Self-adaptive temperature control method and system based on SSD (Solid State Disk)

Abstract

The invention relates to the technical field of self-adaptive temperature control of an SSD, and discloses a self-adaptive temperature control method and system based on the SSD.The method comprises the steps that internal data and external parameters of the SSD are collected, trend extraction is carried out, and a load change trend is obtained; according to the load change trend, identifying a heat accumulation risk point, carrying out heat dissipation adjustment, collecting temperature data, and generating a temperature distribution diagram; performing regional division on the temperature distribution diagram, and generating a load distribution scheme in combination with preset load adjustment parameters; screening tasks in the load distribution scheme, and determining an initial task list; if the continuous high-intensity read-write operation exists in the initial task list, redistributing the continuous high-intensity read-write operation to other partitions of the SSD to generate an optimized task list; and recording temperature and performance fluctuation conditions in the operation process of the optimization task list, and generating a temperature stable state record. According to the method, the temperature can be controlled, and the SSD performance is improved.

Description

Technical Field

[0001] This invention relates to the field of adaptive temperature control technology for SSDs, and more particularly to an adaptive temperature control method and system based on SSDs. Background Technology

[0002] Currently, SSDs are a core component for data storage and processing, and their performance and stability directly affect the overall system efficiency and user experience. Especially under high load scenarios, SSD operating temperature often becomes a key factor affecting performance and lifespan. How to effectively control temperature in dynamically changing working environments to ensure stable device operation is a crucial research direction.

[0003] In one existing technology, a fixed temperature threshold is preset. When the device's internal smart sensor detects that the temperature has reached this threshold, a fan at a specific speed is activated. Simultaneously, a fixed load threshold is set; when the read / write load exceeds this threshold, some read / write tasks are restricted. However, because this existing technology only uses fixed parameters for heat dissipation once the fixed temperature threshold is reached, it struggles to handle sudden temperature spikes. Furthermore, when the load threshold is exceeded, it simply restricts tasks without optimizing resource allocation. Additionally, this existing technology relies on real-time data without analyzing the correlation between load and temperature or predicting future trends, leading to uncontrolled temperature and performance fluctuations. In summary, existing technologies suffer from performance fluctuations due to the inability to effectively control SSD temperature. Summary of the Invention

[0004] This invention provides an adaptive temperature control method and system based on SSD to solve the problem of performance fluctuations caused by uncontrollable temperature.

[0005] In a first aspect, to solve the above-mentioned technical problems, the present invention provides an adaptive temperature control method based on SSD, comprising:

[0006] An adaptive temperature control method based on SSD, characterized by comprising:

[0007] Data from internal SSD sensors and external environmental parameters are collected and aggregated to form an initial dataset. Trend extraction and correlation analysis are then performed on the initial dataset to obtain the load change trend.

[0008] Based on the external environmental parameters and the load change trend, potential heat accumulation risk points are identified, and risk assessment results are generated.

[0009] If the risk assessment results show that the potential heat accumulation risk point exceeds the preset risk threshold, then heat dissipation adjustment is performed, and temperature data after heat dissipation adjustment is collected to generate a temperature distribution map.

[0010] The temperature distribution map is divided into regions and the data is parsed. A load allocation scheme is generated by combining the preset load adjustment parameters.

[0011] Filter tasks that meet the preset task conditions in the load distribution scheme to determine the initial task list;

[0012] If there are consecutive high-intensity read / write operations in the initial task list, then the consecutive high-intensity read / write operations are reallocated to other partitions of the SSD to generate an optimized task list;

[0013] Record the temperature and performance fluctuations during the operation of the optimization task list, and generate a temperature stability record.

[0014] In one optional implementation, the process of collecting internal sensor data from the SSD and external environmental parameters, summarizing them to form an initial dataset, and performing trend extraction and correlation analysis on the initial dataset to obtain the load change trend includes:

[0015] The internal operating data and temperature data of the SSD are collected through the built-in sensor interface of the SSD, and the ambient temperature and humidity data are obtained by using external environmental monitoring equipment.

[0016] The internal operating data, temperature data, ambient temperature data, and humidity data are aggregated to form an initial dataset;

[0017] Based on a preset fluctuation threshold, the fluctuation of load intensity in the initial dataset is analyzed to determine the load fluctuation.

[0018] The correlation analysis is performed to determine the degree of correlation between the initial dataset and the load fluctuations, thereby identifying the load change trend.

[0019] In one optional implementation, the step of identifying potential heat accumulation risk points and generating risk assessment results based on the external environmental parameters and the load change trend includes:

[0020] Based on the external environment parameters and the load change trend, combined with historical load data, core features are selected and quantified to form a feature dataset;

[0021] The feature dataset is segmented into time segments according to a preset time range to obtain segmented feature datasets; if the load intensity in the segmented feature dataset exceeds a preset load intensity threshold and the ambient temperature data is higher than a preset ambient temperature threshold, it is marked as a high-risk interval.

[0022] The characteristic data of the high-risk area are input into a preset time prediction model to obtain the probability of temperature rise risk. If the probability of temperature rise risk exceeds a preset risk threshold, it is determined that there is a risk of heat accumulation.

[0023] By combining the heat accumulation risk and the feature dataset, risk triggering conditions are extracted, and warning indicators for the risk triggering conditions are marked to identify potential heat accumulation risk points and generate risk assessment results.

[0024] In one optional implementation, the step of dividing the temperature distribution map into regions and parsing the data, and generating a load allocation scheme by combining preset load adjustment parameters, includes:

[0025] The temperature distribution map is divided into regions according to a preset region division method to obtain temperature division regions;

[0026] Extract the temperature data of the temperature division region and the current load data, and combine the temperature data of the temperature division region and the current load data to obtain the core operating parameters;

[0027] Analyze the relationship between the core operating parameters and the preset load adjustment parameters to obtain the load adjustment parameter configuration;

[0028] Based on the configured load adjustment parameters, the preset load allocation rules are dynamically updated to obtain a load allocation scheme.

[0029] In one optional implementation, the step of filtering tasks that meet preset task conditions in the load allocation scheme to determine an initial task list includes:

[0030] Extract the tasks that need to be prioritized from the load allocation scheme and determine the target task list;

[0031] Obtain the real-time hardware status indicators of the temperature division region;

[0032] The real-time hardware status indicators are compared with the target task list. If the real-time hardware status indicators meet the preset task conditions of the target task list, then the tasks in the target task list are marked as transferable tasks.

[0033] The execution priorities of the transferable tasks are sorted out to determine the initial task list.

[0034] In one optional implementation, if there are consecutive high-intensity read / write operations in the initial task list, the consecutive high-intensity read / write operations are reallocated to other partitions of the SSD to generate an optimized task list, including:

[0035] The read / write frequency and task type are obtained from the initial task list, and key features are extracted. If the extraction result exceeds the preset intensity threshold, a preliminary task set is obtained.

