Software-driven processing system based on artificial intelligence

Abstract

The invention discloses a software drive processing system based on artificial intelligence, and belongs to the technical field of artificial intelligence. The system specifically comprises the following steps: S1, acquiring various data required by a software drive processing system through a data acquisition module, preprocessing the data, and outputting a data packet; and S2, a threshold setting module sets a threshold required by the system, and a performance evaluation module analyzes data in the data packet to obtain a dynamic resource allocation index. And dynamically updating the second threshold value of the dynamic resource allocation index based on the mean value and the standard deviation of the dynamic resource allocation index of the sliding window. The method can adapt to different load scenes, and misjudgment caused by a fixed threshold value is reduced. For example, when the load fluctuation of the system is large, the dynamically updated threshold can reflect the actual situation of the system more accurately, an unreasonable resource allocation decision is avoided, frequent strategy switching caused by extreme fluctuation is prevented by setting lower limit constraints, and the stability of the system is enhanced.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a software-driven processing system based on artificial intelligence. Background Art

[0002] The software driver processing system bridges the gap between computer hardware, the operating system, and applications. It includes a series of drivers and related management mechanisms. These drivers are software codes written specifically for specific hardware devices. They enable communication and collaboration between the hardware devices and the rest of the computer, allowing the operating system and applications to correctly control and use the hardware devices.

[0003] Existing drive processing systems have the following flaws: static resource allocation, which relies on fixed rules to allocate CPU, memory and other resources and cannot adapt to dynamic load changes, resulting in idle or overloaded resources; single evaluation dimensions, with traditional methods relying solely on single parameters such as response time or utilization, and lacking a multi-dimensional comprehensive performance evaluation mechanism; rigid strategies, with optimization strategies based on manually preset thresholds and unable to adjust autonomously through real-time feedback, resulting in poor system robustness. Summary of the Invention

[0004] Purpose of the invention: The purpose of the present invention is to provide a software-driven processing system based on artificial intelligence; it can solve the problem of static resource allocation in existing drive systems.

[0005] Technical Solution: To solve the above technical problems, according to one aspect of the present invention, more specifically, a software-driven processing system based on artificial intelligence, the system specifically includes the following steps: S1, collect various data required by the software-driven processing system through the data acquisition module, pre-process the data and output the data package; S2. The threshold setting module sets the threshold required by the system, and the performance evaluation module analyzes the data in the data packet to obtain the dynamic resource allocation index; S3. The policy decision module generates optimization strategies marked as high-performance state, triggered balance state, and emergency state based on the comparison results between the dynamic resource allocation index and the preset dynamic resource allocation index first threshold and dynamic resource allocation index second threshold, and generates different optimization strategies from the preset policy library according to different states; S4. The driving execution module executes the resource allocation instruction according to the optimization strategy generated by the strategy generation module. When the execution of the resource allocation instruction fails, the fault tolerance mechanism is activated; S5. The data collection module collects various data of the optimized system, and dynamically updates the second threshold value of the dynamic resource allocation index by calculating the new threshold value based on the mean and standard deviation of the dynamic resource allocation index in the past 5 minutes.

[0006] Furthermore, in step S1, the data collected by the data collection module includes: CPU utilization, CPU computing power factor, memory usage, I / O waiting time and the number of lock contentions per second.

[0007] Furthermore, in step S1, Kalman filtering is used to eliminate instantaneous noise during data preprocessing, and the data is encapsulated in JSON format, including timestamp, various parameters and metadata.

[0008] Furthermore, in step S2, the performance evaluation module obtains CPU utilization, CPU computing power factor, memory usage, I / O waiting time, and the number of lock contentions per second, and analyzes and processes the above data to obtain the dynamic resource allocation index of the software-driven processing system:

[0009] Among them, FI is the dynamic resource allocation index, is the CPU utilization, is the CPU computing power factor, is the number of lock contentions per second, is the I / O wait time, The preset benchmark I / O wait time is 1ms. is the memory usage, is the total memory, The memory pressure baseline is the quotient of the preset memory usage and the total memory.

