Composite stress testing methods, apparatus, equipment and media for solid-state drives

Abstract

This invention discloses a method, apparatus, device, and medium for composite stress testing of solid-state drives (SSDs). The method includes: loading test configuration parameters and establishing a communication connection with the SSD under test, which is pre-placed in a preset test environment; performing performance baseline calibration and preprocessing on the SSD under test; performing composite stress performance testing, dynamically coupling at least two of temperature stress, capacity occupancy stress, power interruption stress, garbage collection induced stress, and mixed load stress; synchronously collecting test data and constructing a time-series data warehouse; and performing intelligent deep analysis based on the time-series data warehouse, including performance consistency deep analysis and health and risk prediction analysis. This invention simulates a real, complex environment through composite stress testing, solving the problem of limited scenarios in traditional testing methods. Simultaneously, intelligent deep analysis addresses the issue of superficial data analysis, enabling a comprehensive and accurate assessment of SSD reliability.

Description

Technical Field

[0001] This invention relates to the field of solid-state drive (SSD) testing technology, and more particularly to a composite stress testing method, apparatus, equipment, and medium for SSDs. Background Technology

[0002] Solid-state drives (SSDs), as core storage components of computer systems, are widely used in consumer electronics, enterprise storage, and data centers due to their high performance, low latency, and shock resistance. The performance and reliability of SSDs directly impact the performance of the entire computing system; therefore, rigorous stress testing is crucial during their research, development, quality control, and selection. Currently, the industry-standard stress testing methods are typically based on standard performance testing tools, which can provide a preliminary assessment of the basic performance of SSDs, forming the current paradigm for SSD stress testing. Despite the widespread adoption of these existing testing methods, with the development of SSD technology and the increasing complexity of application scenarios, their inherent limitations are becoming increasingly apparent, leading to the following prominent issues in SSD reliability testing:

[0003] Firstly, in practical applications, solid-state drives often face the combined effects of multiple stress conditions such as high temperature, high capacity usage, power interruption, and garbage collection pressure. Traditional testing methods usually apply stress in isolation, such as only testing high temperature or only testing full disk, which cannot reproduce the complex scenario of multiple harsh conditions occurring simultaneously or alternately in the real world, resulting in deviations between test results and reality.

[0004] Secondly, current testing methods mostly focus on the statistical level of performance indicators, paying only attention to the average values ​​of basic indicators such as IOPS and bandwidth. They lack in-depth analysis of test data and fail to make reasonable use of the test data, resulting in test results that cannot fully reflect the true reliability of solid-state drives and are difficult to provide effective support for product quality improvement and predictive maintenance. Summary of the Invention

[0005] This invention provides a composite stress testing method, apparatus, computer equipment, and medium for solid-state drives (SSDs) to address the technical problems of limited testing scenarios and insufficient depth of test data analysis in existing SSD testing methods.

[0006] Firstly, a composite stress testing method for solid-state drives is provided, including:

[0007] Load test configuration parameters and establish a communication connection with the solid-state drive under test that has been placed in the preset test environment;

[0008] Perform performance baseline calibration and pre-testing processing on the solid-state drive under test;

[0009] Based on the preset test strategy, benchmark performance test and composite stress performance test are performed on the solid-state drive under test. Test data during the test process are collected simultaneously, and a time-series data warehouse is built. The composite stress includes at least two of the following: temperature stress, capacity occupancy stress, power interruption stress, garbage collection induced stress, and mixed load stress, which are dynamically coupled.

[0010] Based on a time-series data warehouse, intelligent deep analysis is performed, including at least one of performance consistency deep analysis and health and risk prediction analysis.

[0011] Secondly, a composite stress testing device for solid-state drives is provided, comprising:

[0012] The configuration loading module is used to load test configuration parameters and establish a communication connection with the solid-state drive under test that has been placed in a preset test environment.

[0013] The preprocessing module is used to perform performance baseline calibration and preprocessing of the solid-state drive under test.

[0014] The testing module is used to perform benchmark performance tests and composite stress performance tests on the solid-state drive under test based on a preset testing strategy. It simultaneously collects test data during the testing process and builds a time-series data warehouse. The composite stress includes at least two of the following: temperature stress, capacity occupancy stress, power interruption stress, garbage collection induced stress, and mixed load stress, which are dynamically coupled.

[0015] The analysis module is used to perform intelligent deep analysis based on the time-series data warehouse. The intelligent deep analysis includes at least one of performance consistency deep analysis and health and risk prediction analysis.

[0016] Thirdly, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the composite stress testing method for the solid-state drive described above.

[0017] Fourthly, a computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements the steps of the composite stress testing method for the solid-state drive described above.

[0018] The aforementioned composite stress testing method, apparatus, equipment, and storage medium for solid-state drives (SSDs) utilizes a highly realistic testing environment by dynamically coupling multiple stresses, including temperature stress, capacity occupancy stress, power interruption stress, garbage collection-induced stress, and mixed load stress. This effectively solves the technical problem of traditional testing methods being isolated and unable to reproduce complex real-world application scenarios. This composite stress testing method can simulate various extreme conditions that SSDs may encounter in actual use, significantly improving test coverage and realism, and allowing potential defects to be more fully exposed. At the data analysis level, this invention breaks through the limitations of traditional testing that only focuses on average performance indicators. By constructing a time-series data warehouse and an intelligent deep analysis system, it achieves a leap from shallow statistics to deep mining, accurately quantifying the performance stability of SSDs, realizing quantitative health scoring, risk pattern identification, and remaining lifespan prediction, providing a scientific basis for product quality assessment and predictive maintenance. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart illustrating a composite stress testing method for a solid-state drive according to an embodiment of the present invention.

