant. The value of K is used to adjust the numerical range of the dynamic urgency level D to match the static service level L, so as to ensure that it can have an effective impact in the comprehensive priority calculation. T_remaining is the remaining time from the current time to the scheduled start time of the task. For tasks without a specified time limit: D = Completion_Ratio; where Completion_Ratio is the current percentage of the task's progress, meaning the higher the completion rate, the larger the D value. The value of C is taken as the reciprocal of the cost of occupying resources or the value of the task itself, in order to optimize costs. It is estimated based on the resource specifications requested by the task and the unit resource price. .

[0008] Another aspect of the present invention provides a service level-aware dynamic scheduling method for computing power networks for broadcasting services, comprising: Step 1: Receive new business tasks or detect events where multiple tasks are competing for the same scarce resource; Step 2: The system calls the comprehensive priority calculation engine to calculate or update the comprehensive priority score, i.e., P value, for the task, and determines whether there are enough idle resources to meet the needs of the task with the highest P value. Step 3: If so, the required computing power, network and storage resources are directly allocated to the task with the highest P value through the multi-resource collaborative allocation module. The process ends after the task is completed. Step 3: If not, trigger the intelligent preemption process.

[0009] In one implementation, the intelligent preemption process includes: Step A: List all low-P-value tasks that can be preempted; evaluate the preemption cost of each candidate task, which takes into account the sunk cost and restart overhead of the task; the system selects one or more task combinations with the lowest total preemption cost as the preemption target. Step B: For the selected low P-value tasks, perform a checkpoint save operation to persist their current running state instead of terminating them directly. Move the preempted tasks into a compensation queue with a high scheduling weight and immediately allocate resources to the high P-value tasks that are blocked due to insufficient resources. Step C: When the system has idle resources later, it prioritizes scheduling tasks in the compensation queue and resumes the transmission from the saved checkpoint until the task is completed.

[0010] Another aspect of the present invention provides a service level-aware dynamic scheduling method for computing power networks for broadcasting services, comprising: Regularly collect key performance indicator data from the business scheduling engine and resource monitoring agent module, including: service level agreement achievement rate for each business level, system global and local resource utilization, and success / failure rate of task scheduling and preemption. The collected real-time data is compared and analyzed with the preset system optimization target threshold. When one or more indicators deviate from the expected range, it is determined that the system is in a non-optimal operating state, thereby triggering parameter adjustment decisions. The dynamic feedback control module generates specific parameter adjustment instructions and sends them to the corresponding functional modules in the system to adjust the weight coefficients in the comprehensive priority calculation model or dynamically scale the size of the elastic resource pool reserved for a specific business level. Return to the data acquisition step to continuously monitor and evaluate the actual effect of this parameter adjustment.

[0011] Another aspect of the present invention provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the service level-aware dynamic scheduling method for computing power networks for broadcast services as described above.

[0012] Another aspect of the present invention provides a computer storage medium storing a computer program that is executed to implement the service level-aware dynamic scheduling method for computing power networks for broadcast services as described above.

[0013] By adopting the above technical solution, the present invention has at least the following advantages: The system provided by this invention employs an intelligent scheduling method based on a weighted comprehensive priority model. This method integrates multi-dimensional scheduling factors into a quantifiable priority score through a specific and explicit mathematical calculation of weighted summation, thereby driving subsequent intelligent decision-making. Attached Figure Description

[0014] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 This is a schematic diagram of a service level-aware dynamic scheduling system for computing power networks for broadcasting services according to an embodiment of the present invention. Figure 2 This is a flowchart illustrating a service level-aware dynamic scheduling method for computing power networks for broadcast services according to an embodiment of the present invention (resource scheduling and preemption). Figure 3 This is a flowchart illustrating a service level-aware dynamic scheduling method (dynamic feedback optimization) for broadcast services according to an embodiment of the present invention. Detailed Implementation

[0015] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the present invention will be described in detail below with reference to the accompanying drawings and preferred embodiments.