[0036] Obtain the resource utilization rate of each SSD partition, and filter out the partitions whose resource utilization rate is lower than a preset utilization rate threshold as target partitions;

[0037] According to the preset task allocation rules, the initial task set is redistributed to the target partition to generate an optimized task list.

[0038] In one optional implementation, recording the temperature and performance fluctuations during the operation of the optimized task list and generating a temperature stability record includes:

[0039] Obtain temperature feedback data and performance indicators during the operation from the list of optimization tasks;

[0040] The timestamps of the temperature feedback data and the performance indicators are classified and organized, and features are extracted to obtain temperature fluctuation records and performance indicator fluctuation records.

[0041] By associating the temperature fluctuation record and the performance index fluctuation record, a temperature stability record is generated.

[0042] Secondly, the present invention provides an SSD-based adaptive temperature control system, comprising:

[0043] The data acquisition and trend analysis module is used to collect internal sensor data and external environmental parameters of the SSD, summarize them to form an initial dataset, and perform trend extraction and correlation analysis on the initial dataset to obtain the load change trend.

[0044] The risk identification and assessment module is used to identify potential heat accumulation risk points based on the external environmental parameters and the load change trend, and generate risk assessment results;

[0045] The temperature data processing module is used to adjust the heat dissipation if the risk assessment result shows that the potential heat accumulation risk point exceeds a preset risk threshold, and to collect the temperature data after the heat dissipation adjustment and generate a temperature distribution map.

[0046] The scheme generation module divides the temperature distribution map into regions and parses the data, and generates a load allocation scheme by combining preset load adjustment parameters.

[0047] The task list determination module is used to filter tasks that meet preset task conditions in the load allocation scheme and determine the initial task list.

[0048] The task optimization module, if there are consecutive high-intensity read and write operations in the initial task list, will reallocate the consecutive high-intensity read and write operations to other partitions of the SSD and generate an optimized task list;

[0049] The recording module is used to record the temperature and performance fluctuations during the operation of the optimization task list and generate a temperature stability record.

[0050] Thirdly, the present invention also provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement an SSD-based adaptive temperature control method as described in any one of the above.

[0051] Fourthly, the present invention also provides a computer-readable storage medium comprising a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform an SSD-based adaptive temperature control method as described above.

[0052] Compared with the prior art, the present invention has the following beneficial effects:

[0053] (1) This invention collects data from inside and outside the solid-state drive in real time, extracts load trends by combining time series analysis, predicts the risk of temperature rise, locks in the heat risk in advance, reduces performance problems caused by sudden overheating, improves the stability of equipment operation, and solves the problem of performance fluctuation caused by temperature runaway in the prior art.

[0054] (2) After predicting the heat risk, the present invention first activates the auxiliary heat dissipation in a specific area, then analyzes the relationship between overheating and load through the temperature distribution map, optimizes the load allocation and transfers high-priority tasks to low-temperature areas, disperses the heat source, achieves temperature balance and efficient load allocation, and improves read and write efficiency and resource utilization.

[0055] (3) The present invention uses a dynamic scheduling mechanism to allocate tasks to different partitions for continuous high-intensity read and write operations, monitors temperature feedback throughout the process to adjust the execution sequence, achieves dynamic task adaptation through multi-stage scheduling, and forms closed-loop control through real-time temperature feedback, thereby solving the problem of heat accumulation and mitigating performance fluctuations. Attached Figure Description

[0056] Figure 1 This is a schematic diagram of an adaptive temperature control method based on SSD provided in the first embodiment of the present invention;

[0057] Figure 2 This is a schematic diagram of an SSD-based adaptive temperature control system provided in the second embodiment of the present invention. Detailed Implementation

[0058] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0059] Reference Figure 1 The first embodiment of the present invention provides an adaptive temperature control method based on SSD, comprising the following steps:

[0060] S11. Collect data from internal sensors of the SSD and external environmental parameters, summarize them to form an initial dataset, and perform trend extraction and correlation analysis on the initial dataset to obtain the load change trend.

[0061] S12, Based on the external environmental parameters and the load change trend, identify potential heat accumulation risk points and generate risk assessment results;

[0062] S13, if the risk assessment result shows that the potential heat accumulation risk point exceeds the preset risk threshold, then heat dissipation adjustment is performed, and temperature data after heat dissipation adjustment is collected to generate a temperature distribution map.

[0063] S14, Divide the temperature distribution map into regions and parse the data, and generate a load allocation scheme by combining the preset load adjustment parameters;

[0064] S15, filter the tasks that meet the preset task conditions in the load distribution scheme, and determine the initial task list;

[0065] S16, If there are continuous high-intensity read and write operations in the initial task list, the continuous high-intensity read and write operations are redistributed to other partitions of the SSD to generate an optimized task list;

[0066] S17, record the temperature and performance fluctuations during the operation of the optimization task list, and generate a temperature stability record.

[0067] In step S11, the collection of internal sensor data and external environmental parameters of the SSD is summarized to form an initial dataset. Trend extraction and correlation analysis are then performed on the initial dataset to obtain the load change trend, including:

[0068] The internal operating data and temperature data of the SSD are collected through the built-in sensor interface of the SSD, and the ambient temperature and humidity data are obtained by using external environmental monitoring equipment.

[0069] The internal operating data, temperature data, ambient temperature data, and humidity data are aggregated to form an initial dataset;

[0070] Based on a preset fluctuation threshold, the fluctuation of load intensity in the initial dataset is analyzed to determine the load fluctuation.

[0071] The correlation analysis is performed to determine the degree of correlation between the initial dataset and the load fluctuations, thereby identifying the load change trend.

[0072] Specifically, the SSD's SMART interface is first set to collect data once per second to obtain internal operating data and internal temperature data in real time. For example, at 09:00:00 on May 21, 2024, internal operating data with a read / write speed of 90MB / s, a task queue length of 15, and a partition resource utilization rate of 70% is collected, as well as internal temperature data of 46°C for the flash memory area and 43°C for the controller. At the same time, using the temperature and humidity sensors deployed in the data center, external environmental data of 27°C and 60%RH at that moment are collected synchronously. This data collection operation is continuously performed for 24 hours.

[0073] In one possible implementation, all data collected within 24 hours is integrated based on timestamps, and the data format is unified. The read / write speed unit is fixed in MB / s, the temperature unit is unified in ℃, and the humidity unit is retained as %RH. For example, the "read / write 90MB / s, flash memory 46℃, controller 43℃, ambient temperature 27℃, humidity 60%RH" corresponding to "2024-05-21 09:00:00" is integrated into a complete data record, ultimately forming an initial dataset in CSV format containing 86,400 data records.