[0021] Figure 2 This is a schematic diagram of a composite stress testing device for a solid-state drive according to an embodiment of the present invention.

[0022] Figure 3 This is a schematic diagram of the structure of a computer device according to an embodiment of the present invention. Detailed Implementation

[0023] 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, not all, of the embodiments of the present invention. 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.

[0024] Please see Figure 1 As shown, Figure 1 A schematic flowchart of a composite stress testing method for a solid-state drive provided in an embodiment of the present invention includes the following steps:

[0025] Step S1: Load test configuration parameters and establish a communication connection with the solid-state drive under test that is pre-placed in the preset test environment.

[0026] Specifically, in this embodiment, the test environment is pre-deployed. The solid-state drive under test is installed in a programmable high and low temperature chamber and connected to a high-performance test platform equipped with a power circulator, and its parameters are initialized. The basic operating parameters of the programmable high and low temperature chamber include the initial temperature setpoint (usually set to 25°C standard ambient temperature), temperature variation range (e.g., -10°C to 85°C), heating / cooling rate (e.g., 5°C / minute), etc. The basic operating parameters of the power circulator include the power-off duration (usually 10 to 30 seconds), power-on interval, maximum number of power-offs limit, etc. In addition, the sampling frequency, data format, storage path, etc. of each monitoring tool (e.g., fio, smartctl, iostat) also need to be set.

[0027] Step S2: Perform performance baseline calibration and pre-testing preprocessing on the solid-state drive under test.

[0028] Specifically, before the formal test, a series of standardized, single loads were applied to the solid-state drive under test in a standard environment (25℃) using a professional I / O load generation tool (such as fio). Key performance indicators in the above test were collected and recorded, mainly including: IOPS (read and write operations per second), bandwidth (MB / s), and latency (average latency, tail latency such as P99.9). These data were stored as performance baseline calibration, which is an important standard for "performance recovery judgment" and "performance degradation calculation" in subsequent tests.

[0029] The purpose of preprocessing before testing is to reset the solid-state drive from its "brand new" or "unknown" state and simulate a stable "test start state" that meets the testing requirements, thereby eliminating the interference of residual data and historical states on the test results. Specifically, this includes:

[0030] 1. Raw device mapping:

[0031] The test is performed directly on the solid-state drive under test (such as / dev / nvme0n1) to eliminate interference from file system caching and overhead, and to obtain the lowest-level storage performance data.

[0032] 2. Secure erasure and metadata reset:

[0033] Use SSD-specific commands (such as the Format NVM command or Sanitize command in the NVMe standard) to perform a secure erase on the drive. This operation will erase all user data and force the SSD's flash translation layer (FTL) and other critical metadata to reset, logically restoring it to a state similar to that of a new drive at the factory, ensuring that the starting point for each test is absolutely consistent.

[0034] 3. Simulation of fragmented state:

[0035] To simulate the state after a period of actual user use, rather than an idealized new state, it is necessary to actively create data fragmentation within the hard drive. Using testing tools, the hard drive is filled to a preset occupancy rate (e.g., 50%) in batches using random write mode. During the filling process, large-scale sequential writes and small-scale random writes are alternated. This irregular write pattern can actively create fragmented data distribution in the physical space of the NAND flash memory, more realistically simulating the scenario after long-term use.

[0036] 4. Metadata stress preheating:

[0037] Perform short, high-intensity random write operations (e.g., 1 million 4KB random writes with a high queue depth). The purpose of this operation is to quickly trigger the SSD's internal garbage collection (GC) and wear leveling mechanisms, and to bring metadata such as the FTL mapping table to a certain level of activity. This ensures that when subsequent formal testing begins, the SSD's firmware and controller are already in an active working state, rather than a cold-start initial state, allowing the tests to better reflect its true performance during stable operation.

[0038] Step S3: Based on the preset test strategy, perform benchmark performance test and composite stress performance test on the solid-state drive under test, collect test data during the test process simultaneously, and build a time-series data warehouse. The composite stress includes at least two of the following: temperature stress, capacity occupancy stress, power interruption stress, garbage collection induced stress, and mixed load stress, which are dynamically coupled.

[0039] Specifically, before applying composite stress, sequential read / write (1M) and random read / write (4K) tests are repeatedly performed for 15 minutes each under a relatively mild baseline environment (e.g., 25°C, 50% utilization) to establish a performance benchmark. This benchmark is used to more accurately measure relative performance changes in subsequent composite stress tests, avoiding directly treating the pre-processed state as the benchmark. After benchmark performance testing, a central control unit dynamically couples temperature stress, capacity utilization stress, power interruption stress, garbage collection induced stress, and mixed load stress, coordinating multiple tests to be performed simultaneously or alternately, rather than sequentially executing a single test. "Dynamic coupling" in composite stress performance testing refers to applying at least two of the following stresses—temperature stress, capacity utilization stress, power interruption stress, garbage collection induced stress, and mixed load stress—to the solid-state drive under test in a combined manner, either simultaneously, alternately, or in a preset logical order, to simulate the combined influence of multiple stress factors in a real, complex environment. For example, coupling between temperature stress and capacity occupancy stress can be achieved by adjusting temperature changes while performing capacity occupancy tests, and testing the performance of the solid-state drive under test in this coupled environment.

[0040] Furthermore, the composite stress performance testing includes temperature and capacity coupling testing, specifically including:

[0041] 1.1 Increase the utilization rate of the solid-state drive under test to the preset utilization rate.

[0042] Specifically, capacity occupancy stress is first applied. High occupancy significantly increases the metadata management pressure on the flash translation layer (FTL) inside the solid-state drive (SSD), reduces available spare blocks, and makes garbage collection (GC) operations more frequent and inefficient. The preset occupancy rate is usually set to an extremely high level, such as 90%, 95%, or higher, to simulate the extreme case of almost being filled. Using I / O load generation tools (such as fio), data is continuously written to the SSD under test in sequential or random write modes until its user-available space reaches the preset occupancy target. After this stage, the SSD enters a "high-pressure" initial state, preparing for the subsequent introduction of temperature stress.

