A qtl1999-based ssd energy consumption test intelligent management system

Abstract

The application discloses a kind of SSD energy consumption test intelligent management systems based on QTL1999, belong to solid state disk technical field.Layered modular architecture of test execution layer, data processing layer, storage layer and application layer is used.System realizes the automatic synchronization collection of energy consumption and performance data through pyquarch interface, time difference is controlled at ≤1 millisecond, ensure data accurate alignment;Utilize the definition test flow of tag configuration file, support the flexible configuration of pre-processing condition and test model;Establish the data management mechanism with SN number, test timestamp and model label as triple association mark, realize the unified integration and batch retrieval of multiple models, multiple SN test data.Application layer integrates multidimensional manual comparison module and AI interaction double mode of large model, support natural language query and automatic report generation.The application significantly reduces the data analysis threshold, improves the value conversion efficiency of test data and system expansibility.

Description

Technical Field

[0001] This invention relates to the field of solid-state drive technology, and in particular to an intelligent management system for SSD power consumption testing based on QTL1999. Background Technology

[0002] With the rapid development of cloud computing, big data, and mobile terminal technologies, SSDs have been widely used in servers, data centers, and consumer electronics due to their advantages such as high IOPS (input / output operations per second), low latency, and strong shock resistance. Against the backdrop of large-scale data center deployments and increasing battery life requirements for terminal devices, energy efficiency ratio (i.e., a synergistic indicator of performance and power consumption, such as IOPS / watt) has become a core parameter for evaluating the competitiveness of SSD products, directly impacting device operating costs, battery life, and stability.

[0003] In the entire process of SSD power consumption test data management, due to limitations of existing technical solutions and process design, there are several bottlenecks in efficiency and accuracy at each stage, specifically as follows: In the SSD testing stage, the testing process based on QTL1999 and existing supporting tools requires manual intervention to complete data acquisition and connection, resulting in insufficient time axis alignment accuracy of I / O performance and power consumption data, making it difficult to meet the needs of refined test analysis; In the data retrieval stage, the existing system only supports a single retrieval mode, which cannot adapt to the traditional operating habits of testers, nor can it cope with the need for efficient retrieval across dimensions and multiple conditions, significantly reducing data acquisition efficiency; In terms of data utilization, the existing system requires high professional analysis capabilities, and non-professionals cannot quickly obtain customized analysis results, and the report generation process is cumbersome and inefficient, restricting the value conversion of test data; In the field of data management, the test data of multiple models and multiple SNs of SSDs lacks a unified integration and management mechanism, and is scattered in different files or systems, making it impossible to quickly conduct horizontal (multiple models / multiple SNs) and vertical (same model, different batches / SNs) comparative analysis in batches, affecting the comprehensiveness and accuracy of test conclusions.

[0004] Currently, enterprises face complex scenarios in SSD power consumption testing, involving multiple models (such as NVMe and SATA interfaces) and multiple serial numbers (different individuals of the same model). During the R&D phase, performance bottlenecks of different models need to be compared; during production quality control, consistency between different serial numbers of the same model needs to be screened; and during the selection phase, the differences in power efficiency among multiple models need to be evaluated. However, existing testing technologies have systemic shortcomings, particularly inefficient data collection, management, retrieval, and utilization, making it difficult to meet the needs of refined testing.

[0005] Existing technical solutions and their defects 1. Data Acquisition Stage: Insufficient Accuracy and Efficiency Due to Manual Intervention: Existing technical solutions employ a separate architecture of "QTL1999 power consumption testing tool + independent performance testing tool (such as fio)". This solution accurately acquires SSD voltage, current, and power data through QTL1999, while relying on manual recording of IOPS, latency, and other indicators output by the performance testing tool. However, due to the lack of an automated synchronization mechanism, manual intervention is required to start the test, record data, and follow up, resulting in the following drawbacks: Low time axis alignment accuracy: Manual operation introduces second-level time differences, making it impossible to achieve millisecond-level alignment between power consumption and performance data, and making it difficult to support "performance-power consumption" collaborative analysis; Poor efficiency and reliability: Manual recording is prone to data omissions or errors, resulting in low acquisition efficiency, and human error affects data reliability.