[0074] For example, dividing the initial 24-hour dataset into 24 analysis periods, with each period lasting one hour, direct weighted calculations would be affected by the numerical range of each indicator due to differences in the dimensions of read / write speed, task queue length, and partition resource utilization. Therefore, linear normalization is first used to eliminate the influence of these dimensions. The extreme values ​​of each indicator in the initial 24-hour dataset are first calculated: minimum read / write speed of 10MB / s and maximum of 100MB / s; minimum task queue length of 2 and maximum of 20; and minimum partition resource utilization of 10% and maximum of 90%. Then, the normalization formula is applied... norm= Normalization is performed, where x is the original value of the indicator and xnorm is the normalized value, ranging from 0 to 1.

[0075] The indicator values ​​for each time period are converted, and then the comprehensive load intensity value is calculated using read / write speed (weight 40%), task queue length (weight 40%), and partition resource utilization (weight 20%) as indicators. For example, during the period from 09:00 to 10:00, the average read / write speed is 85MB / s, the average task queue length is 14, and the average partition resource utilization is 68%. Their normalized values ​​are (85-10) / (100-10)=0.833, (14-2) / (20-2)=0.667, and (68-10) / (90-10)=0.725, respectively. The comprehensive value is 0.833×0.4+0.667×0.4+0.725×0.2=0. .745; After calculating the normalized composite value for 24 time periods, it was found that the composite value increased from 0.467 to 0.689 during the period from 08:00 to 11:00, with a load fluctuation of 0.222. During the period from 22:00 to 07:00 the next day, the composite value stabilized between 0.200 and 0.244, with a load fluctuation of 0.044. Considering the characteristics of SSD hardware and actual application scenarios, and combining the normalized composite value range (0-1), the load fluctuation threshold was set to ±0.15. The fluctuation of 0.222 during the morning peak period exceeded the preset threshold and was judged as a significant characteristic of load fluctuation. The fluctuation of 0.044 during the night period did not exceed the threshold and was judged as a characteristic of stable load.

[0076] It should be noted that the load fluctuation threshold setting is based on the chip's tolerance for sudden temperature rise from a hardware perspective. Combining the correlation between comprehensive value fluctuation and temperature runaway in historical data, in this embodiment, when the load fluctuation exceeds ±0.15, the risk of the core temperature rising by more than 0.5℃ per minute increases, and when the fluctuation amplitude is ≥0.15, the probability of temperature exceeding the standard exceeds 80%. Therefore, the threshold is set to ±0.15.

[0077] It is worth noting that read and write speeds directly reflect the real-time data transfer volume of the SSD and are the most intuitive manifestation of load intensity. The task queue length represents the backlog of tasks to be processed. Both together determine the real-time workload of the SSD and have the strongest core driving effect on the load, so they are both given a high weight of 40%. On the other hand, partition resource utilization reflects the long-term usage status of storage resources. Although it will affect storage efficiency, its direct impact on real-time load intensity is weaker than the first two indicators, so it is given a lower weight of 20%.

[0078] For example, a scatter plot is drawn between the overall load intensity and the flash memory region temperature. A clear positive correlation is observed between the two. Then, the Pearson correlation coefficient is calculated using the following formula: , where r represents the Pearson correlation coefficient, xi represents the comprehensive load intensity value of the i-th time period, and yi represents the flash memory region temperature of the i-th time period. This represents the average value of the overall load intensity. This represents the average temperature of the flash memory region. Substituting this into the calculation yields a Pearson correlation coefficient of 0.91.

[0079] In this invention, a linear regression equation Y=a+bX is established based on two sets of data, where a is the intercept and b is the slope, representing the average temperature increase for every 1 unit increase in load. Then, the linear regression formula is used... Calculating the slope and substituting it into the example, we get b=0.32, meaning that for every 1 increase in load, the temperature rises by an average of 0.32℃. This establishes a quantitative correlation: "For every 1 increase in the overall load intensity, the flash memory temperature rises by an average of 0.32℃." Combining load and temperature data from 24 time periods, we found that from 08:00 to 11:00 each day, as the load increases, the flash memory temperature gradually rises from 41℃ to 49℃. From 11:00 to 18:00, the load remains in the 55-61℃ range, and the temperature stabilizes at 47-49℃. After 18:00, the load gradually decreases, and the temperature drops back to 41-43℃. Ultimately, we can extract a cyclical trend in the SSD's load and temperature during the workday: a continuous increase in the morning, a high and stable level at midday, and a gradual decrease in the evening. This provides data support for subsequent temperature prediction and heat dissipation control.

[0080] In step S12, identifying potential heat accumulation risk points based on the external environmental parameters and the load change trend, and generating risk assessment results, includes:

[0081] Based on the external environment parameters and the load change trend, combined with historical load data, core features are selected and quantified to form a feature dataset;

[0082] The feature dataset is segmented into time segments according to a preset time range to obtain segmented feature datasets; if the load intensity in the segmented feature dataset exceeds a preset load intensity threshold and the ambient temperature data is higher than a preset ambient temperature threshold, it is marked as a high-risk interval.

[0083] The characteristic data of the high-risk area are input into a preset time prediction model to obtain the probability of temperature rise risk. If the probability of temperature rise risk exceeds a preset risk threshold, it is determined that there is a risk of heat accumulation.

[0084] By combining the heat accumulation risk and the feature dataset, risk triggering conditions are extracted, and warning indicators for the risk triggering conditions are marked to identify potential heat accumulation risk points and generate risk assessment results.

[0085] In one possible implementation, the SSD's historical load data from the past 30 days is integrated, such as peak morning load of 42-61, nighttime load of 18-22, real-time collected external environmental parameters, such as temperature of 25-29℃ and humidity of 55-65%RH, as well as the analyzed load change trend of continuous increase in the morning, high and stable at noon, and gradual decrease in the evening. From this, load intensity, ambient temperature, and load fluctuation amplitude are selected as core features, and each feature is quantified into specific values, such as a certain period recorded as load of 58, ambient temperature of 28℃, and fluctuation amplitude of 12, forming a feature dataset.

[0086] For example, the 24-hour feature dataset is segmented according to a preset 1-hour time range, resulting in 24 segmented feature datasets. Each segment is then examined. If the load intensity within a segment does not exceed 70 and the ambient temperature does not exceed 28°C, it is not marked; if the load intensity within a segment exceeds 70 or the ambient temperature exceeds 28°C, that time period is marked as a high-risk zone.

[0087] It should be noted that, from the perspective of SSD hardware operation logic, the preset duration of 1 hour means that when the load intensity exceeds 70 or the ambient temperature is higher than 28°C, the core components will take 30-50 minutes to exceed the safety threshold. In addition, the cooling system takes 20-30 minutes from startup to temperature stabilization. The 1-hour duration can fully cover the entire cycle, avoiding the mistaken judgment of no risk when the heat peak is not captured due to too short a segment, and also preventing the risk handling from being delayed due to too long a segment.