[0043] 1.2 While maintaining the preset occupancy rate, control the ambient temperature to change from the first preset temperature to the second preset temperature at a predetermined rate, and continuously apply I / O load during the temperature change process, and record the test data.

[0044] Specifically, the capacity utilization test is not conducted at a stable temperature, but rather during a process of continuous temperature change, thereby exposing the SSD's sensitivity to temperature gradients. The first preset temperature is typically room temperature or a standard temperature, such as 25°C. The second preset temperature is a target high temperature, set according to product specifications and test severity levels, such as 70°C or 85°C. The preset rate controls the chamber to linearly increase in temperature at a specific rate (e.g., 5°C / minute) to simulate rapid deterioration of ambient temperature. Throughout the heating process, a low to moderate I / O load (e.g., 4KB random write) is continuously applied to keep the SSD controller and flash memory actively operating. The dynamic response curves of performance parameters (e.g., latency, IOPS) as a function of temperature are observed. During observation, changes in temperature, performance data, and SMART attributes (e.g., temperature sensor readings, error counts) are simultaneously collected and recorded to establish a time correlation.

[0045] 1.3. Under steady-state conditions of maintaining the preset occupancy rate and the third preset temperature, continuously apply the I / O load for the first preset duration and record the test data.

[0046] Specifically, after completing the temperature variation test, the SSD under test undergoes a long-term durability and stability test under steady-state conditions. During this phase, data errors caused by thermal throttling, metadata overflow, or charge leakage are highly likely to occur. A third preset temperature is pre-set, which can also be set to the same as the second preset temperature, maintaining the highest temperature reached in the previous steps (e.g., 70°C) and a preset utilization rate to ensure consistent capacity pressure throughout the test. The first preset duration is set to a relatively long time, such as 4 hours, 8 hours, or longer, to fully expose potential defects. Mixed read / write loads are applied within the first preset duration to simulate mixed read / write scenarios in real-world applications, continuously testing the SSD's FTL management, garbage collection, and data retention capabilities. Throughout the test, all performance, health, and environmental data are monitored and recorded, with a focus on any sudden performance drops, spikes in error counts, or other abnormal behaviors.

[0047] Furthermore, the composite stress performance test includes intelligent waste recycling stress testing, specifically including:

[0048] 2.1 During the burst write phase, data is continuously written until the available space on the hard disk is lower than the preset threshold or the duration reaches the second preset duration.

[0049] It should be noted that the purpose of the burst write phase is to write data to the solid-state drive at an extremely high speed, quickly consume its available space, artificially create an extreme fragmentation scenario and force the triggering of an aggressive garbage collection mechanism, thereby simulating scenarios such as users continuously copying large files or continuously writing to the database, and putting the solid-state drive in a state of "pressure saturation". After the burst write phase ends, the FTL mapping table of the solid-state drive will be very complex, the available block pool will be exhausted, and efficient garbage collection must be started immediately to release space. Specifically, I / O tools (such as fio) can be used to write large blocks of data (such as 128KB) sequentially and a high queue depth (such as 32) can be set to maximize the write throughput and fill the write bandwidth of the hard drive. The termination condition is controlled by two preset thresholds, and it will stop when one of them is met: (1) Space threshold: when the available space of the hard drive is lower than the preset threshold (such as 1%), it indicates that the hard drive has been almost full and the garbage collection pressure has reached its maximum; (2) Time threshold: the duration reaches the second preset duration (such as 5 minutes), which is to prevent endless writing to hard drives with particularly conservative recycling strategies and play a safety protection role.

[0050] 2.2 Idle Recycling and Monitoring Phase: Stop all I / O operations and leave the area idle for a third preset duration, while monitoring the activity level of idle garbage recycling.

[0051] Specifically, after external I / O pressure is stopped, the SSD controller is given a time window (i.e., the third preset duration) to perform background idle garbage collection. This is achieved by stopping all I / O operations and allowing the system to remain idle for the third preset duration (e.g., 2 minutes) to remove external interference and focus on background data cleanup and block erasure. During this process, the activity of idle garbage collection is monitored synchronously by periodically (e.g., once per second) querying relevant SMART attributes or specific logs. For example, monitoring changes in parameters such as Available Spare and Media and Data Integrity Errors allows for an indirect assessment of the efficiency and activity of garbage collection by observing the trends of these parameters during the idle period.

[0052] 2.3. In the performance recovery detection phase, a short-term performance test is performed to determine whether the performance indicators of the solid-state drive have recovered to the predetermined percentage of the baseline. If the performance indicators of the solid-state drive have not recovered to the predetermined percentage of the baseline, the idle reclamation and monitoring phase and the performance recovery detection phase are repeated, and the test data is recorded.

[0053] Specifically, immediately after the idle phase ends, a short-term (e.g., 1 minute) performance benchmark test is performed, typically using a load that sensitively reflects the hard drive's state, such as 4KB random write, with a low queue depth (e.g., 1). The average IOPS of this short-term test is then calculated. This average IOPS is then compared to a predetermined percentage (e.g., 90%) of the corresponding item under the "ideal performance baseline" established in the benchmark performance test. This determines whether the SSD's performance metrics have recovered to the predetermined percentage of the baseline. If the SSD's performance metrics have not recovered to the predetermined percentage of the baseline, the idle reclamation and monitoring phase and the performance recovery detection phase are repeated. In this embodiment, the number of times the idle reclamation and monitoring phase and the performance recovery detection phase are repeated is limited, for example, set to 3 times, to prevent entering an infinite loop.