[0010] Furthermore, in step S3, the policy decision module generates the high performance state, the trigger balance state and the emergency optimization strategy respectively according to the comparison result between the preset threshold values of the dynamic resource allocation index. When the dynamic resource allocation index reaches the first threshold, it is marked as high performance state and the current resource allocation is maintained; when the dynamic resource allocation index reaches the second threshold When the dynamic resource allocation index reaches the first threshold, a triggering balance state strategy is generated to reduce the resources of low-priority tasks; when When the dynamic resource allocation index reaches the second threshold, an emergency optimization strategy is generated to release redundant cache and allocate an exclusive CPU core.

[0011] Furthermore, in step S3, the first threshold of the dynamic resource allocation index is greater than the second threshold of the dynamic resource allocation index.

[0012] Furthermore, in step S4, when the execution of the resource allocation instruction fails, the system rolls back to the last stable configuration and reports an error.

[0013] Furthermore, in step S5, the new threshold is obtained by calculating the mean and standard deviation of the dynamic resource allocation index in the past 5 minutes:

[0014] Among them, NY is the new threshold, The value of the dynamic resource allocation index for each minute in the past 5 minutes.

[0015] Furthermore, in step S5, when the second threshold value of the dynamic resource allocation index is dynamically updated by calculating the new threshold value based on the mean and standard deviation of the dynamic resource allocation index in the past 5 minutes, the dynamically updated second threshold value of the dynamic resource allocation index satisfies:

[0016] When the new threshold calculation result obtained by calculating the mean and standard deviation of the dynamic resource allocation index in the past 5 minutes does not exceed the preset dynamic resource allocation index second threshold lower limit, the dynamic resource allocation index second threshold is forcibly set to the dynamic resource allocation index second threshold lower limit.

[0017] Beneficial effects: The present invention has the following beneficial effects: 1. Real-time system performance evaluation using the Dynamic Resource Allocation Index (FI). This index combines multiple parameters, including CPU utilization, CPU power factor, memory usage, I / O wait time, and the number of lock contentions per second, avoiding the limitations of traditional methods that rely on a single metric. When calculating the FI, the numerator of the formula combines actual CPU utilization with the hardware power, eliminating the impact of performance differences between different CPUs. This makes the formula applicable across hardware platforms and improves heterogeneous system compatibility. A logarithmic function is used to transform the product into nonlinear growth, suppressing the impact of extreme high values on the results. Based on the comparison of the FI with the preset first and second thresholds of the Dynamic Resource Allocation Index, the system automatically adjusts the resource allocation strategy.

[0018] 2. When the driver execution module fails to execute a resource allocation instruction, the system automatically activates the fault-tolerance mechanism, rolls back to the last stable configuration, and reports the error. This mechanism avoids system crashes caused by execution failures, ensures the basic operation of the system, and even in the event of an error, quickly restores to a stable state, reducing the impact on system operation. The feedback optimization module continuously monitors execution results and optimizes the policy library through incremental learning. As the system operates, execution data and feedback information are continuously accumulated. The policy library can optimize and adjust based on this information, so that the system can continuously improve its robustness during long-term operation, better cope with various complex operating conditions and potential failures, and improve the overall performance and reliability of the system.

[0019] 3. The system dynamically updates the second threshold of the dynamic resource allocation index based on the mean and standard deviation of the dynamic resource allocation index of the sliding window. This approach can adapt to different load scenarios and reduce misjudgments caused by fixed thresholds. For example, when the system load fluctuates greatly, the dynamically updated threshold can more accurately reflect the actual situation of the system and avoid unreasonable resource allocation decisions. By setting a lower limit constraint, when the calculated new threshold does not exceed the lower limit of the second threshold of the dynamic resource allocation index, the second threshold of the dynamic resource allocation index is forcibly set to the lower limit of the second threshold of the dynamic resource allocation index. This can prevent extreme fluctuations from causing frequent policy switching and enhance the stability of the system. In actual operation, avoiding frequent switching of resource allocation policies helps maintain the smooth operation of the system and reduce the additional overhead and potential risks caused by policy changes. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a schematic diagram of the system principle. DETAILED DESCRIPTION

[0021] In order to make the technical solution of the present invention clearer, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0022] Example 1 The data acquisition module collects various data required by the software-driven processing system, including: CPU utilization, CPU computing power factor, memory usage, I / O waiting time and the number of lock contentions per second, to provide original input for subsequent analysis. The data is then filtered using Kalman filtering to eliminate transient noise and improve data reliability. The data is encapsulated in JSON format, containing a timestamp, various parameters and metadata, to facilitate efficient transmission and analysis between modules. Multi-dimensional data is collected to comprehensively evaluate the system status, solving the problem that the transmission software-driven processing system relies on static resource distribution and cannot dynamically adapt to load changes.