[0088] It is worth noting that, based on the SSD hardware parameters, its rated continuous operating load limit is 90, with 20% redundancy reserved to cope with sudden read and write peaks. Therefore, the load threshold for daily monitoring is set to 90×(1-20%)≈70. Laboratory data shows that the chip's heat dissipation efficiency decreases by 30% after the temperature exceeds 28℃, which can easily lead to heat accumulation. Therefore, the ambient temperature threshold is set to 28℃.

[0089] It should be noted that the preset time prediction model is constructed using a support vector machine. The structure of the support vector machine is as follows: taking the 3D risk feature data of the data center rack area, and taking the load intensity, ambient temperature, and load fluctuation amplitude as inputs, a binary classification SVM model is constructed to establish the mapping relationship between the 3D feature data and the risk judgment results, and output the probability of temperature rise risk.

[0090] In this invention, the training process of the support vector machine is as follows: First, synchronous samples of 3D features and risk results within a 2-year period are extracted from the historical operation and maintenance database of the data center. The 3D feature data needs to undergo outlier removal and standardization processing. For example, the load intensity is taken as a normalized dimensionless value (range 0-1), the ambient temperature is taken as the measured value of 20-50℃, and the load fluctuation range is 0-50. It is transformed into data in the range [-1,1] through Z-score standardization to ensure that the sample timestamp deviation is ≤1 minute. The training set and the validation set are divided into a 7:3 ratio for model training. The 3D feature data in the training set is used as input, and the temperature rise risk probability is used as the output label. The nonlinear relationship between the features and the temperature rise risk probability is processed through the RBF kernel function. For example, the temperature rise rate is nonlinearly increased after the load exceeds 70℃. To quickly identify the relationship between features, a binary hyperplane is constructed to learn the mapping relationship between different feature combinations and different probabilities of temperature rise risk. The Platt scaling method is used to convert the decision function values ​​into probability values, obtaining the temperature rise risk probability as the output label. Validation and optimization are then performed. The model's classification accuracy is verified using a validation set (e.g., temperature rise risk probability prediction accuracy ≥ 90%, recall ≥ 88%). If the requirements are not met, samples from extreme operating conditions, such as sudden load increases or high-risk samples caused by cooling system failures, are added, or the model's penalty parameter C (to control overfitting) and kernel function parameter γ (to optimize high-dimensional mapping effects) are adjusted. Iterative training continues until the model can stably output the association result of 3D feature data and temperature rise risk probability, and the error in judging the temperature rise risk probability of new scene samples is ≤ 10%. If the probability calculated by the support vector machine model is 86%, exceeding the preset 80% probability threshold, then it is determined that there is a risk of heat accumulation in that interval.

[0091] It should be noted that setting the probability threshold to 80% has several advantages. From a model validation perspective, this threshold ensures that the model maintains a recall rate of over 88% for risky samples and keeps the false positive rate for risk-free samples below 10%. This avoids both the potential risk of missed detections due to an excessively high threshold and the frequent false positives caused by an excessively low threshold. From a hardware safety perspective, the 80% threshold means that when the model determines there is a risk, there is an 80% probability that the temperature will exceed the limit within one hour. This allows for early intervention and sufficient time for response, preventing over-warning.

[0092] It is worth noting that the temperature risk threshold is set based on the SSD hardware manual, which clearly states that the critical tolerance temperature of the chip is 60℃. If this temperature is exceeded, the device will be forced to shut down. In order to reserve safety redundancy, this value should be avoided. Furthermore, according to historical data, when the temperature of the SSD core components exceeds 50℃, the SSD bad block generation rate soars from 0.01% to 0.15%. Therefore, the temperature risk threshold is set at 50℃.

[0093] For example, by combining the determined heat accumulation risk and feature dataset, the risk triggering conditions of load intensity ≥70 and ambient temperature ≥28℃ are analyzed and extracted. These conditions are marked with a red warning sign, and high-risk intervals such as 12:00-13:00 and 14:00-15:00 are identified as potential heat accumulation risk points. Finally, a risk assessment result containing high-risk intervals, triggering conditions, warning levels, and the probability of temperature rise is generated.

[0094] In step S13, if the risk assessment result shows that the potential heat accumulation risk point exceeds the preset risk threshold, then heat dissipation adjustment is performed, and temperature data after heat dissipation adjustment is collected to generate a temperature distribution map.

[0095] In one possible implementation, the risk assessment results are reviewed. If the results show that there are four potential heat accumulation risk points during the day: 11:00-12:00, 12:00-13:00, 14:00-15:00, and 15:00-16:00, exceeding the preset threshold for three time periods, then heat dissipation adjustments are initiated. For the rack area where the risk point is located, the fan speed is increased from the normal 60% to 85%, and the preset auxiliary cooling fans in the rack are turned on. If the risk is concentrated in a single area, such as all three SSDs in a rack triggering the risk, the angle of the airflow guide plate in the rack is further dynamically adjusted.

[0096] It should be noted that three time-period thresholds are preset, which are linked to load change trend analysis and risk assessment results. Statistically, under normal operating conditions, there are usually 1-2 high-risk intervals for SSDs each day. The three time-period thresholds cover the normal peak values ​​and also reserve one time-period redundancy for sudden loads, avoiding accidental overheating caused by occasional risks. Furthermore, based on the matching characteristics between the cooling system and the SSD load, for every 1 unit increase in load, the average temperature of the SSD core components rises by 0.32℃. When the cooling system continuously handles a high-risk interval of 3 hours, it can stabilize the temperature in that area at 26-28℃ through dynamic speed adjustment, and the power consumption does not exceed 80% of the rated power. If there are more than three high-risk intervals, after the cooling system runs at full speed continuously for more than 3 hours, the airflow decreases by ≥15% due to bearing wear. At this point, the temperature of the SSD core components approaches the risk threshold.

[0097] For example, the SSD's SMART interface is synchronized with the data center temperature sensor to a unified system time, accurate to the second; then, the temperature data after heat dissipation adjustment is collected synchronously every 5 minutes through the SSD's SMART interface and the data center temperature sensor. For example, at 11:30, the target SSD flash memory temperature is collected synchronously from 49°C to 46°C, and the rack ambient temperature is collected from 29°C to 27°C.

[0098] The collected SSD internal temperature and rack ambient temperature data at different time points are imported into a visualization program built on Python Matplotlib or an equivalent open-source library. This program supports time series temperature data plotting. A temperature distribution map is generated by the dimensions of time axis (horizontal axis), temperature value (vertical axis), and high-risk interval labeling. The temperature risk threshold of 50℃ is marked with a red dashed line, and the temperature change trend after heat dissipation is presented with a blue curve. At the same time, the corresponding heat dissipation adjustment measures are associated with and noted below the distribution map, such as increasing the fan speed from the normal 60% to 85% at 11:00.