[0054] In this embodiment, the intelligent garbage collection stress test, through a closed-loop process of applying pressure, observing recovery, quantitative evaluation, and dynamic iteration, not only tests the garbage collection capability of solid-state drives, but also conducts an in-depth evaluation of the efficiency and stability of their performance recovery. It solves the problems of insufficient or excessive testing in traditional fixed-process testing, and can intelligently and accurately expose the differences in garbage collection strategies and performance recovery of different solid-state drives.

[0055] Furthermore, the composite stress performance test includes mixed load and power interruption coupling test, specifically including:

[0056] 3.1 Configure concurrent mixed load mode, including database read and write load and log write load.

[0057] Specifically, by simulating the I / O characteristics of real server applications, rather than a single, idealized load, the test environment is ensured to be highly relevant to the actual working scenarios of SSDs. Database read / write loads are typically configured as small-block random reads and writes (e.g., 4KB), mixed according to a typical database access ratio (e.g., 70% reads / 30% writes). This load is extremely sensitive to latency and puts continuous pressure on the SSD's FTL random access capabilities. Log write loads are typically configured as large-block sequential writes (e.g., 128KB). This load continuously consumes the disk's write bandwidth and spare blocks, and affects garbage collection behavior. By using multiple threads or instances of I / O testing tools (such as fio), both of these loads are simultaneously applied to the SSD under test. This concurrent, heterogeneous I / O flow creates complex and interfering internal states, more realistically reflecting the scheduling and processing capabilities of the SSD controller.

[0058] 3.2. Apply mixed load pressure continuously during the extended test period.

[0059] It's important to understand that many potential defects (such as firmware logic errors, memory leaks, FTL entry conflicts, etc.) only become apparent after prolonged operation. Therefore, this embodiment conducts long-term, stable stress endurance and stability testing on the solid-state drive (SSD) under mixed load conditions. This extended test time is set to a relatively long period, such as 8 hours or 24 hours. Throughout the period, the mixed load is maintained continuously, keeping the SSD under high load at all times. During this time, its performance is monitored for any trend of gradual degradation over time, i.e., "performance degradation."

[0060] 3.3. Inject random power interruption events during mixed load operation and record the test data.

[0061] It is important to understand that sudden power outages during complex I / O operations pose the most severe challenge to solid-state drive (SSD) data protection mechanisms, metadata atomicity operations, and fault recovery processes. Therefore, this embodiment injects random power interruption events during mixed load operation by pre-setting an adaptive power outage strategy. The timing of the power interruption is not fixed but occurs randomly throughout the long test window. This unpredictability ensures that the test covers different I / O processing stages, avoiding omissions due to fixed-timing tests.

[0062] Furthermore, in this embodiment, the random power-off strategy can be temporally random or "intelligently random" in conjunction with the testing phase. This strategy can more effectively expose the deficiencies in the data protection mechanisms of solid-state drives during critical operations. For example, the system can strategically select to trigger a power-off at the following high-risk moments:

[0063] (1) Trigger an unexpected power outage within 10 seconds after the burst write phase of the garbage recycling stress test ends;

[0064] (2) During the steady-state phase of temperature and pressure testing, power outages are performed 1-3 times randomly;

[0065] (3) During the transaction process of updating FTL metadata;

[0066] (4) At a point in time when the write cache may not be completely cleared.

[0067] In this embodiment, by combining long-term, highly complex business pressure with unpredictable catastrophic hardware failures, solid-state drive products that perform well in routine tests but may lose data or fail to recover quickly under extreme abnormal conditions are effectively screened out.

[0068] Furthermore, test data is collected synchronously during the testing process, and a time-series data warehouse is built, including:

[0069] 1. Use fio to collect performance data, smartctl to collect complete SMART log data, and iostat to collect system-level I / O information data, and timestamp all data.

[0070] It's important to understand that performance data is collected via fio and primarily includes IOPS (IOPS, read / write operations per second, distinguishing between random / sequential and read / write), bandwidth (data transfer rate), and latency (response time for a single I / O request, with a focus on average latency and 99.9th percentile latency). SMART log data records the health data of the SSD under test, mainly including: durability metrics such as NAND write volume and average wear count; reliability metrics such as spare block count and bad block count; and errors and anomalies such as unsafe shutdowns, CRC error count, and end-to-end error count. System-level I / O information data includes CPU utilization, I / O queue length, and device utilization, which helps distinguish whether performance bottlenecks originate from the hard drive itself or other parts of the system.

[0071] Specifically, the aforementioned tools and data sources operate independently, with different sampling frequencies and startup times, and the original data cannot be directly correlated. Therefore, in this embodiment, the central control unit (main control script) is used as a unified "clock server." Every time a data acquisition action occurs (e.g., fio outputs a log line, smartctl completes a query), the main control script immediately adds a timestamp obtained from the same high-precision time source. The accuracy of this timestamp is usually at the millisecond or microsecond level, ensuring accurate alignment even under high-frequency sampling.

[0072] 2. Store the collected data in a time-series database to establish a three-dimensional data warehouse for performance, health, and environmental parameters.

[0073] It should be noted that this embodiment preferably uses a time-series database (such as InfluxDB, Prometheus, or MySQL / PostgreSQL tables that support time-series data) as the storage engine, rather than a traditional relational database. Time-series databases are specifically optimized for time-series data, offering overwhelming advantages in write speed, storage efficiency, and performance in querying by time range. Data is organized chronologically, with each performance metric and each SMART parameter stored as an independent time series.