[0006] 2. Data Management: Distributed storage hinders batch comparison analysis. Another similar solution uses a "distributed storage + manual aggregation" model, storing test data for different models and serial numbers (SNs) in scattered local files or isolated databases. Subsequent manual extraction and organization of data for horizontal and vertical comparisons are required. Its drawbacks include: lack of a management mechanism: data lacks unified association identifiers, making it impossible to quickly retrieve data for multiple models or SNs in batches; low comparison efficiency: manual aggregation is time-consuming and labor-intensive; for example, screening for consistency of the same model for 100 SNs takes approximately 2 hours, and omissions can easily affect the comprehensiveness of the analysis.

[0007] 3. Data Retrieval Stage: The single-mode adaptability is insufficient. Existing retrieval tools (such as SSD-Z) only support queries by device model, serial number, or a single metric, failing to consider the actual needs of testers for "cross-dimensional combined queries." Its limitations are reflected in: poor retrieval flexibility: unable to adapt to multi-condition combined queries (such as "comparing the energy consumption ratio of models A and B when randomly written in 4K"); and low operational efficiency: in complex scenarios, users need to manually filter data, significantly reducing data acquisition speed.

[0008] 4. Data Utilization: High professional barriers limit value conversion. Existing systems require advanced data analysis skills, making it difficult for non-professionals to quickly obtain customized conclusions. Report generation relies on manual plotting and organization, a cumbersome process that results in insufficient value conversion of test data.

[0009] This invention aims to solve the following problems: 1. Test data acquisition: Existing processes require manual intervention, and the timeline alignment accuracy of I / O performance and energy consumption data is insufficient. Automated synchronous acquisition is needed to improve efficiency and accuracy. 2. Data retrieval and query: Existing systems have limited retrieval methods and low efficiency in cross-dimensional queries. A dual-mode system of "manual multi-dimensional comparison + natural language AI interaction" needs to be built to balance user habits and efficiency. 3. Data integration and management: Test data from multiple models and multiple serial numbers are scattered and lack a unified mechanism, making it difficult to quickly conduct batch horizontal and vertical comparative analysis. A unified integration and management mechanism needs to be established. 4. Data utilization and output: Professional analysis has a high barrier to entry, and non-professionals cannot quickly obtain customized results. The report generation process is cumbersome and inefficient. The utilization barrier needs to be lowered and report generation optimized. Summary of the Invention

[0010] The purpose of this invention is to overcome the shortcomings of the prior art and provide an intelligent management system for SSD power consumption testing based on QTL1999.

[0011] The objective of this invention is achieved through the following technical solution: an intelligent management system for SSD power consumption testing based on QTL1999, which adopts a layered modular architecture, including a test execution layer, which is connected to a data processing layer, which is connected to a storage layer, which is connected to an application layer, and the application layer is connected to a user terminal. The test execution layer is based on QTL1999 devices and a Linux test platform. It builds standardized scripts through the pyquarch interface, configures business models using configuration files, and automatically and synchronously collects SSD energy consumption and performance data. The data processing layer is used to integrate multi-source test data and preprocess it through a unified data association rule. The data association rule uses the SN number, test timestamp, and model label as unique identifiers. The storage layer is used to store test equipment information, test records of the test disk at various time points, power consumption data and performance data corresponding to each test record; The application layer has data retrieval, comparison analysis, and report generation functions, including a multi-dimensional manual comparison module and an interactive AI question-and-answer module. The multi-dimensional manual comparison module is used to perform batch retrieval, horizontal and vertical comparison, and consistency screening based on model, SN data, test model, and time conditions, and automatically generate visual charts. The interactive AI question-and-answer module parses user intent through natural language queries, batch retrieves and integrates underlying multi-source data, and finally outputs structured analysis results, realizing integrated management of SSD energy consumption test data.

[0012] Preferably, the configuration file adopts a lightweight YAML format, and defines the test process through tag-based encapsulation and modular configuration, replacing manual script writing; the tag-based encapsulation includes preprocessing condition tagging and test model tagging; the preprocessing condition tagging encapsulates the pre-operations before testing; the test model tagging encapsulates the performance-energy consumption test scenarios of different IO modes.

[0013] Preferably, the data association rules use a triple unique identifier as an index to establish a hierarchical association relationship of "SSD basic information - test environment information - original collected data", realizing hierarchical data binding traceability and batch retrieval; the triple unique identifier includes: The SSD basic information is associated with the SN number as a unique identifier, which includes the model, capacity and firmware version; The test environment information is associated with the SN number and the test timestamp as unique identifiers. The test environment information includes the test configuration, host CPU model, memory capacity, and operating system version. The original acquired data is associated with the serial number (SN), test timestamp, and model label as unique identifiers. The original acquired data includes voltage, current, power, IOPS, bandwidth, and latency metrics with different model labels.