[0099] In step S14, the process of dividing the temperature distribution map into regions and parsing the data, and generating a load allocation scheme by combining preset load adjustment parameters, includes:

[0100] The temperature distribution map is divided into regions according to a preset region division method to obtain temperature division regions;

[0101] Extract the temperature data of the temperature division region and the current load data, and combine the temperature data of the temperature division region and the current load data to obtain the core operating parameters;

[0102] Analyze the relationship between the core operating parameters and the preset load adjustment parameters to obtain the load adjustment parameter configuration;

[0103] Based on the configured load adjustment parameters, the preset load allocation rules are dynamically updated to obtain a load allocation scheme.

[0104] In one possible implementation, taking data center area A as an example, there are four rack areas: A1, A2, A3, and A4. A1 and A4, where the temperature is consistently below 26℃, are classified as low-load adaptation areas; A2, where the temperature is between 26-28℃, is classified as a medium-load adaptation area; and A3, where the temperature once exceeded 28℃ but has since cooled down, is classified as a limited-load area, resulting in four temperature-divided areas.

[0105] Extract the core operating parameters of each region: A1 current temperature 25℃, load 49; A2 current temperature 27℃, load 65; A3 current temperature 26.5℃, load 68; A4 current temperature 25.5℃, load 52.

[0106] For example, the relationship between load allocation rules and temperature data is analyzed by combining preset load adjustment parameters (upper limit 70, fluctuation ≤15). The load value is normalized dimensionless data, ranging from 0 to 100, and fluctuation ≤15 means that the absolute change value does not exceed 15 units. Low load adaptation zones A1 and A4 have low temperatures (consistently below 26℃) and can handle more load, with the load increasing from the current value to 55-64 (fluctuation ≤15); medium load adaptation zone A2 has a temperature close to the threshold, and the load needs to be maintained at 65-68 to avoid exceeding the upper limit of 70; restricted load zone A3 has a history of high risk, and the load needs to be reduced to 53-55. Based on the real-time resource utilization of each zone, specific values ​​are selected within the preset range, reserving more safety redundancy, resulting in the load adjustment parameter configurations: A1: 49→58, A2: 65→65, A3: 68→53, A4: 52→58.

[0107] It should be noted that the upper limit of 70 is explained in step S12, and the load fluctuation ≤15 is to avoid temperature runaway and hardware shock. Because the temperature rises by 0.32℃ for every 1% increase in load, if the fluctuation exceeds 15, such as rising from 50 to 66 within 10 minutes, the temperature will rise sharply by more than 4.8℃. The heat dissipation system takes 15-30 minutes to respond, which can easily cause short-term heat accumulation. At the same time, the data read and write error correction rate of SSD storage chips will decrease when the load fluctuates drastically. Fluctuation ≤15 can ensure stable hardware operation.

[0108] In this embodiment, the preset load allocation rules are dynamically updated according to the configuration. The original rule was average allocation, which is updated to allocation based on regional temperature levels, generating a load allocation scheme. The excess load in region A3 is transferred to A1 and A4 respectively, while A2 remains unchanged. The scheme clearly defines the adjusted load value, load transfer path, and corresponding temperature control target for each region. The temperature is controlled based on the linear relationship between load and temperature, according to the conclusion in step S11 that for every increase of 1 in load intensity, the average temperature of the flash memory region rises by 0.32℃. For example, if the load in A1 increases from 49 to 58, the expected temperature rises from 25℃ to 27.88℃, which is below the 50℃ safety threshold.

[0109] For example, the load transfer path from A3 to A1 and A4 needs to be executed according to the logic of redundancy matching, task characteristic adaptation, and risk control. First, calculate the excess load of A3. The initial load of A3 is 68, which needs to be reduced to 53, an excess of 15. Then, match the load-bearing redundancy of A1 and A4. The current load of A1 is 49, which needs to be increased to 58, a redundancy of 9. The current load of A4 is 52, which needs to be increased to 58, a redundancy of 6. The total redundancy of 15 can fully handle the excess load of A3. Next, according to the task characteristic adaptation, prioritize transferring the 8 tasks in A3 that can be processed with delay, such as offline log analysis tasks, to A1. Since the current temperature of A1 is 25℃, which is lower than that of A4 at 25.5℃, the heat dissipation redundancy is more sufficient, which can buffer the temperature changes caused by the increase in load. The remaining 7 tasks with high requirements for storage I / O adaptability, such as high-frequency small file read and write tasks, are transferred to A4 to match the I / O characteristics of A4 storage chip and reduce task execution loss. Simultaneously, the load transfer order is clearly defined: first, transfer loads that can be processed with a delay to A1; after the load on A1 stabilizes, transfer loads with high I / O adaptability to A4, avoiding short-term temperature fluctuations caused by both areas simultaneously handling loads. This ultimately forms a complete transfer path: A3→A1: 8 loads that can be processed with a delay, A3→A4: 7 loads with high I / O adaptability.

[0110] In step S15, the process of filtering tasks that meet preset task conditions in the load allocation scheme to determine an initial task list includes:

[0111] Extract the tasks that need to be prioritized from the load allocation scheme and determine the target task list;

[0112] Obtain the real-time hardware status indicators of the temperature division region;

[0113] The real-time hardware status indicators are compared with the target task list. If the real-time hardware status indicators meet the preset task conditions of the target task list, then the tasks in the target task list are marked as transferable tasks.

[0114] The execution priorities of the transferable tasks are sorted out to determine the initial task list.

[0115] In one possible implementation, task priority rules are clearly defined, using temperature risk as the basis for classification. These rules apply to all tasks involved in the load balancing scheme. For example, tasks that are likely to cause high temperature risk after execution, such as single read / write operations exceeding 100MB and lasting for more than 2 hours, are classified as P0 level, such as real-time user data read / write and cross-regional big data synchronization; tasks that only generate medium temperature risk after execution, such as single read / write operations of 30-100MB and lasting for 0.5-2 hours, are classified as P1 level, such as data backup and log summary analysis; and tasks that have almost no temperature risk after execution, such as single read / write operations of less than 30MB and lasting for less than 0.5 hours, are classified as P2 level, such as system redundancy detection and routine disk space inspection.

[0116] It should be noted that from the tasks involved in the load balancing scheme, all P0 and P1 level tasks are selected. For example, among the 8 tasks currently running in region A3, 4 are P0 level and 3 are P1 level; among the tasks currently running in regions A1 and A4, all are P1 and P2 level, with no P0 level tasks. The final target task list is then determined, forming a target task list containing 7 tasks with the 4 P0 level and 3 P1 level tasks in region A3 as the core.