[0023] To solve the problem that the transmission method relies on a single indicator and cannot comprehensively reflect the system performance, the threshold setting module sets the threshold required by the system. The performance evaluation module analyzes the data in the data packet to obtain the dynamic resource allocation index. The performance evaluation module obtains CPU utilization, CPU computing power factor, memory usage, I / O waiting time and the number of lock contentions per second. The above data is analyzed and processed to obtain the dynamic resource allocation index of the software-driven processing system:

[0024] Among them, FI is the dynamic resource allocation index, is the CPU utilization, is the CPU computing power factor, is the number of lock contentions per second, is the I / O wait time, The preset benchmark I / O wait time is 1ms. is the memory usage, is the total memory, The memory pressure baseline is the quotient of the preset memory usage and the total memory.

[0025] Combining the actual CPU utilization with the hardware computing power, the impact of different CPU performance differences is eliminated, making the formula applicable across hardware platforms and improving the compatibility of heterogeneous systems. The numerator part uses a logarithmic function to convert the product into nonlinear growth, suppressing the impact of extreme high values on the result. 1+ is used to prevent the dynamic resource allocation index from being overly compressed by the denominator due to a too small numerator, avoiding the system from misjudging it as a low-load state. +1 ensures that the logarithmic function input is always certificated to prevent mathematical errors. Normalize the I / O wait time and use the square root to reduce the sensitivity to high I / O latency and avoid the dynamic resource allocation index being too low due to the denominator being too large. Standardize memory usage and balance the impact of memory pressure to generate a dynamic resource allocation index that comprehensively reflects the performance status of multiple dimensions such as CPU, memory, and I / O. The multi-dimensional comprehensive evaluation combines parameters such as CPU, memory, I / O, and lock contention to avoid the one-sidedness of traditional single indicators. Through hardware parameter and logarithmic standardization, it adapts to heterogeneous hardware environments. Logarithms and square roots suppress extreme values and improve system robustness.

[0026] The policy decision module generates the high performance state, trigger balance state and emergency optimization strategy respectively according to the comparison result between the dynamic resource allocation index and the preset dynamic resource allocation index first threshold and dynamic resource allocation index second threshold. When the dynamic resource allocation index reaches the first threshold, it is marked as high performance state and the current resource allocation is maintained; when the dynamic resource allocation index reaches the second threshold When the dynamic resource allocation index reaches the first threshold, a triggering balance state strategy is generated to reduce the resources of low-priority tasks; when When the dynamic resource allocation index reaches the second threshold, an emergency optimization strategy is generated to release redundant cache and allocate an exclusive CPU core. The first threshold of the dynamic resource allocation index is greater than the second threshold of the dynamic resource allocation index.

[0027] System performance is assessed in real time using the Dynamic Resource Allocation Index (DRI). This index combines multiple parameters, including CPU utilization, CPU power factor, memory usage, I / O wait time, and the number of lock contentions per second, avoiding the limitations of traditional methods that rely solely on a single metric. When calculating the FI, the numerator of the formula combines actual CPU utilization with the hardware power, eliminating the impact of performance differences between different CPUs. This makes the formula applicable across hardware platforms and improves compatibility across heterogeneous systems. A logarithmic function is used to transform the product into a nonlinear growth, mitigating the impact of extreme high values on the results. Based on the comparison of the FI with the preset first and second thresholds of the Dynamic Resource Allocation Index, the system automatically adjusts the resource allocation strategy, selecting predefined strategies based on different loads to ensure responsiveness and strategy diversity.

