A method, device, and medium for dynamic task isolation and scheduling based on time constraint vectors.
By adopting a task dynamic isolation scheduling method based on time constraint vectors, the entry opportunities of non-real-time tasks are dynamically adjusted, which solves the dynamic balance problem of real-time and non-real-time tasks in multi-core processors and realizes stable execution and flexible scheduling among tasks.
Patent Information
- Application Number
- CN202611601448.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-16
- Publication Date
- 2026-11-17
AI Technical Summary
In multi-core processors, existing technologies struggle to achieve dynamic balance and stable execution between tasks in the dynamic running states of real-time and non-real-time tasks. In particular, when the running states of real-time and non-real-time tasks are constantly changing, scheduling methods with fixed priorities and fixed running regions cannot guarantee the stability between tasks.
A task dynamic isolation scheduling method based on time constraint vectors is adopted. By generating time constraint vectors corresponding to real-time job instances, collecting the current actual execution results, comparing the time constraint vectors with the actual execution results, determining the admission restrictions and execution budget in the scheduling window, and forming a protected execution window in the next scheduling window, the entry opportunities of non-real-time job instances are dynamically adjusted to form a dynamic isolation scheduling closed loop.
It achieves dynamic balancing and stable execution of real-time and non-real-time job instances, reduces the possibility of long-term waiting for non-real-time job instances, and improves the stability and flexibility of task scheduling.
Smart Images

Figure CN123114707A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of task scheduling technology, and in particular to a method, device and medium for dynamic isolation scheduling of tasks based on time constraint vectors. Background Technology
[0002] In industrial control, vehicle control, embedded control, edge computing control, and other real-time control scenarios, multi-core processors typically need to run real-time and non-real-time tasks simultaneously. Real-time tasks usually have explicit time constraints, such as release time, start time, completion time, deadline, allowable jitter, allowable execution duration, and expected number of executions. While non-real-time tasks have relatively less stringent requirements regarding completion time, they still have upper bounds on waiting and runnability requirements. Therefore, how to balance the determinism of real-time job instances and the runnability of non-real-time job instances in multi-core processors is a crucial problem that scheduling systems need to solve.
[0003] To address the aforementioned issues, existing technologies assign fixed priorities based on task cycles and worst-case execution times, ensuring that real-time tasks (real-time job instances) are executed before non-real-time tasks (non-real-time job instances); alternatively, during system initialization, processor cores are divided into real-time and non-real-time execution regions, allowing different tasks to run within a preset range. These methods can improve the execution stability of real-time tasks to some extent when the task set is stable, the operational relationships are simple, or the resource contention pattern is relatively fixed.
[0004] However, in reality, the running states of real-time and non-real-time tasks are constantly changing. Relying solely on static task attributes, fixed priorities, fixed running areas, or single scheduling configurations is insufficient to guarantee dynamic balance and stable execution among tasks. Summary of the Invention
[0005] To address the aforementioned problems and technical requirements, the applicant proposes a task dynamic isolation scheduling method, device, and medium based on time constraint vectors. This method aims to solve the problem of poor execution stability in existing technologies when real-time job instances and non-real-time job instances run together, thereby achieving dynamic balance and stable execution of real-time job instances and non-real-time job instances.
[0006] This application provides a method for dynamic task isolation scheduling based on time constraint vectors. The method includes: Obtain the tasks to be processed and create job instances corresponding to the tasks to be processed. Job instances include: real-time job instances and non-real-time job instances. Generate a time constraint vector corresponding to the real-time job instance, and collect the current actual execution result of the real-time job instance in the current observation window, as well as the current actual execution status of the non-real-time job instance. The time constraint vector is used to describe the expected constraints of the real-time job instance in the observation window. Compare the time constraint vector with the current actual execution result. Based on the comparison result, determine if there are cases where the windowing constraint is not satisfied in the classification, and obtain the current classification result where the windowing constraint is not satisfied. Based on the current classification results, determine the admission restrictions and execution budget for non-real-time job instances in the next scheduling window, as well as the list of protected objects for real-time job instances; In the next scheduling window, real-time job instances are added to the candidate execution set, and non-real-time job instances are added to the candidate execution set based on the current actual execution status, admission restrictions, and execution budget restrictions. A protected execution window is formed in the next scheduling window, where job instances that meet the scheduling rules of the protected execution window are executed first. After the next scheduling window is completed, the next actual execution result of the real-time job instance and the next waiting time of the non-real-time job instance are collected in the next observation window. Based on the next actual execution result, the real-time constraint satisfaction is determined. Based on the determination result and the next waiting time, one of the following operations is performed: relax, remove or maintain the access restriction and execute the budget.
[0007] According to the task dynamic isolation scheduling method based on time constraint vector provided in the embodiments of this application, based on the judgment result and the next waiting time, one of the following operations is performed: relaxing, removing or maintaining the admission restriction and the execution budget, including: If the judgment result indicates that the real-time constraint is not continuously satisfied, maintain the admission restrictions and execution budget, and continue data acquisition in the next two observation windows; When the judgment result indicates that the real-time constraint is continuously satisfied but the preset release condition has not been met, it is determined whether the next waiting time of the non-real-time job instance has reached the preset upper bound. If the preset upper bound is reached, the admission restrictions and execution budget are relaxed. If the preset upper bound is not reached, the admission restrictions and execution budget are maintained, and data collection continues in the next two observation windows. When the judgment result indicates that the real-time constraint is continuously satisfied and the preset release condition is met, the admission restriction and execution budget are released.
[0008] According to the task dynamic isolation scheduling method based on time constraint vector provided in the embodiments of this application, the method further includes: Create a task group object that has job instances with sequential dependencies or shared resource relationships; Specifically, if any real-time job instance in the same task group has a windowing constraint that does not meet the classification, the real-time job instance and other real-time job instances in the task group that have sequential dependencies or shared resource relationships are added to the list of protected objects.