[0117] Resource data is extracted based on the hardware resources required for task transfer. Specific data for each region is as follows:

[0118] A1 (Low Load Adaptation Zone): CPU utilization 52%, memory free capacity 120GB, storage IOPS remaining 8000, heat dissipation redundancy 4℃ (current 25℃, threshold 29℃).

[0119] A4 (Low Load Adaptation Zone): CPU utilization 48%, free memory capacity 150GB, remaining storage IOPS 9500, thermal redundancy 3.5℃ (currently 25.5℃, threshold 29℃).

[0120] A3 (Limited Load Area): CPU utilization 68%, memory free capacity 30GB, storage IOPS remaining 2000, thermal redundancy 0.5℃ (currently 26.5℃, threshold 27℃).

[0121] A2 (Medium Load Adaptation Zone): CPU utilization 65%, memory free 50GB, maintain load according to plan, and will not participate in task migration for the time being.

[0122] For example, the migration conditions are set as follows: remaining CPU utilization in the target area ≥ task requirements, free memory capacity ≥ task requirements, and storage IOPS ≥ task requirements. Based on historical SSD cluster operation data, system overhead averages approximately 20% of resources, therefore an 80% safety margin is added. The minimum value among CPU, memory, and IOPS that can be handled is taken and rounded down. The resource requirements of each task in the target task list are compared with the resource data of A1 and A4. A3 has four P0-level tasks, each requiring 8% CPU utilization, 20GB memory, and 1200 IOPS; A1 has 48% remaining CPU (38% after considering the safety margin, which can handle 4 tasks), 120GB memory (96GB after considering the safety margin, which can handle 4 tasks), and 8000 IOPS (6400 after considering the safety margin, which can handle 5 tasks), meeting the migration requirements.

[0123] For example, each of A3's three P1-level tasks requires 5% CPU utilization, 15GB of memory, and 800 IOPS. The remaining resources in A1 and A4 can also meet these requirements, but priority must be given to P0-level tasks. Adding new P1-level tasks: These will be allocated based on remaining resources after the A3 tasks are transferred. Finally, all four P0-level and three P1-level tasks in A3 meet the transfer criteria and are all marked as transferable tasks.

[0124] The tasks were prioritized based on task priority > temperature risk level. Priority was given to transferring P0-level tasks from A3, with two tasks requiring higher IOPS (1500, while typical tasks only require 800) allocated to A4 (which has more IOPS remaining), and the other two to A1. P1-level tasks from A3 were also transferred, with one allocated to A1 and two to A4 to avoid overloading a single region and causing resource strain. Newly added P1-level tasks were not included in the initial list but will be added after the core task transfers are completed. The final initial task list included two P0-level and one P1-level tasks transferred from A3 to A1, and two P0-level and two P1-level tasks transferred to A4, clearly defining the name, resource requirements, and target region for each task.

[0125] In step S16, if there are consecutive high-intensity read / write operations in the initial task list, the consecutive high-intensity read / write operations are reallocated to other partitions of the SSD to generate an optimized task list, including:

[0126] The read / write frequency and task type are obtained from the initial task list, and key features are extracted. If the extraction result exceeds the preset intensity threshold, a preliminary task set is obtained.

[0127] Obtain the resource utilization rate of each SSD partition, and filter out the partitions whose resource utilization rate is lower than a preset utilization rate threshold as target partitions;

[0128] According to the preset task allocation rules, the initial task set is redistributed to the target partition to generate an optimized task list.

[0129] In one possible implementation, the preset intensity threshold for continuous high-intensity read and write operations is set based on the SSD hardware characteristics. The standard is a read / write frequency of ≥800 times / minute and a single read / write data volume of ≥100MB within 10 minutes. Exceeding this threshold will cause the instantaneous heat generation of the SSD chip to increase by 40%, which can easily lead to a sudden increase in local temperature.

[0130] Extract read / write data from the initial task list. Task T1 (P0 level), handling real-time writing of user orders, had a read / write frequency of 920 times / minute and a single read / write volume of 120MB / s over 10 minutes, exceeding the threshold. Task T5 (P0 level), handling cross-regional data synchronization, had a read / write frequency of 880 times / minute and a single read / write volume of 150MB / s over 10 minutes, also exceeding the threshold. Other tasks, such as daily data aggregation (P1 level), had read / write frequencies below 700 times / minute, not reaching the threshold. T1 and T5 are combined into a preliminary task set.

[0131] It should be noted that read / write frequency statistics require first defining a fixed statistical period of 10 consecutive minutes. The number of read and write requests per second associated with the task is collected once per second via the SSD's SMART interface, with timestamps accurate to milliseconds recorded synchronously. The total IOPS per second is calculated, and all total IOPS within the 10-minute period are summed to obtain the total number of requests. Finally, the total number of requests is divided by 10 to obtain the read / write frequency per minute.

[0132] Data volume statistics are calculated by collecting the actual data transfer size of each valid read / write request through the SSD hardware interface within the same 10-minute period, excluding non-task-related data related to system-level overhead, and averaging the collected single read / write data volume to obtain the single read / write data volume.

[0133] For example, preset thresholds for resource utilization are set: CPU utilization ≤ 60%, free memory ≥ 80GB, and SSD read / write bandwidth usage ≤ 50%, ensuring the target partition has sufficient redundancy to handle high-intensity tasks. Real-time resource data is collected from the four rack partitions in Area A. A1's CPU utilization is 62% (above 60%), free memory is 85GB, and bandwidth usage is 48%, which does not meet the standards; A2's CPU utilization is 58%, free memory is 95GB, and SSD bandwidth usage is 40%, which meets the standards; A3's CPU utilization is 62% (above 60%), free memory is 90GB, and bandwidth usage is 42%, which does not meet the standards; A4's CPU utilization is 61%, free memory is 65GB (below 80GB), and bandwidth usage is 68% (above 50%), which does not meet the standards. Ultimately, only partition A2 meets the requirements and is selected as the target partition.

[0134] It should be noted that the preset threshold for resource utilization should be set with reference to the resource consumption of high-intensity tasks (such as T1 and T5) in the initial task set. A single task should occupy 15%-20% of CPU utilization, 20GB-30GB of memory, and 15%-20% of SSD read / write bandwidth. The threshold should reserve space for at least two such tasks. For example, CPU ≤ 60% can accommodate 3 tasks, and memory ≥ 80GB can accommodate 3 tasks to avoid resource overload after taking on more tasks. Secondly, considering the characteristics of SSD hardware, CPU utilization exceeding 60% will lead to task scheduling delays, SSD bandwidth usage exceeding 50% can easily cause IO blocking, and less than 80GB of free memory will increase data exchange time. All of these will exacerbate hardware heat generation. Setting the threshold within a safe range can ensure stable hardware operation and controllable heat generation.