[0074] Specifically, after acquiring the collected data, a three-dimensional data warehouse integrating performance, health, and environmental parameters is constructed. For any given point in time, information across all three dimensions can be viewed simultaneously and in relation to each other within this data warehouse. For example, in terms of performance, the IOPS, bandwidth, and latency at that given time point are as follows; in terms of health, the NAND write volume, wear count, and error count at that given time point are as follows; and in terms of environment, the stability of the hard drive temperature, chassis temperature, and power supply at that given time point is as follows. This three-dimensional database enables complex correlation analysis and root cause analysis. For instance, an IOPS curve overlaid with a temperature curve over time can be plotted, revealing whether performance fluctuations are related to temperature changes. Alternatively, a query can be used to identify all events with an increase in CRC error count and trace back to the corresponding performance and temperature conditions at the time these events occurred.

[0075] Step S4: Based on the time-series data warehouse, perform intelligent deep analysis, which includes at least one of performance consistency deep analysis and health and risk prediction analysis.

[0076] Specifically, this embodiment uses a time-series data warehouse to achieve comprehensive and multi-dimensional evaluation and prediction of solid-state drive reliability.

[0077] Further, in-depth performance consistency analysis includes:

[0078] 1. Calculate the statistical distribution characteristics of IOPS and latency. The statistical distribution characteristics of latency include the mean, standard deviation and 99.9th percentile latency index.

[0079] Specifically, IOPS = Total I / O operations / Test time (seconds). The average latency reflects the average performance level during the test, the standard deviation of latency measures the dispersion of IOPS data, and the 99.9th percentile latency metric reflects the latency performance of the worst 0.1% of requests.

[0080] 2. Analyze the coefficient of variation and performance fluctuation frequency of the performance data. The coefficient of variation is calculated based on the mean and standard deviation. The performance fluctuation frequency is determined based on the number of times the performance data exceeds the stable range per unit time in a continuous time series. The stable range is set based on the mean and standard deviation.

[0081] Specifically, the coefficient of variation = (standard deviation / mean) × 100%, used to eliminate the influence of dimensions and facilitate stability comparisons between different test scenarios and different devices. Performance jitter frequency refers to the number of times performance data points exceed the stable range per unit time (usually per minute) in a continuous time series. The stable range is set with the mean as the center, extended both above and below by two standard deviations as the boundaries of the stable interval.

[0082] 3. Generate a comprehensive performance consistency assessment report based on the statistical distribution characteristics of IOPS and latency, the 99.9% high quantile latency index, the coefficient of variation, and the performance jitter frequency.

[0083] Specifically, this embodiment uses multi-dimensional data for evaluation, pre-assigning weights based on the importance of each dimension, such as: 99.9% high percentile latency: 30%, performance jitter frequency: 25%, coefficient of variation: 20%, latency statistical distribution: 15%, IOPS: 10%. Then, according to the pre-set scoring rules for each dimension, a score evaluation is generated for each dimension: Dimension Score = Base Score × Weighting Coefficient × Adjustment Factor. The adjustment factor is pre-set based on the complexity of the test environment, workload intensity, and differences in equipment specifications. Finally, a comprehensive consistency score is generated: Comprehensive Score = Σ (Scores of each dimension).

[0084] Furthermore, health and risk prediction analysis includes quantitative assessment of health, risk pattern identification, and remaining life expectancy prediction.

[0085] Among them, the quantitative assessment of health includes:

[0086] 1.1 Extract characteristic parameters, including SMART parameters and performance degradation curves, from the test data. SMART parameters include at least NAND write volume, average wear count, and spare block count.

[0087] 1.2 Input the feature parameters into the pre-trained machine learning regression model for analysis and output a comprehensive health score.

[0088] Specifically, the following feature parameters are systematically extracted from the time-series data warehouse:

[0089] The SMART parameter feature group includes: Durability parameters: total NAND writes (reflecting historical workload), average wear count (measuring the wear leveling status of flash memory cells); Reliability parameters: spare block count (monitoring reliability reserve consumption), bad block growth rate, read error count; Anomaly logging parameters: number of unsafe shutdowns, CRC check error count, end-to-end protection error count.

[0090] The dynamic characteristics of performance degradation include: performance degradation curve characteristics: the slope of performance metrics (IOPS, latency) over time during continuous stress testing; performance recovery characteristics: the efficiency and time required for performance to recover to baseline levels during garbage collection testing; and stability characteristics: consistency metrics such as performance coefficient of variation and jitter frequency.

[0091] The machine learning regression model is constructed using ensemble learning algorithms such as Gradient Boosting Decision Tree (GBDT) or Random Forest. Training data is derived from a large amount of complete lifecycle data of SSDs of the same type as the one being tested, collected from historical tests (including historical data of SMART parameter feature groups and historical data of dynamic performance degradation feature groups). Through expert annotation and actual failure data, training samples are assigned realistic health labels (0-100 points). After model training, the extracted feature parameters are standardized to eliminate dimensional differences before being input into the model to obtain a health score calculated based on multi-dimensional features. Furthermore, the model can simultaneously output the influence weights of each feature on the score, providing interpretability analysis.

[0092] Risk pattern recognition includes:

[0093] 2.1 Real-time statistical analysis of abnormal events during the testing process.

[0094] Specifically, abnormal events include performance abnormalities and health abnormalities. Performance abnormalities include: sudden performance drop (IOPS or bandwidth suddenly drops by more than 50%), sudden latency increase (average latency or P999 latency spikes abnormally), and performance jitter (performance fluctuates drastically within a short period of time). Health abnormalities include: sudden changes in SMART parameters (accelerated decrease in spare block count, rapid increase in error count), abnormal temperature (constantly exceeding the operating temperature threshold), and recovery failure (performance fails to return to normal levels after garbage collection).

[0095] 2.2 Risk warning is carried out by using pre-constructed risk pattern combinations and abnormal events. The risk pattern combinations are formed by mining the potential correlations between abnormal events using association rule learning algorithms.