[0016] While exemplary embodiments of the invention are shown in the accompanying drawings, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the invention and to fully convey its scope to those skilled in the art. The invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0017] This invention provides a service level-aware dynamic scheduling system for computing power networks for broadcasting services, wherein the logical structure of the system is as follows: Figure 1 As shown, it includes: The business profile management module has built-in static calculation rules and parameters; The multi-resource monitoring agent module continuously reports the collected real-time system data to the comprehensive priority calculation engine; The comprehensive priority calculation engine, based on the configured weighted priority formula and the calculation rules and parameters, calculates a real-time, quantitative comprehensive priority score for each task, and generates and maintains a global task priority queue. The intelligent preemption decision module receives the priority queue, uses an algorithm to select the task combination with the lowest total preemption cost as the target, and issues scheduling and preemption instructions to the multi-resource collaborative allocation module. The multi-resource collaborative allocation module issues resource allocation instructions to the underlying infrastructure and reports the execution results and other status information to the dynamic feedback control module. The dynamic feedback control module aggregates monitoring data from all modules, compares the monitoring data with preset target thresholds, determines whether the system's operating state is optimal, and dynamically generates optimization instructions based on the analysis results.

[0018] The following will combine Figure 1 The framework structure and Figure 2 , Figure 3 The workflow of the system provided in this embodiment will be described in detail.

[0019] 1) Business Profile Management Module The business profile management module stores and manages the service level rules (L) for all businesses, defines the static priority of different business types, maintains the dynamic urgency model (D) for various businesses, and builds a cost factor library (C) to quantify the resource cost or business value of task execution.

[0020] Interaction relationship: Provides static calculation rules and parameters (definitions of L, D, C and weight coefficients α, β, γ) to the comprehensive priority calculation engine, which is the strategic basis for scheduling decisions.

[0021] 2) Multi-resource monitoring agent module The multi-resource monitoring agent module continuously collects real-time operational data from the underlying computing infrastructure (CPU / GPU utilization), network infrastructure (bandwidth, latency), and storage infrastructure (IOPS, capacity). It also monitors the task's running status (such as start, completion, and progress).

[0022] Interaction relationship: The collected real-time data is continuously reported to the comprehensive priority calculation engine, and raw monitoring data such as resource utilization and task status are provided to the dynamic feedback control module.

[0023] 3) Comprehensive Priority Calculation Engine The comprehensive priority calculation engine receives rules and real-time data from upstream modules. Based on the core algorithm of this invention—the weighted priority formula—it... For each task, a real-time, quantified comprehensive priority score (P) is calculated, and a global task priority queue is generated and maintained.

[0024] Interaction relationship: The calculated priority queue is sent to the intelligent preemption decision module, and the monitoring data such as its own queue status is reported to the dynamic feedback control module.

[0025] 4) Intelligent preemption decision module The intelligent preemption decision module monitors the system's resource status and is triggered when a high-priority task cannot be scheduled due to insufficient resources. It assesses the "cost" of preempting different low-priority tasks (including "sunk cost" and "restart overhead." Sunk cost: the quantification of resources already consumed by the task, such as GPU time and bandwidth. Restart overhead: the additional resources required to restore the task, including reloading the environment and the time spent calculating from the checkpoint). It then uses an algorithm to select the task combination with the lowest total preemption cost as the target. Finally, it issues a command to schedule the high-priority task, and performs checkpoint saving and moves the preempted task to a compensation queue for subsequent priority scheduling.

[0026] Interaction: It receives priority queues from the computing engine and issues scheduling and preemption instructions to the multi-resource collaborative allocation module. At the same time, it reports monitoring data such as preemption count and success rate to the dynamic feedback control module.

[0027] 5) Multi-resource collaborative allocation module The multi-resource collaborative allocation module receives instructions from the decision-making module; it collaboratively calls the interfaces of the underlying computing power management platform, SDN controller, and storage management system to ensure that computing power, network, and storage resources are allocated in an integrated manner to avoid resource bottlenecks; and it returns the success / failure confirmation information of resource allocation to the system.

[0028] Interaction relationship: Send resource allocation instructions to the underlying infrastructure and report the status such as the execution results of the instructions to the dynamic feedback control module. 6) Dynamic feedback control module The dynamic feedback control module aggregates monitoring data from all modules (such as SLA compliance rate, resource utilization rate, and preemption success rate). It compares the monitoring data with preset target thresholds to determine if the system's operating state is optimal. Based on the analysis results, it dynamically generates optimization instructions.