[0009] According to the task dynamic isolation scheduling method based on time constraint vector provided in the embodiments of this application, the time constraint vector includes: deadline, allowable jitter, expected start offset, expected completion offset, expected number of executions, and allowed execution duration; The actual execution results include: window identifier, actual start time, actual completion time, actual execution duration, actual number of executions, measured jitter, number of samples, number of jitter exceedances, job release time, and evidence status; Compare the time constraint vector with the current actual execution result, including: Obtain the comparison priority and comparison method of each constraint field in the preset time constraint vector; Based on the comparison priority and comparison method, the time constraint vector is compared with each field in the current actual execution result in turn.
[0010] According to the task dynamic isolation scheduling method based on time constraint vector provided in the embodiments of this application, after comparing the time constraint vector with the current actual execution result, it further includes: If any one or more fields are found to have an anomaly in the comparison results, determine that the windowing constraint is not satisfied with the classification, obtain the classification label corresponding to the field with the anomaly, and determine the obtained classification label as the classification result of the windowing constraint not being satisfied.
[0011] According to the task dynamic isolation scheduling method based on time constraint vectors provided in the embodiments of this application, the admission restrictions and execution budget of non-real-time job instances in the next scheduling window are determined based on the current classification result, including: If the current classification results include cases where deadline violations and / or consecutive windows are all subject to windowing constraints, generate a first admission limit and a first execution budget. The first admission restriction includes: reducing the number of non-real-time job instances entering the preset candidate execution set in the next scheduling window compared to the current scheduling window, delaying the entry of non-real-time job instances into the candidate execution set, and increasing the entry of non-real-time job instances into the candidate execution set, or any one or more of these restrictions. The first execution budget includes: reducing the maximum execution duration and maximum number of executions of pending non-real-time job instances to a preset threshold within the next scheduling window; If the current classification results are determined to be free of deadline violations and continuous windows, and include jitter exceeding limits, offset exceeding limits, insufficient execution times, and execution time exceeding limits, then a second admission limit and a second execution budget are generated. Wherein, the second admission restriction is a preset multiple of the adjustment range of the first admission restriction, and the second execution budget is a preset multiple of the adjustment range of the first execution budget, wherein the preset multiple is a number greater than zero and less than 1, wherein the adjustment range of the first admission restriction is a preset first admission restriction amount, and the adjustment range of the first execution budget is a preset first execution budget amount.
[0012] According to the task dynamic isolation scheduling method based on time constraint vector provided in the embodiments of this application, the method further includes: Based on the comparison results, it is determined that there are cases where the windowing constraints satisfy the classification, and the current classification result that satisfies the windowing constraints is obtained; Based on the current classification results, determine the admission restrictions and execution budget for non-real-time job instances in the next scheduling window, including: If the previous classification result indicates that the windowing constraint is not satisfied, the current classification result indicates that the windowing constraint is satisfied, and the release condition has not been met, then a third admission constraint and a third execution budget are generated. The third access restriction is a preset multiple of the adjustment range of the second access restriction, and the third execution budget is a preset multiple of the adjustment range of the second execution budget. The adjustment range of the second access restriction is a preset second access restriction amount, and the adjustment range of the second execution budget is a preset second execution budget amount.
[0013] According to the task dynamic isolation scheduling method based on time constraint vectors provided in the embodiments of this application, non-real-time job instances are included in the candidate execution set based on the current actual execution status, admission restrictions, and execution budget restrictions, including: Obtain the admission rules for the candidate execution set; If a non-real-time job instance is determined not to meet the admission rules based on the current actual execution status and admission restrictions, the non-real-time job instance will not be included in the candidate execution set, and the waiting time of the non-real-time job instance will be recorded. If a non-real-time job instance meets the admission rules based on the current actual execution status and admission restrictions, calculate the remaining execution budget for the next scheduling window. If the remaining execution budget is available, add the non-real-time job instance to the candidate execution set. If the remaining execution budget is unavailable, determine that the non-real-time job instance will not enter the candidate execution set, and record the waiting time of the non-real-time job instance.
[0014] This application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps of the task dynamic isolation scheduling method based on time constraint vectors as described above.
[0015] This application also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the task dynamic isolation scheduling method based on time constraint vectors as described above.
[0016] The task dynamic isolation scheduling method, device, and medium based on time constraint vectors provided in this application create real-time job instances and non-real-time job instances corresponding to the tasks to be processed; generate time constraint vectors corresponding to the real-time job instances; and collect the current actual execution results of the real-time job instances and the current actual execution status of the non-real-time job instances within the current observation window. The time constraint vectors describe the expected constraints of the real-time job instances within the observation window. The time constraint vectors are compared with the current actual execution results. Based on the comparison results, if a windowed constraint is not satisfied, a current classification result for the windowed constraint is obtained. This application does not rely on fixed settings but generates time constraint vectors based on real-time job instances, transforming the actual execution results within the observation window into a windowed constraint non-satisfaction classification, allowing the scheduling consequences of the previous observation window to affect the next scheduling window. Based on the current classification result, the admission restrictions and execution budgets of the non-real-time job instances in the next scheduling window, as well as the protection object list of the real-time job instances, are determined. The real-time job instances are added to the candidate execution list within the next scheduling window. This application sets up a candidate execution set based on the current actual execution status, admission restrictions, and execution budget. Within the next scheduling window, a protected execution window is formed. This application does not simply restrict all non-real-time job instances; instead, it derives corresponding admission restrictions and execution budgets based on classification results. This allows for tiered adjustment of candidate entry and window execution opportunities for non-real-time job instances, and utilizes the protected execution window to provide protected real-time job instances with more stable candidate entry and execution opportunities. After the next scheduling window execution ends, the next actual execution result of real-time job instances and the next waiting time of non-real-time job instances are collected within the next observation window. Based on the next actual execution result, the real-time constraint satisfaction status is determined. Based on the determination result and the next waiting time, one of the following operations is performed: relaxing, removing, or maintaining the admission restrictions and execution budget. Finally, a recovery strategy is used to reduce the possibility of long-term waiting for non-real-time job instances. At the scheduling window level, this application achieves dynamic balance and stable execution of real-time and non-real-time job instances by forming a dynamic isolation scheduling closed loop through candidate execution set admission, window execution budget, and subsequent review. Attached Figure Description
[0017] To more clearly illustrate the technical solutions in the embodiments of this application 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 some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0018] Figure 1 This is one of the flowcharts illustrating the task dynamic isolation scheduling method based on time constraint vectors provided in the embodiments of this application; Figure 2 This is the second flowchart of the task dynamic isolation scheduling method based on time constraint vector provided in the embodiments of this application; Figure 3 This is the third flowchart illustrating the task dynamic isolation scheduling method based on time constraint vectors provided in this application embodiment; Figure 4 This is the fourth flowchart of the task dynamic isolation scheduling method based on time constraint vector provided in the embodiments of this application; Figure 5 This is the fifth flowchart illustrating the task dynamic isolation scheduling method based on time constraint vectors provided in this application embodiment; Figure 6 This is the sixth flowchart of the task dynamic isolation scheduling method based on time constraint vector provided in the embodiments of this application; Figure 7 This is the seventh flowchart of the task dynamic isolation scheduling method based on time constraint vector provided in the embodiments of this application; Figure 8 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation
[0019] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.