[0135] The default task allocation rule follows a priority order, prioritizing task priority over resource suitability, when allocating tasks T1 and T5. Priority is assigned to T1 (P0 level, for user order writes) and T5 (P0 level, for cross-region synchronization, with the same priority as T1). Further prioritization is based on resource requirement matching. T5 has a single read / write requirement of 150MB. Considering the overall capacity and usage of the A2 SSD, the actual available space calculated from its remaining SLC cache ratio fully covers T5's 150MB read / write requirement, demonstrating higher suitability. While T1's single read / write requirement of 120MB can also be met by the space corresponding to the A2's current remaining SLC cache ratio, its accuracy in matching requirement with available cache is lower than T5; therefore, T5 is allocated first.

[0136] For example, when reassigning tasks, T5 is moved from A4 to A2. After occupying some of A2's memory and bandwidth, A2's CPU utilization rises to 65% (still ≤70% upper limit), memory free space drops to 75GB (still sufficient), and bandwidth usage rises to 50% (reaching the threshold but not exceeding it). T1 remains on A1 for the time being, and temporary cooling enhancement is enabled for A1, with the fan speed increased to 90% to prevent its temperature from rising due to continuous operation of T1.

[0137] Clearly label the final destination of each task, such as T5→A3, T1→A1, and add heat dissipation measures, task priority, read and write parameters and resource usage to ensure that no single partition bears too many continuous high-intensity read and write tasks, and generate an optimized task list.

[0138] In step S17, recording the temperature and performance fluctuations during the operation of the optimized task list and generating a temperature stability record includes:

[0139] Obtain temperature feedback data and performance indicators during the operation from the list of optimization tasks;

[0140] The timestamps of the temperature feedback data and the performance indicators are classified and organized, and features are extracted to obtain temperature fluctuation records and performance indicator fluctuation records.

[0141] By associating the temperature fluctuation record and the performance index fluctuation record, a temperature stability record is generated.

[0142] In one possible implementation, the core data types and frequencies to be collected are clearly defined. Temperature feedback data is collected every minute, including the ambient temperature of each partition, such as the rack temperature of A1 and A3, and the temperature of SSD core components, such as the chip temperature of the SSDs where T1 and T5 are running. Performance indicators are collected every 30 seconds, covering the task level, the read / write frequency and data throughput of T1 and T5, and the CPU utilization, memory usage, and SSD read / write bandwidth usage of partitions A1 and A3. Actual data collection results: After A3 took over from T5, at the 5th minute, the ambient temperature was 26.1℃, the SSD chip temperature was 44℃, the T5 read / write frequency was 870 times / minute, the throughput was 145MB / s, the A3 CPU utilization was 64%, and the SSD bandwidth usage was 48%. During the T1 operation, at the 8th minute, the ambient temperature was 25.8℃ (the temperature did not rise due to fan enhancement), the SSD chip temperature was 43℃, the T1 read / write frequency was 910 times / minute, the throughput was 118MB / s, the A1 CPU utilization was 59%, and the SSD bandwidth usage was 61%.

[0143] In this embodiment, temperature fluctuation records are organized by partition and task, with data sorted by timestamp, and the maximum, minimum, and fluctuation range of the temperature are marked. For example, A3 (T5) has a temperature range of 26-26.3℃ within 1 hour, with a fluctuation range of 0.3℃, and there are no cases where the temperature exceeds the control target of 26℃; A1 (T1) has a temperature range of 25.7-25.9℃, with a fluctuation range of 0.2℃, indicating that the heat dissipation enhancement measures are effective.

[0144] Performance fluctuation records were also organized by timestamp, and performance stability characteristics were extracted. The T5 read / write frequency was stable at 860-880 times / minute, with fluctuations ≤2.3%, and no stuttering. The T1 read / write frequency was stable at 900-920 times / minute, with fluctuations ≤2.2%, and no significant decrease in data throughput. At the partition level, the A3 CPU utilization was 62%-64% (≤70% upper limit) and the SSD bandwidth usage was 42%-48% (≤50% threshold). The A1 CPU utilization was 58%-59% and the SSD bandwidth usage was 43%-47% (≤50% threshold), both of which were marked as stable performance.

[0145] For example, temperature fluctuation records and performance index fluctuation records are associated with the same timestamp to form a complete record. If the performance index fluctuates greatly at a certain point in time, such as a sudden drop of 30% in read / write frequency, the corresponding temperature is checked simultaneously to determine if it is caused by excessive temperature. If the temperature is stable but the performance fluctuates, it is marked as a non-temperature factor, such as network latency.

[0146] A temperature stability record is generated every hour, including fields such as time interval, partition, task name, temperature range and fluctuation amplitude, performance index range and fluctuation amplitude, and related conclusions. For example, "14:00-15:00, A3 partition, T5 task, temperature 26-26.3℃ (fluctuation 0.3℃), read / write frequency 860-880 times / minute (fluctuation 2.3%). Conclusion: Temperature stable, performance normal, meets operating requirements"; "14:00-15:00, A1 partition, T1 task, temperature 25.7-25.9℃ (fluctuation 0.2℃), read / write frequency 900-920 times / minute (fluctuation 2.2%), conclusion: heat dissipation enhancement effective, temperature and performance both stable."

[0147] It should be noted that the hourly temperature stability record is generated based on the fact that SSD temperature is affected by load. Under normal tasks, temperature fluctuations are mostly slow increases or stabilization. The hourly cycle can completely capture the process of a single fluctuation without missing any key changes. At the same time, 24 records per day can form a clear timeline data chain without data redundancy due to too many records, and can also be quickly located during subsequent review.

[0148] In summary, this invention discloses an adaptive temperature control method based on SSDs, which solves the problem of performance fluctuations caused by the uncontrollable temperature of SSDs.

[0149] Reference Figure 2 The second embodiment of the present invention provides an SSD-based adaptive temperature control system, comprising:

[0150] The temperature data processing module is used to adjust the heat dissipation if the risk assessment results show that the potential heat accumulation risk points exceed the preset risk threshold, and to collect the temperature data after the heat dissipation adjustment and generate a temperature distribution map.

[0151] The scheme generation module is used to divide the temperature distribution map into regions and parse the data, and generate a load allocation scheme by combining preset load adjustment parameters.

[0152] The task list determination module is used to filter tasks that meet preset task conditions in the load allocation scheme and determine the initial task list.

[0153] The task optimization module is used to reallocate continuous high-intensity read and write operations to other partitions of the SSD if there are such operations in the initial task list, thereby generating an optimized task list.

[0154] The recording module is used to record the temperature and performance fluctuations during the operation of the optimization task list and generate a temperature stability record.