[0029] This invention explicitly defines the calculation method for the comprehensive priority score P as a weighted summation model, and its calculation formula is as follows: P: Overall priority score. The higher the value, the higher the scheduling priority.

[0030] L: Static service level, predefined by the business profile.

[0031] D: Dynamic urgency factor, which quantifies the real-time urgency of a task. Its calculation method is as follows: For time-limited tasks: Where K is a preset amplification constant. The value of K is used to adjust the numerical range of the dynamic urgency level D to match the static service level L, ensuring that it can have an effective impact in the comprehensive priority calculation. The specific value of K can be configured according to the sensitivity of different business scenarios to urgency. T_remaining is the remaining time (in seconds) from the current time to the scheduled start time of the task.

[0032] For tasks with no explicit time limit but long duration: D = Completion_Ratio (the current percentage of the task's progress). The higher the completion rate, the larger the D value, resulting in a higher urgency level. This allows the task to be prioritized in resource competition, effectively avoiding the huge sunk cost waste caused by competing for high-completion tasks, and improving the overall economy and fairness of the system.

[0033] C: Cost factor, typically negative, representing the cost of resource usage or the reciprocal of the task's value, used for cost optimization. It can be estimated based on the resource specifications requested by the task (e.g., GPU hours, bandwidth) and the unit resource price. By introducing the cost factor C, the system can proactively optimize the economics of resource allocation in scheduling decisions, significantly reducing the overall operating cost of the computing network.

[0034] α, β, γ: are the weighting coefficients of L, D, and C, respectively, and satisfy the following conditions: The initial values ​​can be set by the system administrator according to business strategies and optimized online by the dynamic feedback control module.

[0035] The second embodiment of this invention provides a service level-aware dynamic scheduling method for computing power networks for broadcasting services, which can be referred to again. Figure 2 It is understood that the testing method provided in this embodiment is an application of the system provided in the first embodiment. This embodiment aims to solve the scheduling strategy conflict problem caused by resource contention in a multi-service mixed deployment scenario. Specifically, it includes the following steps: 1) Process initialization and event triggering The process begins when the system receives a new business task or detects an event where multiple tasks are competing for the same scarce resource.

[0036] 2) Comprehensive priority calculation and resource judgment The system invokes the overall priority calculation engine to calculate or update the overall priority score (P-value) for the relevant tasks. Then, the system determines whether there are sufficient idle resources to meet the needs of the task with the highest P-value.

[0037] 3) Normal scheduling path If the judgment result of step 2) is "yes" (i.e., resources are sufficient), then the normal scheduling process begins. The system directly allocates the required computing power, network, and storage resources to the task with the highest P value through the multi-resource collaborative allocation module. The process ends after the task is completed.

[0038] 4) Intelligent decision-making and execution path If the result of step 2) is "no" (i.e., insufficient resources), then the intelligent preemption process is triggered.

[0039] ① Intelligent preemption decision: The system first lists all low-P-value tasks that can be preempted; then, it evaluates the preemption cost of each candidate task, which takes into account the sunk cost (the value of resources already consumed) and restart overhead (the additional cost required to restore the task); finally, the system selects one or more task combinations with the lowest total preemption cost as the preemption target.

[0040] ② Perform preemption compensation: For the low-P-value tasks selected in step ①, perform a checkpoint save operation to persist their current running state instead of terminating them directly. Move the preempted tasks into a compensation queue with a high scheduling weight. Immediately allocate resources to the high-P-value tasks that are blocked due to insufficient resources.

[0041] ③ Compensation callback: When the system has idle resources later, it will prioritize scheduling the tasks in the compensation queue and resume them from the saved checkpoints until the tasks are completed.

[0042] The third embodiment of this invention provides a service level-aware dynamic scheduling method for computing power networks for broadcasting services, which can be referred to again. Figure 3 It is understood that the testing method provided in this embodiment is an application of the system provided in the first embodiment. This embodiment aims to achieve automatic optimization of system parameters through closed-loop control to adapt to dynamic changes in business load and continuously improve overall performance. Specifically, it includes the following steps: 1) Data Collection Key performance metrics data are collected regularly from core modules such as the business scheduling engine and resource monitoring agent. These mainly include: service level agreement (SLA) achievement rate for each business level, system-wide and local (e.g., edge node) resource utilization, and task scheduling and preemption success / failure rates.