[0014] Preferably, the horizontal and vertical comparisons of the multi-dimensional manual comparison module include: Horizontal comparison is used to compare the energy efficiency differences of SSDs of different models or serial numbers under the same test model, and automatically generate bar charts or line charts. Longitudinal comparison is used to analyze the consistency of SSDs of the same model in different batches or at different test times, and automatically generates data comparison tables and anomaly annotations.

[0015] Preferably, the interactive AI question-answering module is based on an open-source large model and customized for SSD power consumption testing, including: Configure domain terminology dictionaries and semantic parsing rules to improve the recognition accuracy of professional queries for SSD power consumption testing; Convert natural language queries into standardized database commands; Configure multi-turn dialogue context awareness function so that users do not need to repeat basic conditions when asking follow-up questions; After responding to a query, the interactive AI question-answering module automatically outputs visual charts and structured conclusions.

[0016] Preferably, the open-source large model is DeepSeek-V3.1.

[0017] The beneficial effects of this invention are: 1) Dual improvement in data acquisition accuracy and efficiency: This invention uses the PyQuarch interface to achieve automated synchronous acquisition of energy consumption and performance data, and eliminates manual intervention by standardizing the test process through configuration files; at the same time, it strictly controls the acquisition trigger time difference to ≤1ms; on the one hand, it avoids the time difference and error of manual operation, and improves the alignment accuracy of the two types of data time axes from "second level" to "millisecond level", meeting the needs of refined collaborative analysis; on the other hand, there is no need to manually record and connect data, the acquisition efficiency is improved by more than 80%, and the data reliability is significantly improved.

[0018] 2) Exponential optimization of multi-dimensional data comparison efficiency: This invention establishes a unified data integration and storage mechanism, realizing centralized management of multi-model and multi-SN data through triple association identification, and supporting batch retrieval; at the same time, it designs an automated chart generation function to replace manual summarization and drawing; testers do not need to manually screen and organize scattered data, and the horizontal (multiple models) and vertical (multiple batches of the same model) comparison time is shortened from "hours" to "minutes"; for example, the consistency screening of 100 SNs can be compressed from 2 hours to 5 minutes, which greatly improves the comprehensiveness of test conclusions and decision-making speed.

[0019] 3) Lowered data analysis threshold and more comprehensive data value conversion: This invention achieves direct output of natural language to structured results through AI domain adaptation. At the same time, the automated report generation function integrates analysis results with visualization charts, which can be operated without professional data analysis skills. Non-professionals can quickly obtain customized analysis results (such as "the energy consumption fluctuation trend of a certain model of SSD under different loads"), reducing the data analysis threshold from "professional level" to "entry level". The report generation time is shortened from "hours" to "minutes", enabling the value conversion of test data to cover more user groups and improving data utilization.

[0020] 4) This invention adopts a layered modular architecture, with each layer connected through standardized interfaces; the configuration file supports adding new test models and preprocessing conditions without modifying the core code; the storage layer is compatible with test data formats of multiple types of SSDs (NVMe, SATA, etc.); the system can flexibly add SSD models and test scenarios without reconstructing the architecture. Attached Figure Description

[0021] Figure 1 This is a system architecture diagram of the present invention. Detailed Implementation

[0022] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. 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.

[0023] First, some terms in this invention will be explained: 1. SSD (Solid State Drive): A solid-state drive is a storage device that uses flash memory chips as the storage medium. It has advantages such as high IOPS, low latency, and strong shock resistance. It is divided into interface types such as NVMe and SATA.

[0024] 2. NVMe (Non-Volatile Memory Express): A non-volatile memory host controller interface specification designed specifically for PCIe interface flash memory devices, with a transfer rate much higher than that of the SATA interface.

[0025] 3. SATA (Serial Advanced Technology Attachment): A standard interface for connecting storage devices to a host, widely used in traditional hard disk drives and some SSDs.

[0026] 4. IOPS (Input / Output Operations Per Second): The number of input / output operations per second, a core metric for measuring the random read / write performance of storage devices.

[0027] 5. Power efficiency ratio: The performance that an SSD can provide per unit power consumption (W) (such as IOPS / W) is the core parameter for evaluating the energy efficiency synergy of an SSD.

[0028] 6. QTL1999: A high-precision power consumption test device that supports real-time acquisition of voltage, current and power, and is widely used in power consumption testing scenarios for electronic equipment.