[0028] The driver execution module executes resource allocation instructions according to the optimization strategy generated by the strategy generation module. When the execution of the resource allocation instruction fails, the fault tolerance mechanism is activated. When the execution of the resource allocation instruction fails, it rolls back to the last stable configuration and reports the error. This mechanism avoids system crashes caused by execution failures and ensures the basic operation of the system. Even in the event of an error, it can quickly recover to a stable state, reducing the impact on system operation. The feedback optimization module continuously monitors the execution effect and optimizes the strategy library through incremental learning. As the system runs, execution data and feedback information are continuously accumulated. The strategy library can be optimized and adjusted based on this information, so that the system can continuously improve its robustness during long-term operation, better cope with various complex operating conditions and potential failures, and improve the overall performance and reliability of the system.

[0029] The data collection module collects various data of the optimized system and dynamically updates the second threshold of the dynamic resource allocation index by calculating the new threshold value based on the mean and standard deviation of the dynamic resource allocation index in the past 5 minutes:

[0030] Among them, NY is the new threshold, The value of the dynamic resource allocation index every minute in the past 5 minutes. Reflects the average load level in the past 5 minutes. Double the standard deviation captures the load fluctuation range, and dynamic adjustments based on historical data improve robustness to prevent fixed thresholds from being unable to adapt to load fluctuations. The lower limit of the threshold is dynamically adjusted to prevent short-term fluctuations from causing the threshold to be too low.

[0031] When dynamically updating the second threshold value of the dynamic resource allocation index by calculating the new threshold value based on the mean and standard deviation of the dynamic resource allocation index over the past five minutes, the dynamically updated second threshold value of the dynamic resource allocation index satisfies the following conditions:

[0032] When the new threshold calculation result obtained by calculating the mean and standard deviation of the dynamic resource allocation index in the past 5 minutes does not exceed the preset dynamic resource allocation index second threshold lower limit, the dynamic resource allocation index second threshold is forcibly set to the dynamic resource allocation index second threshold lower limit. The dynamic resource allocation index second threshold lower limit is constrained by preventing the dynamically updated dynamic resource allocation index second threshold from being too low, resulting in frequent measurement switching, thereby enhancing stability.

[0033] By dynamically adjusting the threshold using sliding window statistics, the fixed threshold is prevented from misjudging when the load fluctuates. The lower limit limits the impact of extreme fluctuations, ensuring smooth system policy switching. The system dynamically updates the second threshold of the dynamic resource allocation index based on the mean and standard deviation of the dynamic resource allocation index within the sliding window. This approach adapts to different load scenarios and reduces misjudgments caused by fixed thresholds. For example, when the system load fluctuates significantly, the dynamically updated threshold can more accurately reflect the actual system conditions and avoid unreasonable resource allocation decisions. By setting a lower limit constraint, if the calculated new threshold does not exceed the lower limit of the dynamic resource allocation index second threshold, the dynamic resource allocation index second threshold is forcibly set to the lower limit. This prevents frequent policy switching caused by extreme fluctuations and enhances system stability. In actual operation, avoiding frequent resource allocation policy switching helps maintain smooth system operation and reduces the additional overhead and potential risks associated with policy changes.

[0034] Example 2 When calculating the dynamic resource allocation index, =80%, =2, =10, =1.5, =4, =8, =0.4, then:

[0035] when When the dynamic resource allocation index reaches the first threshold, it is marked as a high-performance state and the current resource allocation is maintained; When the dynamic resource allocation index second threshold When the dynamic resource allocation index reaches the first threshold, a triggering balance state strategy is generated to reduce the resources of low-priority tasks; when When the dynamic resource allocation index reaches the second threshold, an emergency optimization strategy is generated to release redundant cache and allocate an exclusive CPU core.

[0036] Example 3 When calculating the new threshold, =3.5, =0.025, then:

[0037] At this time, 0.56 is compared with the lower limit value of the second threshold of the dynamic resource allocation index: when When the second threshold value of the dynamic resource allocation index reaches the lower limit, the second threshold value of the dynamic resource allocation index is dynamically adjusted to 0.56; when When the dynamic resource allocation index reaches the lower limit of the second threshold, the dynamic resource allocation index second threshold is dynamically adjusted to the lower limit of the second threshold.

[0038] The above-described embodiments merely illustrate several implementations of the present invention, and while their descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art would be able to make numerous variations and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be determined by the appended claims.