[0020] This application provides a task dynamic isolation scheduling method based on time constraint vectors. This method can be applied to smart terminals and servers. This application uses the application of this method in a server as an example for illustration, and some other descriptions in the embodiments are for illustrative purposes only and are not intended to limit the scope of protection of this application, and will not be described in detail thereafter. The specific implementation of the method is as follows... Figure 1 As shown: Step 101: Obtain the tasks to be processed and create job instances corresponding to the tasks to be processed.
[0021] The job instances include: real-time job instances and non-real-time job instances.
[0022] Step 102: Generate the time constraint vector corresponding to the real-time job instance, and collect the current actual execution result of the real-time job instance and the current actual execution status of the non-real-time job instance within the current observation window.
[0023] The time constraint vector is used to describe the expected constraints of real-time job instances within the observation window.
[0024] The actual implementation status includes: waiting time, resource conflict status, budget consumption, and actual implementation progress.
[0025] Step 103: Compare the time constraint vector with the current actual execution result. Based on the comparison result, determine if there are cases where the windowing constraint is not satisfied with the classification, and obtain the current classification result where the windowing constraint is not satisfied.
[0026] Step 104: Based on the current classification results, determine the admission restrictions and execution budget for non-real-time job instances in the next scheduling window, as well as the list of protected objects for real-time job instances.
[0027] Step 105: In the next scheduling window, add real-time job instances to the candidate execution set, and non-real-time job instances are added to the candidate execution set based on the current actual execution status, admission restrictions, and execution budget restrictions, and a protected execution window is formed in the next scheduling window.
[0028] Among them, job instances that conform to the scheduling rules of the protected execution window are executed first.
[0029] Step 106: After the next scheduling window is completed, collect the next actual execution result of the real-time job instance and the next waiting time of the non-real-time job instance in the next observation window, and determine the real-time constraint satisfaction based on the next actual execution result. Based on the determination result and the next waiting time, perform one of the following operations: relax, remove or maintain the access restriction and execute the budget.
[0030] Specifically, the access restriction corresponds to a specific access restriction amount (i.e., the adjustment range of the transfer restriction), which can be preset according to the user's actual needs, and this application does not impose any restrictions; the execution budget corresponds to a specific execution budget amount (i.e., the adjustment range of the execution budget), which can be preset according to the user's actual needs, and this application does not impose any restrictions.
[0031] The task dynamic isolation scheduling method based on time constraint vectors provided in this application does not rely on fixed settings. Instead, it generates time constraint vectors based on real-time job instances, transforming the actual execution results in the observation window into windowed constraint non-compliance classifications, so that the scheduling consequences of the previous observation window can be applied to the next scheduling window. It does not simply restrict all non-real-time job instances, but obtains corresponding admission restrictions and execution budgets based on the classification results, allowing non-real-time job instance candidate entry and window execution opportunities to be adjusted hierarchically. It also utilizes protected execution windows to provide protected real-time job instances with more stable candidate entry and execution opportunities. Finally, it reduces the possibility of long-term waiting for non-real-time job instances through recovery strategies. At the scheduling window level, this application achieves dynamic balance and stable execution of real-time and non-real-time job instances by forming a dynamic isolation scheduling closed loop through candidate execution set admission, window execution budget, and subsequent review.
[0032] In one specific embodiment, the object relationships between job instances, time constraint vectors, observation window results (actual execution results), and constraint non-satisfaction classifications specifically include: Specifically, a job instance includes: task identifier, job instance identifier, task group identifier, trigger type, and expected release time.
[0033] When multiple real-time job instances belong to the same task group, the associated protection scope of the task group is determined based on the time constraints of each real-time job instance.
[0034] Among them, the task identifier is used to indicate the task to which the job instance belongs, the job instance identifier is used to distinguish different trigger instances of the same task, the task group identifier is used to indicate the association relationship between the job instance and other job instances, and the trigger type can be periodic trigger, event trigger, or mixed trigger.
[0035] Specifically, the time constraint vector includes: deadline, allowable jitter, expected start offset, expected completion offset, expected number of executions, and allowed execution duration.
[0036] Among them, the deadline is used to determine whether a deadline violation has occurred; the allowed jitter is used to determine whether jitter has exceeded the limit; the expected start offset and expected finish offset are used to determine whether the execution deviation has exceeded the limit (offset exceeds the limit); the expected number of executions is used to determine whether the actual number of executions is insufficient; and the allowed execution duration is used to determine whether the actual execution duration has exceeded the limit.
[0037] Specifically, the observation window results are used to describe the actual execution consequences.
[0038] The observation window results include: window identifier, actual start time, actual completion time, actual execution duration, actual execution count, measured jitter, sample size, number of deadline violations, number of jitter exceedances, and evidence status.
[0039] Among them, the number of deadline violations and the number of jitter exceedances belong to the observation result field or classification result field, and do not belong to the time constraint vector field.
[0040] Specifically, each constraint field in the time constraint vector corresponds to at least one actual execution field. If the time constraint vector contains a desired number of executions, the actual number of executions is collected in the observation window results; if the time constraint vector contains an allowed execution duration, the actual execution duration is collected in the observation window results. This avoids fields being listed without participating in subsequent judgments.