[0043] 2) Analysis and Decision Making The collected real-time data is compared and analyzed with preset system optimization target thresholds. When one or more indicators deviate from the expected range, the system is determined to be in a suboptimal operating state, thus triggering parameter adjustment decisions.

[0044] 3) Parameter adjustment and execution Based on the analysis results, the dynamic feedback control module generates specific parameter adjustment instructions and sends them to the corresponding functional modules in the system. These adjustment instructions aim to fundamentally optimize the system's scheduling strategy and resource allocation behavior, including adjusting the weight coefficients (α, β, γ) in the comprehensive priority calculation model, or dynamically scaling the size of the elastic resource pool reserved for specific business levels.

[0045] 4) Closed-loop feedback and continuous optimization After the parameter adjustment command is executed, the process does not terminate, but immediately returns to step (1), i.e., the data acquisition step, to continuously monitor and evaluate the actual effect of the parameter adjustment. This constitutes a complete closed-loop feedback control loop of "acquisition-analysis-decision-adjustment-reacquisition". Through this closed loop, the system can continuously learn and self-correct, thereby maintaining a high-performance and highly adaptable optimal operating state for a long time.

[0046] According to a fourth embodiment of the present invention, an electronic device includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the service level-aware dynamic scheduling method for computing power networks for broadcast services as described in the second and third embodiments.

[0047] According to the fifth embodiment of the present invention, a computer storage medium is provided, wherein a computer program is stored on the medium, and the computer program is executed to implement the service level-aware dynamic scheduling method for computing power networks for broadcasting services as described in the second and third embodiments.

[0048] Compared with the prior art, this embodiment has at least the following advantages: (1) Multi-dimensional comprehensive priority model: using a weighted summation formula This approach quantifies business scheduling decisions into a calculable score. It overcomes the limitations of traditional single-indicator scheduling, achieving multi-objective collaborative optimization of business priority, real-time urgency, and economic cost, thus providing a precise data foundation for subsequent intelligent scheduling.

[0049] (2) Collaborative design and functional integration of system modules: such as Figure 1 As shown, the system consists of a business profile management module, a multi-resource monitoring agent module, an intelligent preemption decision module, a multi-resource collaborative allocation module, and a dynamic feedback control module. These modules interact through clear data and control flows, forming an organic whole. This layered and decoupled design ensures the system's scalability and maintainability, while the dedicated responsibilities of each module achieve high cohesion and low coupling, improving the overall efficiency and reliability of the system.

[0050] (3) Cost-Aware Intelligent Preemption and Compensation Mechanism: This mechanism addresses the resource waste and fairness issues inherent in traditional preemption mechanisms. When resource competition occurs, the system does not simply prioritize tasks but evaluates the "total preemption cost" of each candidate task, selecting the combination of tasks with the lowest total cost as the preemption target, thereby minimizing global losses. Subsequently, the system saves checkpoints (persistent runtime states) for the preempted tasks and moves them to a high-weight compensation queue. When resources are idle, these tasks are prioritized for scheduling and resumed from where they left off. This mechanism not only avoids the waste of sunk costs but also ensures the fairness of low-priority task completion through the compensation queue.

[0051] (4) Dynamic feedback closed-loop optimization process: The process includes four steps: data collection, analysis and decision-making, parameter adjustment and closed-loop feedback, which enables the system to maintain optimal performance in the long term, adapt to changes in business load, and improve the intelligence and robustness of the system without manual intervention.

[0052] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims. All of these forms are within the protection scope of this application.

Claims

1. A service level-aware dynamic scheduling system for computing power networks for broadcasting services, characterized in that: include: The business profile management module has built-in static calculation rules and parameters; The multi-resource monitoring agent module continuously reports the collected real-time system data to the comprehensive priority calculation engine; The comprehensive priority calculation engine, based on the configured weighted priority formula and the calculation rules and parameters, calculates a real-time, quantitative comprehensive priority score for each task, and generates and maintains a global task priority queue. The intelligent preemption decision module receives the priority queue, uses an algorithm to select the task combination with the lowest total preemption cost as the target, and issues scheduling and preemption instructions to the multi-resource collaborative allocation module. The multi-resource collaborative allocation module issues resource allocation instructions to the underlying infrastructure and reports the execution results and other status information to the dynamic feedback control module. The dynamic feedback control module aggregates monitoring data from all modules, compares the monitoring data with preset target thresholds, determines whether the system's operating state is optimal, and dynamically generates optimization instructions based on the analysis results.