[0029] 7. pyquarch: "A third-party Python component / library that provides a low-level communication interface with power analyzers such as the QTL1999. It supports sending acquisition commands via Python scripts, reading voltage / current / power data in real time, and returning structured data with timestamps. It is a key technology for achieving synchronous acquisition of energy consumption and performance data." 8. SMART (Self-Monitoring, Analysis and Reporting Technology): Self-monitoring, analysis and reporting technology, the SSD's built-in status monitoring function, can provide information such as remaining lifespan and number of bad blocks.

[0030] 9. IDENTIFY Information: The SSD's inherent identification information, including core parameters such as model, serial number, capacity, firmware version, and interface type.

[0031] 10. Layered Modular Architecture: The system is divided into a test execution layer, a data processing layer, a storage layer, and an application layer according to their functions. Each layer is designed independently and interacts through standardized interfaces to improve the system's scalability and maintainability.

[0032] 11. Domain-specific semantic understanding: For professional terminology and query scenarios in the field of SSD power consumption testing, the AI ​​model is fine-tuned and rules are optimized to improve the accuracy of semantic recognition.

[0033] 12. Multi-turn dialogue context awareness: The AI ​​model has the ability to remember the content of previous dialogues, allowing users to gradually deepen their query needs in continuous interactions without having to repeatedly input basic conditions.

[0034] 13. Horizontal and vertical comparison: Horizontal comparison refers to the performance / energy consumption difference analysis between multiple models and multiple serial numbers; vertical comparison refers to the consistency analysis of different batches and different test times of the same model.

[0035] 14. pywinauto: A Python automation library that can simulate mouse clicks, keyboard input, and other operations in Windows systems to automate the control of graphical user interface (GUI) programs.

[0036] 15. Prompt Engineering: Guide AI models to understand user intent and output results that meet requirements by designing precise text prompts, without modifying the underlying code of the model.

[0037] 16. fio (Flexible I / O Tester): An open-source IO performance testing tool that supports multiple IO modes (such as sequential read / write, random read / write, and mixed read / write) and can collect key performance indicators such as IOPS, latency, and bandwidth. It is a core software tool for SSD performance testing.

[0038] 17. MySQL: An open-source relational database management system with high stability, ease of use, and good compatibility. It is used to store pre-processed SSD energy consumption and performance data, providing data support for data querying and report generation.

[0039] 18. Vue 3.0: A popular front-end development framework built on JavaScript. It can quickly build interactive and user-friendly interfaces for building front-end systems. It supports users to manually select test conditions, view visual charts, and trigger the generation of PDF reports.

[0040] 19. Latency: refers to the total time from initiating an I / O request to its completion. It is a key indicator for measuring the response speed of storage devices, and the unit is usually milliseconds (ms). In testing, ensuring that the time axis of energy consumption and performance data is aligned is the core of controlling the deviation of this indicator.

[0041] 20. iodepth (I / O queue depth): I / O concurrency in single-process / single-thread scenarios, is a key indicator representing the intensity of SSD IO processing capacity usage by a single task.

[0042] 21. numjobs (number of test jobs): A core configuration parameter of fio (Flexible I / O Tester), referring to the total number of independent test jobs (processes or threads) started simultaneously during the test.

[0043] See Figure 1 The present invention provides a technical solution: an intelligent management system for SSD power consumption testing based on QTL1999, which adopts a layered modular architecture, including a test execution layer, the test execution layer connected to a data processing layer, the data processing layer connected to a storage layer, the storage layer connected to an application layer, and the application layer connected to a user terminal; The test execution layer is based on QTL1999 devices and a Linux test platform. It builds standardized scripts through the pyquarch interface, configures business models using configuration files, and automatically and synchronously collects SSD energy consumption and performance data. The data processing layer is used to integrate multi-source test data and preprocess it through a unified data association rule. The data association rule uses the SN number, test timestamp, and model label as unique identifiers. The storage layer is used to store test equipment information, test records of the test disk at various time points, power consumption data and performance data corresponding to each test record; The application layer has data retrieval, comparison analysis, and report generation functions, including a multi-dimensional manual comparison module and an interactive AI question-and-answer module. The multi-dimensional manual comparison module is used to perform batch retrieval, horizontal and vertical comparison, and consistency screening based on model, SN data, test model, and time conditions, and automatically generate visual charts. The interactive AI question-and-answer module parses user intent through natural language queries, batch retrieves and integrates underlying multi-source data, and finally outputs structured analysis results, realizing integrated management of SSD energy consumption test data.