[0041] Specifically, any real-time job instance within the same task group that is classified as having a deadline violation, jitter exceeding limits, execution deviation exceeding limits, execution duration exceeding limits, insufficient execution attempts, or a non-consecutive window not being met is marked as a protected real-time job instance. If there are other real-time job instances in the task group that have a sequential dependency on the real-time job instance, these other real-time job instances are added to the associated protection list. For example, other real-time job instances in the task group that have a sequential dependency on the real-time job instance that has a time constraint violation are added to the associated protection list; for non-real-time job instances that have critical resource conflicts with the aforementioned real-time job instance, corresponding access restrictions or execution budget restrictions are imposed.
[0042] Specifically, the status of evidence includes sufficient evidence, missing fields, insufficient samples, and insufficient boundary evidence.
[0043] Specifically, when the evidence status is sufficient, the windowing constraint does not meet the classification process; when the evidence status is missing fields, insufficient samples, or insufficient boundary evidence, the current observation window is classified as insufficient evidence, and the previous window's admission restrictions and execution budget restrictions are maintained; if there is no previous window restriction, the default admission restrictions and default execution budget are adopted.
[0044] In one specific embodiment, the process for determining whether windowing constraints meet the classification requirements is implemented as follows: Figure 2 As shown: Step 201: Obtain the time constraint vector and the execution result of the observation window.
[0045] The execution result of the observation window is the actual execution result of the job instance within the observation window.
[0046] Step 202: Determine whether the time constraint field, execution result field, and number of valid samples meet the judgment requirements. If yes, proceed to step 203; otherwise, proceed to step 211.
[0047] A valid sample refers to an execution record of a real-time job instance that includes necessary fields such as actual start time, actual completion time, and actual execution duration. If the number of valid samples in the current observation window is less than a preset sample number threshold, or if any necessary field is missing, the system will mark the current observation window as insufficient evidence.
[0048] Step 203: Determine whether the actual completion time exceeds the deadline. If yes, proceed to step 212; otherwise, proceed to step 204.
[0049] Step 204: Determine whether the actual start offset or the actual finish offset exceeds the allowed offset. If yes, proceed to step 213; otherwise, proceed to step 205.
[0050] Step 205: Determine whether the measured jitter exceeds the allowable jitter. If yes, proceed to step 214; otherwise, proceed to step 206.
[0051] Step 206: Determine whether the actual execution time exceeds the allowed execution time. If yes, proceed to step 215; otherwise, proceed to step 207.
[0052] Step 207: Determine whether the actual number of executions is lower than the expected number of executions. If yes, proceed to step 216; otherwise, proceed to step 208.
[0053] Step 208: Determine if there are any cases where consecutive windows do not meet the requirements. If so, proceed to step 209; otherwise, proceed to step 210.
[0054] Each existing field is judged based on a preset priority.
[0055] Here, the number of consecutive abnormal windows is set to Kbad, and the number of windows that continuously satisfy the real-time constraint is set to K, where Kbad is an integer greater than or equal to 2. Kbad is used to determine if consecutive windows are not satisfied, and K is used to determine whether the release condition has been met.
[0056] Specifically, "continuous window non-compliance" means that in K consecutive observation windows, each observation window experiences at least one of the following: deadline violation, jitter exceeding limits, execution deviation exceeding limits, execution time exceeding limits, or insufficient execution count; or in K consecutive observation windows, the margin between the actual completion time and the deadline is less than a preset margin threshold. "Real-time constraint continuous satisfaction" means that in K consecutive observation windows, there are no deadline violations, jitter exceeding limits, execution deviation exceeding limits, execution time exceeding limits, or insufficient execution count, and the evidence status of each observation window is sufficient.
[0057] Step 209: The classification result is determined to be that the continuous window does not meet the requirements.
[0058] Step 210: Determine that the classification result satisfies the constraints.
[0059] Step 211: Determine the classification result as insufficient evidence.
[0060] Specifically, when the time constraint field is missing, the actual execution result is missing, or the number of samples in the observation window is lower than the sample number threshold, the observation window is classified as insufficient evidence.
[0061] In this case, insufficient evidence does not trigger stronger admission restrictions or lower execution budgets. Instead, the system maintains the previous window restrictions or adopts the default restrictions and waits for subsequent window data collection.
[0062] Step 212: Determine the classification result as a violation of the deadline.
[0063] Among them, deadline violation is a stronger category, which can trigger the first admission restriction and the first execution budget.
[0064] Step 213: Determine the classification result as an execution deviation exceeding the limit.
[0065] Among them, exceeding the execution deviation limit can trigger the second admission restriction and the second execution budget.
[0066] Step 214: Determine the classification result as excessive jitter.
[0067] Among them, jitter exceeding the limit can trigger the second admission restriction and the second execution budget.
[0068] Step 215: Determine the classification result as execution time exceeding the limit.
[0069] Among them, exceeding the execution time limit can be a subcategory of exceeding the execution deviation limit, or it can be an independent category that triggers the second access restriction and the second execution budget.
[0070] Step 216: Determine the classification result as insufficient execution attempts.
[0071] Insufficient execution counts can trigger the second admission restriction and the second execution budget; when multiple consecutive observation windows have insufficient execution counts, they can be merged into consecutive windows that do not meet the classification.
[0072] Specifically, Figure 2 The illustrated example specifies the priority of fields involved in the time constraint window. In reality, the fields involved may be less than or equal to... Figure 2 The fields shown in the example can be judged according to this priority.
[0073] Specifically, a single observation window can have multiple category labels simultaneously. For example, both deadline violations and jitter exceedances can occur within the same observation window. A primary category can be determined based on a preset priority, and other categories can be stored as auxiliary category fields.
[0074] In one specific embodiment, the specific implementation of determining the admission restrictions and execution presets for non-real-time job instances is as follows: Figure 3 As shown: The next scheduling window constraint is generated based on the classification results.
[0075] Step 301: Obtain the classification results.
[0076] Step 302: If the classification result indicates insufficient evidence, maintain the admission restrictions and execution budget restrictions of the previous observation window.
[0077] If there are no previous observation window admission restrictions and execution budget restrictions, then the default admission restrictions and default execution budget will be used.
[0078] Step 303: If the classification result is a violation of the deadline or a failure to meet the continuous window requirement, generate the first admission restriction and the first execution budget.