2. The service level-aware dynamic scheduling system for computing power networks for broadcasting services according to claim 1, characterized in that, The calculation rules and parameters include: The service level rule L for all services is used to define the static priority of different service types and their weight coefficient α; Maintain the dynamic urgency model D for various business operations, and its weight coefficient β; Construct a cost factor library C to quantify the resource costs or business value of task execution, and its weighting coefficient γ.

3. The service level-aware dynamic scheduling system for computing power networks for broadcasting services according to claim 2, characterized in that, The weighted priority formula is as follows: Where P: Overall priority score, the higher the value, the higher the scheduling priority; For time-limited tasks: Where K is a preset amplification constant. The value of K is used to adjust the numerical range of the dynamic urgency level D to match the static service level L, so as to ensure that it can have an effective impact in the comprehensive priority calculation. T_remaining is the remaining time from the current time to the scheduled start time of the task. For tasks without a specified time limit: D = Completion_Ratio; where Completion_Ratio is the current percentage of the task's progress, meaning the higher the completion rate, the larger the D value. The value of C is taken as the reciprocal of the cost of occupying resources or the value of the task itself, in order to optimize costs. It is estimated based on the resource specifications requested by the task and the unit resource price. 。 4. A service level-aware dynamic scheduling method for computing power networks for broadcasting services, characterized in that: include: Step 1: Receive new business tasks or detect events where multiple tasks are competing for the same scarce resource; Step 2: The system calls the comprehensive priority calculation engine to calculate or update the comprehensive priority score, i.e., P value, for the task, and determines whether there are enough idle resources to meet the needs of the task with the highest P value. Step 3: If so, the required computing power, network and storage resources are directly allocated to the task with the highest P value through the multi-resource collaborative allocation module. The process ends after the task is completed. Step 3: If not, trigger the intelligent preemption process.

5. The service level-aware dynamic scheduling method for computing power networks for broadcasting services according to claim 4, characterized in that, The intelligent preemption process includes: Step A: List all low-P-value tasks that can be preempted; evaluate the preemption cost of each candidate task, which takes into account the sunk cost and restart overhead of the task; the system selects one or more task combinations with the lowest total preemption cost as the preemption target. Step B: For the selected low P-value tasks, perform a checkpoint save operation to persist their current running state instead of terminating them directly. Move the preempted tasks into a compensation queue with a high scheduling weight and immediately allocate resources to the high P-value tasks that are blocked due to insufficient resources. Step C: When the system has idle resources later, it prioritizes scheduling tasks in the compensation queue and resumes the transmission from the saved checkpoint until the task is completed.

6. A service level-aware dynamic scheduling method for computing power networks for broadcasting services, characterized in that, include: Regularly collect key performance indicator data from the business scheduling engine and resource monitoring agent module, including: service level agreement achievement rate for each business level, system global and local resource utilization, and success / failure rate of task scheduling and preemption. The collected real-time data is compared and analyzed with the preset system optimization target threshold. When one or more indicators deviate from the expected range, it is determined that the system is in a non-optimal operating state, thereby triggering parameter adjustment decisions. The dynamic feedback control module generates specific parameter adjustment instructions and sends them to the corresponding functional modules in the system to adjust the weight coefficients in the comprehensive priority calculation model or dynamically scale the size of the elastic resource pool reserved for a specific business level. Return to the data acquisition step to continuously monitor and evaluate the actual effect of this parameter adjustment.

7. An electronic device, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the service level-aware dynamic scheduling method for computing power networks for broadcast services as described in any one of claims 4 to 6.

8. A computer storage medium, characterized in that, The medium stores a computer program that is executed to implement the service level-aware dynamic scheduling method for computing power networks for broadcast services as described in any one of claims 4 to 6.

Images