[0044] In this embodiment, the invention adopts a "layered modular" architecture design, with the core consisting of four layers: a test execution layer, a data processing layer, a storage layer, and an application layer. Each layer achieves data flow through standardized interfaces, supporting energy efficiency testing and integrated data management for multiple SSD models and multiple serial numbers. The overall architecture is as follows: Figure 1 As shown. Test Execution Layer: Based on QTL1999 devices and a Linux testing platform, and utilizing the open-source PyQuarch interface, standardized scripts are built to enable synchronous collection of power consumption and performance data from multiple SSD models through configuration file-based business model configuration. Data Processing Layer: Integrates test data from multiple sources (basic information, environmental information, and raw collected data), establishes unified data association rules, and provides a standardized data foundation for subsequent storage and batch comparison. After data preprocessing, the preprocessed data is uploaded via an upload interface. Storage Layer: Used to store test device information (including test environment information, SMART and IDENTIFY information of the test disk), test records at various time points of the test disk, and detailed power consumption and performance data corresponding to each record. Application Layer: Provides a dual-mode database retrieval system of "manual multi-dimensional comparison + AI interactive query," combined with visualization and customized report export to meet the batch data analysis needs of different users.

[0045] The following is a detailed explanation of the core module's functions and principles: 1. An automated execution module based on QTL1999 pyquarch. Module function: Configure business models through configuration files to synchronously collect power consumption and performance data of multiple SSD models.

[0046] Key Technology Description: Configuration File Design: The configuration file is the core input carrier of the automated execution module. It adopts the lightweight YAML format and its core goal is to replace the traditional manual writing of repetitive test scripts with "tag-based encapsulation + modular configuration," enabling rapid definition and flexible adjustment of test logic. The specific design is as follows: Tag-based encapsulation rules: The script pre-encapsulates the preprocessing conditions and test models that are frequently reused in SSD power consumption test scenarios with standardized tags. The tags are mapped one-to-one with the underlying execution logic. Users do not need to write complex test execution code. They only need to call the tags to complete the test process configuration.

[0047] a. Preprocessing condition tagging: For pre-test operations such as SSD initialization and load preprocessing, core tags are encapsulated as shown in Table 1 (tag name - corresponding execution logic): Table 1

[0048] b. Test Model Tagging: For performance-energy consumption test scenarios with different IO modes, core tags are encapsulated as shown in Table 2 (tag name - corresponding execution logic): Table 2

[0049] Configuration file structure and usage: The configuration file supports flexible adaptation to multiple scenarios. A typical example of test_config.yaml is shown below, demonstrating the initial 4K random write energy efficiency ratio under PS1 testing: test_flow: - PRE_FORMAT_ALL # Step 1: Perform a full safe wipe; - PRE_PM_SET_PS1 # Step 2: Set the power management status to PS1; - MODEL_4K_RND_W_Q32_T1 # Step 3: Measure the initial state 4K random write energy efficiency (iodepth: 32, numjobs: 1).

[0050] Synchronous data acquisition: By leveraging the open-source Pyquarch interface, the acquisition trigger time difference between energy consumption data (voltage / current / power) and performance data (IOPS / latency) is ensured to be ≤1ms, achieving precise time axis alignment.

[0051] 2. Data preprocessing module: This module provides a standardized data foundation for subsequent storage and batch comparison. Data integration rules are shown in Table 3. Table 3

[0052] Data Upload / Import Unit: Method 1: After data preprocessing, the script uploads the preprocessed data to the database via the upload interface. Method 2: After data preprocessing, click the upload button on the front-end page to upload the preprocessed data to the database.

[0053] 3. Data storage module: This module is used to store basic SSD information, "test + environment information", and a database of raw collected data.

[0054] 4. Multi-dimensional manual comparison module: This module supports batch retrieval, horizontal and vertical comparison, and consistency screening of data from multiple models and multiple serial numbers. It also provides intuitive display of comparison results through automated chart generation.

[0055] Core Function Description: Batch Search Flexibility: Supports multi-select filtering and searching by model, serial number (SN), model, and time; Automated Comparison Charts: No manual plotting is required; the system automatically generates various types of charts, such as: Horizontal Comparison: Supports horizontal comparison between multiple models and multiple SNs. Vertical Comparison: Supports vertical comparison between different models of the same model and SN. Consistency Screening: Supports automatic and rapid screening of consistency differences between models in different test data of the same model. All of the above comparison methods can automatically generate corresponding data comparison tables and interactive bar charts or line charts.