[0155] It should be noted that the SSD-based adaptive temperature control system provided in this embodiment of the invention is used to execute all the process steps of the SSD-based adaptive temperature control method in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0156] This invention also provides an electronic device. The electronic device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, such as an SSD-based adaptive temperature control program. When the processor executes the computer program, it implements the steps described in the various embodiments of the SSD-based adaptive temperature control method, for example... Figure 1 The step S11 shown. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above-described device embodiments, such as the temperature data processing module.

[0157] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the electronic device.

[0158] The electronic device may be a desktop computer, laptop, handheld computer, or smart tablet, etc. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. It may include more or fewer components than described above, or combine certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0159] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the electronic device, connecting all parts of the electronic device via various interfaces and lines.

[0160] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0161] Wherein, if the modules / units integrated in the electronic device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0162] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and 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. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0163] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for adaptive temperature control based on SSD, characterized in that, The method comprises the following steps: Collecting internal sensor data and external environment parameters of the SSD, forming an initial data set, and performing trend extraction and correlation analysis on the initial data set to obtain a load change trend; According to the external environment parameters and the load change trend, identify potential heat accumulation risk points, and generate a risk assessment result; If the risk assessment result shows that the potential heat accumulation risk point exceeds the preset risk threshold, perform heat dissipation adjustment, collect temperature data after heat dissipation adjustment, and generate a temperature distribution map; Divide the temperature distribution map into regions and analyze the data, combine the preset load adjustment parameters to generate a load distribution scheme; Screen the tasks in the load distribution scheme that meet the preset task conditions to determine an initial task list; If there are continuous high-intensity read-write operations in the initial task list, the continuous high-intensity read-write operations are redistributed to other partitions of the SSD to generate an optimized task list; Record the temperature and performance fluctuations during the running of the optimized task list to generate a temperature stable state record.

2. The adaptive temperature control method based on SSD according to claim 1, wherein, The method comprises the following steps: Collecting internal sensor data and external environment parameters of the SSD, forming an initial data set, and performing trend extraction and correlation analysis on the initial data set to obtain a load change trend, comprising: Collecting internal running data and temperature data of the SSD through an internal sensor interface of the SSD, and obtaining environmental temperature data and humidity data using external environmental monitoring equipment; Summarize the internal running data, temperature data, environmental temperature data, and humidity data to form an initial data set; Based on a preset fluctuation threshold, analyze the load intensity fluctuations in the initial data set to determine the load fluctuations; 3. The adaptive temperature control method based on SSD according to claim 2, wherein, Correlation analysis of the initial data set and the load fluctuations to determine the load change trend. The method comprises the following steps: According to the external environment parameters and the load change trend, combine historical load data, screen and quantify core features to form a feature data set; According to the external environment parameters and the load change trend, combine historical load data, screen and quantify core features to form a feature data set; According to the external environment parameters and the load change trend, combine historical load data, screen and quantify core features to form a feature data set; 4. The adaptive temperature control method based on SSD according to claim 1, wherein, If the load intensity in the segmented feature data set exceeds the preset load intensity threshold and the environmental temperature data is higher than the preset environmental temperature threshold, mark it as a high-risk interval; Input the feature data of the high-risk interval into a preset time prediction model to obtain a temperature rise risk probability, if the temperature rise risk probability exceeds the preset risk threshold, it is determined that there is a heat accumulation risk; Combine the heat accumulation risk and the feature data set to extract the risk trigger condition, label the warning identifier of the risk trigger condition, determine the potential heat accumulation risk point, and generate a risk assessment result. The method comprises the following steps: According to the external environment parameters and the load change trend, combine historical load data, screen and quantify core features to form a feature data set; According to the external environment parameters and the load change trend, combine historical load data, screen and quantify core features to form a feature data set; According to the external environment parameters and the load change trend, combine historical load data, screen and quantify core features to form a feature data set; extracting temperature data of the temperature division area and current load data, combining the temperature data of the temperature division area and the current load data to obtain core operation parameters; analyzing a relationship between the core operation parameters and preset load adjustment parameters to obtain load adjustment parameter configurations; updating a preset load allocation rule dynamically according to the load adjustment parameter configurations, and obtaining a load allocation scheme.

5. The adaptive temperature control method based on SSD according to claim 4, wherein, The method for screening tasks meeting preset task conditions in the load allocation scheme to determine an initial task list includes: extracting tasks that need to be executed preferentially from the load allocation scheme to determine a target task list; obtaining real-time hardware state indicators of the temperature division area; comparing the real-time hardware state indicators with the target task list, and if the real-time hardware state indicators meet preset task conditions of the target task list, marking tasks in the target task list as transferable tasks; combing execution priorities of the transferable tasks to determine an initial task list.

6. The adaptive temperature control method based on SSD according to claim 1, wherein, If there are continuous high-intensity read-write operations in the initial task list, the continuous high-intensity read-write operations are redistributed to other partitions of the SSD to generate an optimized task list, including: obtaining read-write frequencies and task types from the initial task list and performing key feature extraction, and if the extraction result exceeds a preset intensity threshold, a preliminary task set is obtained; obtaining resource utilization rates of each partition of the SSD, and screening out a target partition with a resource utilization rate lower than a preset utilization rate threshold; redistributing the preliminary task set to the target partition according to a preset task allocation rule to generate an optimized task list.

7. The adaptive temperature control method based on SSD according to claim 1, wherein, The method for recording temperature and performance fluctuation conditions in the running process of the optimized task list to generate a temperature stable state record includes: obtaining temperature feedback data and performance indicators in the running process from the optimized task list; classifying and organizing time stamps of the temperature feedback data and the performance indicators, and extracting features to obtain temperature fluctuation records and performance indicator fluctuation records; associating the temperature fluctuation records and the performance indicator fluctuation records to generate a temperature stable state record.

8. An adaptive temperature control system based on SSD, characterized in that, The method includes: a data acquisition and trend analysis module for acquiring internal sensor data and external environmental parameters of the SSD, aggregating to form an initial data set, and performing trend extraction and correlation analysis on the initial data set to obtain a load change trend; a risk identification and evaluation module for identifying potential heat accumulation risk points according to the load change trend, and generating a risk evaluation result; a temperature data processing module for performing heat dissipation adjustment if the risk evaluation result shows that the potential heat accumulation risk points exceed a preset risk threshold, and collecting temperature data after the heat dissipation adjustment to generate a temperature distribution map; a scheme generation module for performing area division and data analysis on the temperature distribution map, and generating a load allocation scheme in combination with preset load adjustment parameters; a task list determination module for screening tasks meeting preset task conditions in the load allocation scheme to determine an initial task list; The task optimization module is configured to, if there is a continuous high-intensity read-write operation in the initial task list, re-allocate the continuous high-intensity read-write operation to other partitions of the SSD, and generate an optimized task list. The recording module is configured to record temperature and performance fluctuation conditions during running of the optimized task list, and generate a temperature stable state record.

Images