Claims

1. A software-driven processing system based on artificial intelligence, characterized in that: The system specifically includes the following steps: S1, collect various data required by the software-driven processing system through the data acquisition module, pre-process the data and output the data package; S2. The threshold setting module sets the threshold required by the system, and the performance evaluation module analyzes the data in the data packet to obtain the dynamic resource allocation index; S3. The policy decision module generates optimization strategies marked as high-performance state, triggered balance state, and emergency state based on the comparison results between the dynamic resource allocation index and the preset dynamic resource allocation index first threshold and dynamic resource allocation index second threshold, and generates different optimization strategies from the preset policy library according to different states; S4. The driving execution module executes the resource allocation instruction according to the optimization strategy generated by the strategy generation module. When the execution of the resource allocation instruction fails, the fault tolerance mechanism is activated; S5. The data collection module collects various data of the optimized system, and dynamically updates the second threshold value of the dynamic resource allocation index by calculating the new threshold value based on the mean and standard deviation of the dynamic resource allocation index in the past 5 minutes.

2. The artificial intelligence-based software-driven processing system according to claim 1, characterized in that: In step S1, the data collected by the data collection module includes: CPU utilization, CPU computing power factor, memory usage, I / O waiting time and the number of lock contentions per second.

3. The artificial intelligence-based software-driven processing system according to claim 1, characterized in that: In step S1, Kalman filtering is used to eliminate transient noise during data preprocessing, and the data is encapsulated in JSON format, including timestamp, various parameters and metadata.

4. The artificial intelligence-based software-driven processing system according to claim 1, wherein: In step S2, the performance evaluation module obtains CPU utilization, CPU computing power factor, memory usage, I / O waiting time and the number of lock contentions per second, and analyzes and processes the above data to obtain the dynamic resource allocation index of the software-driven processing system: ； Among them, FI is the dynamic resource allocation index, is the CPU utilization, is the CPU computing power factor, is the number of lock contentions per second, is the I / O wait time, The preset benchmark I / O wait time is 1ms. is the memory usage, is the total memory, The memory pressure baseline is the quotient of the preset memory usage and the total memory.

5. The artificial intelligence-based software-driven processing system according to claim 4, characterized in that: In step S3, the policy decision module generates the high performance state, the trigger balance state and the emergency optimization strategy respectively according to the comparison result between the preset threshold values of the dynamic resource allocation index. When the dynamic resource allocation index reaches the first threshold, it is marked as high performance state and the current resource allocation is maintained; when the dynamic resource allocation index reaches the second threshold When the dynamic resource allocation index reaches the first threshold, a triggering balance state strategy is generated to reduce the resources of low-priority tasks; when When the dynamic resource allocation index reaches the second threshold, an emergency optimization strategy is generated to release redundant cache and allocate an exclusive CPU core.

6. The artificial intelligence-based software-driven processing system according to claim 1, characterized in that: In step S3, the first threshold of the dynamic resource allocation index is greater than the second threshold of the dynamic resource allocation index.

7. The artificial intelligence-based software-driven processing system according to claim 1, characterized in that: In step S4, when the execution of the resource allocation instruction fails, the system rolls back to the last stable configuration and reports an error.

8. The artificial intelligence-based software-driven processing system according to claim 1, characterized in that: In step S5, the new threshold is obtained by calculating the mean and standard deviation of the dynamic resource allocation index in the past 5 minutes: ； Among them, NY is the new threshold, The value of the dynamic resource allocation index for each minute in the past 5 minutes.

9. The artificial intelligence-based software-driven processing system according to claim 1, characterized in that: In step S5, when dynamically updating the second threshold value of the dynamic resource allocation index by calculating the new threshold value based on the mean and standard deviation of the dynamic resource allocation index in the past 5 minutes, the dynamically updated second threshold value of the dynamic resource allocation index satisfies: ； When the new threshold calculation result obtained by calculating the mean and standard deviation of the dynamic resource allocation index in the past 5 minutes does not exceed the preset dynamic resource allocation index second threshold lower limit, the dynamic resource allocation index second threshold is forcibly set to the dynamic resource allocation index second threshold lower limit.

Images