[0079] The first admission restriction is to reduce the number of non-real-time job instances allowed to enter the candidate execution set, postpone the entry of some non-real-time job instances into the candidate execution set, or raise the requirements for non-real-time job instances to enter the candidate execution set.
[0080] The first execution budget is to reduce the maximum execution time, maximum number of executions, or window budget count of non-real-time job instances within the scheduling window to the minimum guaranteed range.
[0081] Step 304: If the classification result is jitter exceeding the limit, execution deviation exceeding the limit, execution time exceeding the limit, or insufficient execution times, but the deadline is not met or the continuous window is not satisfied, generate a second admission limit and a second execution budget.
[0082] Among them, the second admission restriction is less restrictive than the first admission restriction, and the second execution budget allows more non-real-time execution opportunities than the first execution budget.
[0083] Step 305: If the classification result indicates that the constraint is satisfied but the release condition has not been met, a third admission restriction is generated.
[0084] The third access restriction is less restrictive than the second access restriction.
[0085] Under the third admission restriction, only non-real-time job instances that conflict with protected real-time job instances in terms of processor cores, shared cache, shared memory, bus access, or peripheral access are delayed.
[0086] Step 306: If the classification result shows that the constraints are satisfied and the release condition is met, remove the additional admission restrictions triggered by the failure of the windowed constraints to satisfy the classification, and restore the default execution budget.
[0087] Specifically, "removing budget constraints" refers to canceling additional budget constraints triggered by failure to meet category requirements due to windowing constraints and restoring the default execution budget, rather than canceling the scheduling system's regular scheduling rules. "Default admission constraints" refers to the rules that allow non-real-time job instances to enter the candidate execution set according to regular scheduling conditions when additional constraints are not triggered by failure to meet category requirements due to windowing constraints.
[0088] Relaxing access restrictions includes lowering the first access restriction to the second access restriction, lowering the second access restriction to the third access restriction, or removing the access restriction when the removal conditions are met. Increasing the execution budget includes raising the first execution budget to the second execution budget, or raising the second execution budget to the default execution budget. The increase in the execution budget shall not exceed the default execution budget; once the default execution budget is reached, no further increases will be made.
[0089] In one specific embodiment, the explanation of the candidate execution set generation, the restricted candidate execution set sorting, and the budget deduction mechanism can be found in [link to documentation]. Figure 4 : The system maintains a pool of job instances to be scheduled. This pool includes a set of real-time job instances and a set of non-real-time job instances.
[0090] Step 401: For the set of real-time job instances, determine the list of real-time protection objects based on the classification results of the previous observation window, and enter the candidate execution set of the next scheduling window into the real-time job instances belonging to the real-time protection object list.
[0091] In the next scheduling window, priority execution opportunity is obtained according to the scheduling rules of the protection execution window.
[0092] Step 402: For the set of non-real-time job instances, determine whether the non-real-time job instances meet the candidate set admission rules. If yes, proceed to step 403; otherwise, proceed to step 406.
[0093] Step 403: Determine if the window execution budget is available. If yes, proceed to step 404; otherwise, proceed to step 407.
[0094] Step 404: Add the non-real-time job instance to the restricted candidate execution set.
[0095] The candidate execution set includes a restricted candidate execution set, which is used to store non-real-time job instances.
[0096] Step 405: For non-real-time job instances in the restricted candidate execution set, determine the execution order according to waiting time, estimated execution duration, resource conflict status, and remaining execution budget.
[0097] Among them, non-real-time job instances with longer waiting times and expected execution times not exceeding the remaining budget can be prioritized for execution over non-real-time job instances with shorter waiting times or resource conflicts with protected real-time job instances.
[0098] Step 406: Delay adding to the candidate execution set and record the non-real-time waiting time.
[0099] The non-real-time waiting time is used for subsequent waiting for the upper bound to be reviewed.
[0100] Step 407: Wait for the next window to execute the budget counting opportunity or the next scheduling window, and record the non-real-time waiting time.
[0101] The non-real-time waiting time refers to the waiting time of non-real-time job instances.
[0102] In one specific embodiment, when a non-real-time job instance is added to the restricted candidate execution set, the estimated budget is deducted in advance, and the deduction result is adjusted according to the actual execution time, the actual number of executions, or the actual budget count after the non-real-time job instance completes execution.
[0103] Specifically, if the actual consumption is lower than the pre-deducted budget, the difference in budget will be refunded; if the actual consumption exceeds the pre-deducted budget, the excess will be recorded in the budget occupancy record of the next window. When the remaining execution budget is insufficient to support the smallest execution unit, subsequent non-real-time job instances will be kept in the waiting pool and their waiting time will be accumulated.
[0104] When the non-real-time waiting time reaches the upper bound of the waiting period but the real-time constraints have not been met consecutively, the current admission restrictions and execution budget restrictions are not lifted, but non-real-time job instances that meet the minimum guarantee conditions are allowed to enter the restricted candidate execution set within the first execution budget.
[0105] This allows us to continue protecting real-time job instances while preventing non-real-time job instances from waiting indefinitely.
[0106] In one specific embodiment, the formation, persistence, and exit of the real-time job instance protection execution window are described in [the following text is incomplete and requires further context]. Figure 5 : Step 501: Determine whether the protection conditions are triggered based on the classification results of the previous observation window. If yes, proceed to step 502; otherwise, proceed to step 506.
[0107] The protection condition is triggered when the classification result is not met.
[0108] Step 502: Determine the protected real-time job instance and determine the entry conditions, duration conditions, and exit conditions of the protection execution window.
[0109] Among them, the protected real-time job instance can be a real-time job instance that has violated the time constraint, or it can be another real-time job instance in its task group that has a sequential dependency relationship with it.
[0110] The entry conditions include any one or more of the following: the start point of the next scheduling window, the release time of the protected real-time job instance, and the time when the time before the deadline of the protected real-time job instance is less than the preset protection lead time.
[0111] The continuous conditions include any one or more of the following: the protected real-time job instance has not yet completed execution, the current protection window has not yet reached the set number of windows, and the real-time constraints have not been continuously satisfied.