[0096] It should be noted that in this embodiment, the FP-Growth algorithm is used to mine the correlation between historical abnormal events and discover strong correlation rules in the form of {preconditions} => {consequence events}, such as: {average wear count > 80, spare block count < 10} => {performance drop}, {CRC error surge, temperature > 70℃} => {data read / write error}, {unsafe shutdown > 10 times, write volume close to nominal value} => {device failure}. Then, the support, confidence, and lift of each rule are calculated to ensure statistical significance.

[0097] Specifically, by continuously monitoring the occurrence of abnormal events during the testing process, when a precondition that meets the known risk pattern is detected, an early warning is immediately triggered, and a risk level assessment and specific response recommendations are provided.

[0098] The remaining lifetime prediction includes:

[0099] 3.1 Construct a lifetime prediction feature vector that includes wear characteristics, performance degradation rate, and environmental stress history.

[0100] Specifically, wear history characteristics include: the percentage of current wear relative to nominal lifespan, historical wear rate and its trend, and wear acceleration factors under different workloads; performance degradation characteristics include: the degradation slope of key performance indicators (write speed, read latency), the degree of degradation of performance recovery capability, and the deterioration trend of stability indicators; environmental stress characteristics include: cumulative high-temperature operating time and temperature distribution, frequency and distribution of power outage events, and intensity and pattern characteristics of workload. This embodiment utilizes wear characteristics, performance degradation rate, and environmental stress to construct a feature vector containing multi-dimensional information.

[0101] 3.2 Input the lifetime prediction feature vector into the pre-trained lifetime prediction model and output the lifetime prediction result of the solid-state drive under test under a specific workload. The lifetime prediction model is trained using test data of solid-state drives of the same type as the solid-state drive under test and is used to establish the mapping relationship between workload and lifetime consumption.

[0102] It should be noted that this lifespan prediction model adopts a hybrid architecture based on survival analysis model combined with regression prediction algorithm. It is trained using complete lifespan data of similar historical hard drives to establish a quantitative relationship between workload intensity, mode and lifespan consumption rate.

[0103] In this embodiment, through systematic health and risk prediction analysis, a shift from passive detection to proactive prediction is achieved, providing users with comprehensive reliability assessment and predictive maintenance support, effectively extending equipment lifespan and reducing the risk of unexpected failures.

[0104] Furthermore, in some embodiments, after obtaining the time-series data warehouse, tools such as Grafana can be used to build dynamic dashboards to display the dynamic changes of performance indicators with temperature, capacity, and time in real time, thereby realizing visualized monitoring of the testing process.

[0105] This embodiment dynamically couples multiple stresses, including temperature stress, capacity occupancy stress, power interruption stress, garbage collection-induced stress, and mixed load stress, to construct a highly realistic test environment. This effectively solves the technical problem of traditional testing methods being isolated and unable to reproduce complex real-world application scenarios. This composite stress testing method can simulate various extreme conditions that solid-state drives (SSDs) may encounter in actual use, significantly improving test coverage and realism, and allowing potential defects to be more fully exposed. At the data analysis level, this invention breaks through the limitations of traditional testing that only focuses on average performance indicators. By constructing a time-series data warehouse and an intelligent deep analysis system, it achieves a leap from shallow statistics to in-depth analysis, accurately quantifying the performance stability of SSDs, realizing quantitative health scoring, risk pattern identification, and remaining lifespan prediction, providing a scientific basis for product quality assessment and predictive maintenance.

[0106] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0107] In one embodiment, a composite stress testing device for a solid-state drive (SSD) is provided, which corresponds one-to-one with the composite stress testing method for SSDs described in the above embodiments. For example... Figure 2 As shown, the composite stress testing device for the solid-state drive includes a configuration loading module 11, a preprocessing module 12, a testing module 13, and an analysis module 14.

[0108] The configuration loading module 11 is used to load test configuration parameters and establish a communication connection with the solid-state drive under test that is pre-placed in a preset test environment.

[0109] Preprocessing module 12 is used to perform performance baseline calibration and preprocessing of the solid-state drive under test;

[0110] Test module 13 is used to perform benchmark performance tests and composite stress performance tests on the solid-state drive under test based on a preset test strategy, synchronously collect test data during the test process, and build a time-series data warehouse. The composite stress includes at least two of the following: temperature stress, capacity occupancy stress, power interruption stress, garbage collection induced stress, and mixed load stress, which are dynamically coupled.

[0111] Analysis module 14 is used to perform intelligent deep analysis based on time-series data warehouse. Intelligent deep analysis includes at least one of performance consistency deep analysis and health and risk prediction analysis.

[0112] Optionally, the test module 13 performs a composite stress performance test including a temperature and capacity coupling test, specifically including: increasing the utilization rate of the solid-state drive under test to a preset utilization rate; while maintaining the preset utilization rate, controlling the ambient temperature to change from a first preset temperature to a second preset temperature at a predetermined rate, and continuously applying I / O load during the temperature change process, and recording test data; and continuously applying I / O load for a first preset duration under steady-state conditions of maintaining the preset utilization rate and the third preset temperature, and recording test data.

[0113] Optionally, the test module 13 performs a composite stress performance test, including an intelligent garbage collection stress test, specifically including: a burst write phase, continuously writing data until the available space of the hard drive is lower than a preset threshold or the duration reaches a second preset duration; an idle recycling and monitoring phase, stopping all I / O operations and leaving it idle for a third preset duration, while monitoring the activity of idle garbage collection; and a performance recovery detection phase, performing a short-term performance test and determining whether the performance index of the solid-state drive has recovered to a predetermined percentage of the baseline. If the performance index of the solid-state drive has not recovered to the predetermined percentage of the baseline, the idle recycling and monitoring phase and the performance recovery detection phase are repeated, and the test data is recorded.

[0114] Optionally, the test module 13 performs composite stress performance testing, including mixed load and power interruption coupling testing, specifically including: configuring a concurrent mixed load mode, including database read / write load and log write load; continuously applying mixed load pressure during extended test time; injecting random power interruption events during mixed load operation and recording test data.

[0115] Optionally, the test module 13 performs synchronous collection of test data during the test process and builds a time-series data warehouse, including: using fio to collect performance data, smartctl to collect complete SMART log data, and iostat to collect system-level I / O information data, and timestamp all data; storing the collected data into a time-series database to establish a three-dimensional data warehouse of performance, health, and environmental parameters.

[0116] Optionally, the analysis module 14 performs a deep performance consistency analysis, including: calculating the statistical distribution characteristics of IOPS and latency, where the statistical distribution characteristics of latency include the mean, standard deviation, and 99.9% high quantile latency index; analyzing the coefficient of variation and performance jitter frequency of the performance data, where the coefficient of variation is calculated based on the mean and standard deviation, and the performance jitter frequency is determined based on the number of times the performance data exceeds the stable range per unit time in a continuous time series, and the stable range is set based on the mean and standard deviation; and generating a comprehensive performance consistency evaluation report based on the statistical distribution characteristics of IOPS and latency, the 99.9% high quantile latency index, the coefficient of variation, and the performance jitter frequency.

[0117] Optionally, the analysis module 14 performs health and risk prediction analysis, including quantitative assessment of health, risk pattern identification, and remaining life expectancy prediction.

[0118] The health measurement assessment includes: extracting feature parameters, including SMART parameters and performance degradation curves, from the test data. The SMART parameters include at least NAND write volume, average wear count, and spare block count; inputting the feature parameters into a pre-trained machine learning regression model for analysis, and outputting a comprehensive health score.

[0119] Risk pattern recognition includes: real-time statistical analysis of abnormal events during testing; risk warning using pre-built risk pattern combinations and abnormal events, with the risk pattern combinations formed by mining potential associations between abnormal events using association rule learning algorithms;

[0120] The remaining lifetime prediction includes: constructing a lifetime prediction feature vector containing wear characteristics, performance degradation rate, and environmental stress history; inputting the lifetime prediction feature vector into a pre-trained lifetime prediction model, and outputting the lifetime prediction result of the solid-state drive under test under a specific workload. The lifetime prediction model is trained using test data of solid-state drives of the same type as the solid-state drive under test, and is used to establish the mapping relationship between workload and lifetime consumption.

[0121] Specific limitations regarding the composite stress testing device for solid-state drives (SSDs) can be found in the limitations of the composite stress testing method for SSDs mentioned above, and will not be repeated here. Each module in the aforementioned composite stress testing device for SSDs can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in the computer device in hardware form, or stored in the memory of the computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0122] In one embodiment, a computer device is provided, the internal structure of which can be shown in the following diagram. Figure 3 As shown. The computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used to communicate with external clients via a network connection. When the computer program is executed by the processor, it performs the following steps:

[0123] Load test configuration parameters and establish a communication connection with the solid-state drive under test that has been placed in the preset test environment;

[0124] Perform performance baseline calibration and pre-testing processing on the solid-state drive under test;

[0125] Based on the preset test strategy, benchmark performance test and composite stress performance test are performed on the solid-state drive under test. Test data during the test process are collected simultaneously, and a time-series data warehouse is built. The composite stress includes at least two of the following: temperature stress, capacity occupancy stress, power interruption stress, garbage collection induced stress, and mixed load stress, which are dynamically coupled.

[0126] Based on a time-series data warehouse, intelligent deep analysis is performed, including at least one of performance consistency deep analysis and health and risk prediction analysis.

[0127] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0128] Load test configuration parameters and establish a communication connection with the solid-state drive under test that has been placed in the preset test environment;

[0129] Perform performance baseline calibration and pre-testing processing on the solid-state drive under test;

[0130] Based on the preset test strategy, benchmark performance test and composite stress performance test are performed on the solid-state drive under test. Test data during the test process are collected simultaneously, and a time-series data warehouse is built. The composite stress includes at least two of the following: temperature stress, capacity occupancy stress, power interruption stress, garbage collection induced stress, and mixed load stress, which are dynamically coupled.

[0131] Based on a time-series data warehouse, intelligent deep analysis is performed, including at least one of performance consistency deep analysis and health and risk prediction analysis.

[0132] It should be noted that the functions or steps that can be implemented by the computer-readable storage medium or computer device described above can be referred to the relevant descriptions in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0133] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0134] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0135] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A composite stress testing method for solid-state drives, characterized in that, include: Load test configuration parameters and establish a communication connection with the solid-state drive under test that has been placed in the preset test environment; The solid-state drive under test was subjected to performance baseline calibration and pre-test processing. Based on a preset testing strategy, the solid-state drive under test is subjected to benchmark performance testing and composite stress performance testing. Test data is collected synchronously during the testing process, and a time-series data warehouse is constructed. The composite stress includes at least two of the following: temperature stress, capacity occupancy stress, power interruption stress, garbage collection induced stress, and mixed load stress, which are dynamically coupled. Based on the time-series data warehouse, intelligent deep analysis is performed, which includes at least one of performance consistency deep analysis and health and risk prediction analysis. The composite stress performance test includes a smart waste recycling stress test, specifically including: During the burst write phase, data is continuously written until the available disk space is lower than a preset threshold or the duration reaches the second preset duration. During the idle garbage collection and monitoring phase, all I / O operations are stopped and the system is left idle for a third preset duration, while the activity of idle garbage collection is monitored. In the performance recovery detection phase, a short-term performance test is performed to determine whether the performance index of the solid-state drive has recovered to a predetermined percentage of the baseline. If the performance index of the solid-state drive has not recovered to the predetermined percentage of the baseline, the idle reclamation and monitoring phase and the performance recovery detection phase are repeated, and the test data is recorded. The performance consistency in-depth analysis includes: Calculate the statistical distribution characteristics of IOPS and latency, wherein the statistical distribution characteristics of latency include the mean, standard deviation and 99.9th percentile latency index; The coefficient of variation and performance fluctuation frequency of the performance data are analyzed. The coefficient of variation is calculated based on the mean and the standard deviation. The performance fluctuation frequency is determined based on the number of times the performance data exceeds the stable range per unit time in a continuous time series. The stable range is set based on the mean and the standard deviation. A comprehensive performance consistency evaluation report is generated based on the IOPS, the statistical distribution characteristics of the latency, the 99.9% high quantile latency index, the coefficient of variation, and the performance jitter frequency.