[0112] The exit conditions for the protection execution window include the completion of the protected real-time job instance, the protection window reaching the set number of windows, or the real-time constraint being satisfied by K consecutive subsequent observation windows. When the non-real-time waiting time reaches the upper bound, only the minimum guarantee access or restriction relaxation judgment is triggered, and it is not directly used as the exit condition for the protection execution window.
[0113] Step 503: Within the protection execution window, restrict the candidate entry or window execution opportunities of non-real-time job instances based on at least one of the admission restrictions and execution budget of non-real-time job instances.
[0114] The aim is to give protected real-time job instances priority access to the candidate execution set or priority execution opportunities.
[0115] The protection execution window is a logical scheduling control state triggered by the classification result of the current observation window and applied to the next scheduling window. It is not a fixed physical time partition, nor does it represent the allocation of an exclusive processor core to a real-time job instance.
[0116] Step 504: Determine whether the protection execution window has met the exit conditions. If yes, proceed to step 505; otherwise, proceed to step 503.
[0117] Step 505: Perform the verification process for subsequent observation windows.
[0118] The review process includes: if the real-time constraints are continuously satisfied and the release conditions are met, then the additional admission restrictions are lifted or the default execution budget is restored; if the real-time constraints are continuously satisfied but the release conditions are not met, then the admission restrictions are relaxed or the execution budget is increased; if the real-time constraints are not continuously satisfied, then the current restrictions are maintained and the process proceeds to the next observation window.
[0119] Therefore, once a protection execution window is established, it has a clear exit mechanism and subsequent state transition.
[0120] Step 506: Based on the default admission restrictions and execution budget, execute real-time job instances and non-real-time job instances.
[0121] In one specific embodiment, the explanation of the subsequent window review and restriction adjustment mechanism can be found in [link to documentation]. Figure 6 : Step 601: After executing the next scheduling window, collect the next actual execution result of the next observation window, and update the real-time constraint satisfaction status and the waiting time of non-real-time job instances.
[0122] Step 602: Determine whether the real-time constraints are continuously satisfied. If yes, proceed to step 603; otherwise, proceed to step 604.
[0123] Step 603: Determine whether the non-real-time wait has reached the upper bound. If yes, proceed to step 605; otherwise, proceed to step 601.
[0124] Among them, the upper bound of non-real-time waiting is the maximum number of windows, the maximum waiting time, or the maximum number of delays that a non-real-time job instance is allowed to wait for.
[0125] Step 604: Maintain the current access restrictions and execute presets, and continue collecting data in another observation window.
[0126] Maintaining the current restrictions does not mean stopping processing; rather, it means that the current access restrictions, current execution budget, non-real-time wait time, and protected object status continue to serve as input for the next observation window.
[0127] Step 605: Lower the level of access restrictions or increase the execution budget.
[0128] Specifically, when real-time constraints have been continuously satisfied but have not yet reached the release condition, it is further determined whether the waiting time of the non-real-time job instance has reached the upper limit of waiting; if it has reached the upper limit of waiting, the admission restriction level is reduced or the execution budget is increased.
[0129] Specifically, if the number of observation windows in which real-time constraints are continuously satisfied reaches K, the release condition is considered met. Instead of relying on the upper bound of the non-real-time waiting period to determine whether to release protection, the additional admission restrictions triggered by the real-time constraint violation are directly released, and the default execution budget is restored. The waiting time of non-real-time job instances continues to be recorded, and non-real-time job instances that have reached the upper bound of the waiting period participate in scheduling according to the restored default admission rules.
[0130] In one specific embodiment, the scheduling constraint state is maintained within each scheduling window.
[0131] The scheduling constraint states include the default state, the first constraint state, the second constraint state, the third constraint state, the insufficient evidence hold state, and the release state.
[0132] When the classification is due to deadline violation or continuous window not being met, the system enters the first restricted state from the default state; when the classification is due to jitter exceeding the limit, execution deviation exceeding the limit, execution time exceeding the limit, or insufficient execution count, the system enters the second restricted state from the default state; when subsequent observation windows continuously meet real-time constraints but do not reach the release condition, the system decreases from the first restricted state to the second restricted state, or from the second restricted state to the third restricted state; when the release condition is reached, the system enters the release state and restores the default admission limit and default execution budget; when fields are missing, samples are insufficient, or evidence of mixed trigger boundaries is insufficient, the system enters the insufficient evidence hold state and maintains the admission limit and execution budget limit of the previous window.
[0133] Each scheduling constraint state corresponds to the candidate execution set admission action, execution budget action, and subsequent review action in the next scheduling window. Therefore, the states in this application are not simply recorded, but participate in the next stage of scheduling control.
[0134] In one specific embodiment, the explanation of the abnormal boundary handling, supplementary sampling, reclassification, and insufficient evidence state closing mechanism is provided below. Figure 7 : Step 701: Obtain job instance, time constraint vector, actual execution result and scheduling status data.
[0135] The scheduling status data includes current access restrictions, remaining execution budget, non-real-time waiting time, and the status of protected objects.
[0136] Step 702: Determine if the time constraint field is complete. If yes, proceed to step 703; otherwise, proceed to step 706.
[0137] Step 703: Determine whether the observation sample is sufficient. If yes, proceed to step 704; otherwise, proceed to step 707.
[0138] Step 704: Determine if the mixed trigger boundary is clear. If yes, proceed to step 705; otherwise, proceed to step 708.
[0139] Among them, the hybrid trigger boundary refers to whether the release time, expected start time, or completion time constraints can be determined when a job instance has a mixture of periodic triggering and event triggering characteristics.
[0140] Step 705: Proceed to the windowed constraint classification process.
[0141] Step 706: Record the reason for the missing time constraint field and set the category to insufficient evidence, while maintaining the previous window's admission restrictions and execution budget restrictions.
[0142] Step 707: Record insufficient window samples and set the classification to insufficient evidence, while maintaining the previous window's admission restrictions and execution budget restrictions.
[0143] Step 708: Record insufficient boundary evidence and set the category to insufficient evidence, while maintaining the previous window's access restrictions and execution budget restrictions.
[0144] Specifically, if there are no previous window access restrictions and execution budget restrictions, then the default access restrictions and default execution budget will be used. Abnormal boundary records are only used to explain the reasons for missing fields, insufficient samples, or insufficient evidence; they do not trigger external device status updates, nor do they serve as risk record trigger statuses.
[0145] The anomaly boundary record can include a record identifier, window identifier, job instance identifier, missing field, sample count, evidence status, and cause code. The anomaly boundary record is used to determine the supplementary fields and resampled objects for the next observation window. When the missing field is supplemented in the next observation window and the sample count reaches the threshold, the system closes the corresponding insufficient evidence status and re-enters the windowed constraint classification process.
[0146] When the mixed trigger boundary is determined in a subsequent observation window, the system will re-include the corresponding job instance in the windowed constraint non-compliance classification process and generate the next scheduling window constraint based on the reclassification result. If the mixed trigger boundary still cannot be determined, the system will remain in the insufficient evidence hold state and wait for resampling in the next window.
[0147] The following is a concrete example to illustrate this: The scheduling system runs on a quad-core processor with an observation window length of 100 milliseconds. The deadline for real-time job instance J1 is 20 milliseconds, with a jitter allowance of 2 milliseconds, an allowed execution time of 8 milliseconds, and an expected execution count of one. The default non-real-time window execution budget is 40 milliseconds, the first execution budget is 10 milliseconds, and the second execution budget is 20 milliseconds. The upper bound for non-real-time waiting is three observation windows. In this embodiment, Kbad is 2, and K is 3. The release condition is met when three consecutive observation windows satisfy the real-time constraints.
[0148] In observation window W1, J1's actual completion time was 24 milliseconds, the measured jitter was 3.5 milliseconds, and the actual execution count was one. The system classifies W1 as a deadline violation and jitter exceeding the limit. Since the constraint strength for deadline violation is higher than that for jitter exceeding the limit, the system generates the first admission constraint and the first execution budget in the next scheduling window W2, and forms a protective execution window.
[0149] Within observation window W2, the actual completion time of J1 is 18 milliseconds, and the measured jitter is 1.2 milliseconds. The real-time constraint is satisfied, but the release threshold has not yet been reached. The system does not immediately release the constraint, but instead maintains a light constraint or reduces the first admission constraint to the second admission constraint, and increases the first execution budget to the second execution budget.
[0150] Within observation window W3, the actual completion time of J1 was seventeen milliseconds, with a measured jitter of one millisecond, and the non-real-time waiting time reached three observation windows. The system determined that the real-time constraints had been continuously satisfied and the non-real-time waiting time had reached its upper bound. Therefore, it lowered the admission restrictions or increased the execution budget, allowing non-real-time job instances that met the minimum guarantee conditions to enter the restricted candidate execution set.
[0151] Within observation window W4, J1's actual completion time was 16 milliseconds, with a measured jitter of 0.8 milliseconds, continuously meeting the threshold for removal. The system removes the additional admission restrictions triggered by category W1 and restores the default execution budget.
[0152] This application does not simply impose long-term restrictions on non-real-time job instances after W1, but rather gradually reduces or removes the restrictions based on the subsequent recovery of real-time constraints and the upper bound of non-real-time wait.
[0153] For example, after classifying deadline violations and jitter exceedances in the first observation window, this application generates a first admission restriction and a first execution budget in the next scheduling window. This delays the entry of some non-real-time job instances into the candidate execution set and reduces the execution opportunities of non-real-time job instances within the scheduling window. If real-time constraints are satisfied again in subsequent windows, the system reduces the first admission restriction to a second admission restriction or increases the execution budget based on continuous satisfaction. When non-real-time waiting approaches or reaches the upper bound of waiting, non-real-time job instances with longer waiting times and meeting the admission conditions are allowed to enter the restricted candidate execution set. When continuous satisfaction meets the release condition, the additional admission restriction is removed and the default execution budget is restored.
[0154] This application does not simply address the issue of increasing the priority of real-time tasks, but rather it passes the results of unmet constraints in the observation window to the next scheduling window and dynamically balances the protection of real-time job instances and the waiting of non-real-time job instances through subsequent review.
[0155] This application achieves dynamic isolation scheduling of real-time and non-real-time tasks in multi-core processors through the overall process of "job instance modeling - time constraint vector generation - observation window acquisition - windowed classification - admission restrictions and budget restrictions generation - protection execution window formation - subsequent window review - relaxation, removal or maintenance of restrictions".
[0156] Figure 8 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 8 As shown, the electronic device may include a processor 801, a communications interface 802, a memory 803, and a communication bus 804. The processor 801, communications interface 802, and memory 803 communicate with each other via the communication bus 804. The processor 801 can call logical instructions from the memory 803 to execute a task dynamic isolation scheduling method based on time constraint vectors.
[0157] Furthermore, the logical instructions in the aforementioned memory 803 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0158] On the other hand, the present invention also provides a computer program product, the computer program product including a computer program stored on a non-transitory computer-readable storage medium, the computer program including program instructions, and when the program instructions are executed by a computer, the computer is able to execute the task dynamic isolation scheduling method based on time constraint vector provided by the above methods.
[0159] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the task dynamic isolation scheduling method based on time constraint vectors provided in the above embodiments.
[0160] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.
[0161] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.
[0162] Finally, it should be noted that the above descriptions are merely preferred embodiments of this application, and this application is not limited to the above embodiments. It is understood that other improvements and variations directly derived or conceived by those skilled in the art without departing from the spirit and concept of this application should be considered to be included within the protection scope of this application.
Claims
1. A task dynamic isolation scheduling method based on time constraint vectors, characterized in that, The method includes: Obtain the tasks to be processed and create job instances corresponding to the tasks to be processed. Job instances include: real-time job instances and non-real-time job instances. Generate a time constraint vector corresponding to the real-time job instance, and collect the current actual execution result of the real-time job instance in the current observation window, as well as the current actual execution status of the non-real-time job instance. The time constraint vector is used to describe the expected constraints of the real-time job instance in the observation window. Compare the time constraint vector with the current actual execution result. Based on the comparison result, determine if there are cases where the windowing constraint is not satisfied in the classification, and obtain the current classification result where the windowing constraint is not satisfied. Based on the current classification results, determine the admission restrictions and execution budget for non-real-time job instances in the next scheduling window, as well as the list of protected objects for real-time job instances; In the next scheduling window, real-time job instances are added to the candidate execution set, and non-real-time job instances are added to the candidate execution set based on the current actual execution status, admission restrictions, and execution budget restrictions. A protected execution window is formed in the next scheduling window, where job instances that meet the scheduling rules of the protected execution window are executed first. After the next scheduling window is completed, the next actual execution result of the real-time job instance and the next waiting time of the non-real-time job instance are collected in the next observation window. Based on the next actual execution result, the real-time constraint satisfaction is determined. Based on the determination result and the next waiting time, one of the following operations is performed: relax, remove or maintain the access restriction and execute the budget.