2. The composite stress testing method for solid-state drives according to claim 1, characterized in that, The composite stress performance test includes temperature and capacity coupling tests, specifically: Increase the utilization rate of the solid-state drive under test to a preset utilization rate; While maintaining the preset occupancy rate, the ambient temperature is controlled to change from a first preset temperature to a second preset temperature at a predetermined rate, and I / O load is continuously applied during the temperature change process, and test data is recorded. While maintaining the preset occupancy rate and the third preset temperature in a steady state, I / O load for a first preset duration is continuously applied, and test data is recorded.

3. The composite stress testing method for solid-state drives according to claim 1, characterized in that, The composite stress performance test includes mixed load and power interruption coupling test, specifically including: Configure a concurrent mixed load mode, including database read / write load and log write load; Apply mixed load pressure continuously for an extended test period; Inject random power interruption events during mixed load operation and record test data.

4. The composite stress testing method for solid-state drives according to claim 1, characterized in that, The synchronous acquisition of test data during the testing process and the construction of a time-series data warehouse include: We use fio to collect performance data, smartctl to collect complete SMART log data, and iostat to collect system-level I / O information data, and timestamp all the data. The collected data is stored in a time-series database to establish a three-dimensional data warehouse for performance, health, and environmental parameters.

5. The composite stress testing method for solid-state drives according to claim 1, characterized in that, The health and risk prediction analysis includes quantitative assessment of health, risk pattern identification, and remaining life expectancy prediction. The health quantification assessment includes: From the test data, feature parameters including SMART parameters and performance degradation curves are extracted. The SMART parameters include at least NAND write volume, average wear count, and spare block count. The feature parameters are input into a pre-trained machine learning regression model for analysis, and a comprehensive health score is output. The risk pattern identification includes: Real-time statistical analysis of abnormal events during the testing process; Risk warning is performed by using a pre-constructed combination of risk patterns and the abnormal events. The combination of risk patterns is formed by mining the potential associations between abnormal events using an association rule learning algorithm. The remaining lifetime prediction includes: Construct a lifetime prediction feature vector that includes wear characteristics, performance degradation rate, and environmental stress history; The lifetime prediction feature vector is input into a pre-trained lifetime prediction model, which outputs the lifetime prediction result of the solid-state drive under test under a specific workload. The lifetime prediction model is trained using test data of solid-state drives of the same type as the solid-state drive under test and is used to establish a mapping relationship between workload and lifetime consumption.

6. A composite stress testing device for solid-state drives, characterized in that, include: The configuration loading module is used to load test configuration parameters and establish a communication connection with the solid-state drive under test that has been placed in a preset test environment. The preprocessing module is used to perform performance baseline calibration and preprocessing on the solid-state drive under test. The testing module is used to perform benchmark performance testing and composite stress performance testing on the solid-state drive under test based on a preset testing strategy, synchronously collect test data during the testing process, and build a time-series data warehouse. The composite stress includes at least two of the following: temperature stress, capacity occupancy stress, power interruption stress, garbage collection induced stress, and mixed load stress, which are dynamically coupled. The analysis module is used to perform intelligent deep analysis based on the time-series data warehouse, the intelligent deep analysis including at least one of performance consistency deep analysis and health and risk prediction analysis; The composite stress performance test includes a smart waste recycling stress test, specifically including: During the burst write phase, data is continuously written until the available disk space is lower than a preset threshold or the duration reaches the second preset duration. During the idle garbage collection and monitoring phase, all I / O operations are stopped and the system is left idle for a third preset duration, while the activity of idle garbage collection is monitored. In the performance recovery detection phase, a short-term performance test is performed to determine whether the performance index of the solid-state drive has recovered to a predetermined percentage of the baseline. If the performance index of the solid-state drive has not recovered to the predetermined percentage of the baseline, the idle reclamation and monitoring phase and the performance recovery detection phase are repeated, and the test data is recorded. The performance consistency in-depth analysis includes: Calculate the statistical distribution characteristics of IOPS and latency, wherein the statistical distribution characteristics of latency include the mean, standard deviation and 99.9th percentile latency index; The coefficient of variation and performance fluctuation frequency of the performance data are analyzed. The coefficient of variation is calculated based on the mean and the standard deviation. The performance fluctuation frequency is determined based on the number of times the performance data exceeds the stable range per unit time in a continuous time series. The stable range is set based on the mean and the standard deviation. A comprehensive performance consistency evaluation report is generated based on the IOPS, the statistical distribution characteristics of the latency, the 99.9% high quantile latency index, the coefficient of variation, and the performance jitter frequency.

7. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the composite stress testing method for the solid-state drive as described in any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the composite stress testing method for the solid-state drive as described in any one of claims 1 to 5.

Images