2. The task dynamic isolation scheduling method based on time constraint vectors according to claim 1, characterized in that, Based on the assessment result and the next waiting time, perform an action to relax, remove, or maintain access restrictions and execute one of the following actions from the budget: If the judgment result indicates that the real-time constraint is not continuously satisfied, maintain the admission restrictions and execution budget, and continue data acquisition in the next two observation windows; When the judgment result indicates that the real-time constraint is continuously satisfied but the preset release condition has not been met, it is determined whether the next waiting time of the non-real-time job instance has reached the preset upper bound. If the preset upper bound is reached, the admission restrictions and execution budget are relaxed. If the preset upper bound is not reached, the admission restrictions and execution budget are maintained, and data collection continues in the next two observation windows. When the judgment result indicates that the real-time constraint is continuously satisfied and the preset release condition is met, the admission restriction and execution budget are released.
3. The task dynamic isolation scheduling method based on time constraint vectors according to claim 1, characterized in that, The method further includes: Create a task group object that has job instances with sequential dependencies or shared resource relationships; Specifically, if any real-time job instance in the same task group has a windowing constraint that does not meet the classification, the real-time job instance and other real-time job instances in the task group that have sequential dependencies or shared resource relationships are added to the list of protected objects.
4. The task dynamic isolation scheduling method based on time constraint vectors according to any one of claims 1-3, characterized in that, The time constraint vector includes: deadline, allowable jitter, expected start offset, expected finish offset, expected number of executions, and allowed execution duration; The actual execution results include: window identifier, actual start time, actual completion time, actual execution duration, actual execution count, measured jitter, sample size, number of jitter exceedances, job release time, and evidence status; Compare the time constraint vector with the current actual execution result, including: Obtain the comparison priority and comparison method of each constraint field in the preset time constraint vector; Based on the comparison priority and comparison method, the time constraint vector is compared with each field in the current actual execution result in turn.
5. The task dynamic isolation scheduling method based on time constraint vectors according to claim 4, characterized in that, After comparing the time constraint vector with the current actual execution result, it also includes: If any one or more fields are found to have an anomaly in the comparison results, determine that the windowing constraint is not satisfied with the classification, obtain the classification label corresponding to the field with the anomaly, and determine the obtained classification label as the classification result of the windowing constraint not being satisfied.
6. The task dynamic isolation scheduling method based on time constraint vectors according to any one of claims 1-3, characterized in that, Based on the current classification results, determine the admission restrictions and execution budget for non-real-time job instances in the next scheduling window, including: If the current classification results include cases where deadline violations and / or consecutive windows are all subject to windowing constraints, generate a first admission limit and a first execution budget. The first admission restriction includes: reducing the number of non-real-time job instances entering the preset candidate execution set in the next scheduling window compared to the current scheduling window, delaying the entry of non-real-time job instances into the candidate execution set, and increasing the entry of non-real-time job instances into the candidate execution set, or any one or more of these restrictions. The first execution budget includes: reducing the maximum execution duration and maximum number of executions of pending non-real-time job instances to a preset threshold within the next scheduling window; If the current classification results are determined to be free of deadline violations and continuous windows, and include jitter exceeding limits, offset exceeding limits, insufficient execution times, and execution time exceeding limits, then a second admission limit and a second execution budget are generated. Wherein, the second admission restriction is a preset multiple of the adjustment range of the first admission restriction, and the second execution budget is a preset multiple of the adjustment range of the first execution budget, wherein the preset multiple is a number greater than zero and less than 1, wherein the adjustment range of the first admission restriction is a preset first admission restriction amount, and the adjustment range of the first execution budget is a preset first execution budget amount.
7. The task dynamic isolation scheduling method based on time constraint vectors according to claim 6, characterized in that, The method further includes: Based on the comparison results, it is determined that there are cases where the windowing constraints satisfy the classification, and the current classification result that satisfies the windowing constraints is obtained; Based on the current classification results, determine the admission restrictions and execution budget for non-real-time job instances in the next scheduling window, including: If the previous classification result indicates that the windowing constraint is not satisfied, the current classification result indicates that the windowing constraint is satisfied, and the release condition has not been met, then a third admission constraint and a third execution budget are generated. The third access restriction is a preset multiple of the adjustment range of the second access restriction, and the third execution budget is a preset multiple of the adjustment range of the second execution budget. The adjustment range of the second access restriction is a preset second access restriction amount, and the adjustment range of the second execution budget is a preset second execution budget amount.
8. The task dynamic isolation scheduling method based on time constraint vectors according to any one of claims 1-3, characterized in that, Non-real-time job instances are included in the candidate execution set based on current actual execution status, admission restrictions, and execution budget constraints, including: Obtain the admission rules for the candidate execution set; If a non-real-time job instance is determined not to meet the admission rules based on the current actual execution status and admission restrictions, the non-real-time job instance will not be included in the candidate execution set, and the waiting time of the non-real-time job instance will be recorded. If a non-real-time job instance meets the admission rules based on the current actual execution status and admission restrictions, calculate the remaining execution budget for the next scheduling window. If the remaining execution budget is available, add the non-real-time job instance to the candidate execution set. If the remaining execution budget is unavailable, determine that the non-real-time job instance will not enter the candidate execution set, and record the waiting time of the non-real-time job instance.
9. 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 steps of the task dynamic isolation scheduling method based on time constraint vector as described in any one of claims 1 to 8.
10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the task dynamic isolation scheduling method based on time constraint vector as described in any one of claims 1 to 8.