[0056] 5. Interactive AI Question Answering Module: This module is based on the open-source DeepSeek-V3.1 model and has been customized and optimized for the SSD power consumption testing field. It supports semantic queries using natural language and can automatically parse key elements such as product model, serial number, test conditions, and analysis indicators from user input. It batch retrieves and integrates underlying multi-source data, ultimately outputting structured conclusions and multi-dimensional visualization charts. This module significantly lowers the barrier to professional data querying and analysis, improving interaction efficiency and result accuracy. Key Technology Description: Domain-Customized Semantic Understanding Technology: Based on the DeepSeek-V3.1 open-source model, it is optimized by combining a professional terminology dictionary and semantic parsing rules for the SSD power consumption testing field, improving the accuracy of recognizing professional natural language query intent and reducing semantic bias of general models in professional scenarios. Intelligent Query Parsing and Instruction Conversion Technology: It automatically extracts key parameters from natural language queries and accurately converts them into standardized database query instructions, achieving accurate understanding and response to users' complex analysis needs. Multi-turn dialogue context-aware technology: Supports context-based continuous questioning and interactive memory, allowing users to deepen their analysis step by step in multiple rounds of dialogue without having to repeatedly enter basic query conditions, thus improving the continuity and efficiency of interaction in complex analysis scenarios.

[0057] 6. Visualization and Report Export Module: This module supports exporting the output of the "Multi-Dimensional Manual Comparison Module" and the "Interactive AI Question Answering Module" to generate offline data analysis reports.

[0058] In some embodiments, the configuration file adopts a lightweight YAML format, and defines the test process through tag-based encapsulation and modular configuration, replacing manual script writing; the tag-based encapsulation includes preprocessing condition tagging and test model tagging; the preprocessing condition tagging encapsulates the pre-operations before testing; the test model tagging encapsulates the performance-energy consumption test scenarios of different IO modes.

[0059] In some embodiments, the data association rules use a triple unique identifier as an index to establish a hierarchical association relationship of "SSD basic information - test environment information - original collected data", realizing hierarchical data binding traceability and batch retrieval; the triple unique identifier includes: The SSD basic information is associated with the SN number as a unique identifier, which includes the model, capacity and firmware version; The test environment information is associated with the SN number and the test timestamp as unique identifiers. The test environment information includes the test configuration, host CPU model, memory capacity, and operating system version. The original acquired data is associated with the serial number (SN), test timestamp, and model label as unique identifiers. The original acquired data includes voltage, current, power, IOPS, bandwidth, and latency metrics with different model labels.

[0060] In some embodiments, the horizontal and vertical comparison of the multi-dimensional manual comparison module includes: Horizontal comparison is used to compare the energy efficiency differences of SSDs of different models or serial numbers under the same test model, and automatically generate bar charts or line charts. Longitudinal comparison is used to analyze the consistency of SSDs of the same model in different batches or at different test times, and automatically generates data comparison tables and anomaly annotations.

[0061] In some embodiments, the interactive AI question-answering module is customized and optimized for SSD power consumption testing based on an open-source large model, including: Configure domain terminology dictionaries and semantic parsing rules to improve the recognition accuracy of professional queries for SSD power consumption testing; Convert natural language queries into standardized database commands; Configure multi-turn dialogue context awareness function so that users do not need to repeat basic conditions when asking follow-up questions; After responding to a query, the interactive AI question-answering module automatically outputs visual charts and structured conclusions.

[0062] In some embodiments, the open-source large model is DeepSeek-V3.1.

[0063] The following is an example: 1. Implementation Environment: Server-side: Windows server with a MySQL database and a front-end system developed based on Vue 3.0, responsible for overall service operation, test data storage, user interaction, and report generation; Test-side: Ordinary Linux PC with a QTL1999 power analyzer to conduct performance and energy consumption tests on SSDs (3 models with a total of 50 SNs, including NVMe / SATA interfaces); The test-side scripts are developed based on Python 3.9 and rely on components such as pyquarch, responsible for data acquisition, preprocessing, and uploading.

[0064] 2. Implementation steps: 1. Configuration file creation: Write the test_config.yaml configuration file on the Linux test PC, and define preprocessing conditions such as 1M sequential write and 4K random write, and test models such as 4K random mixed read and write and 1M sequential mixed read and write. 2. Data Acquisition and Upload: The test script is launched to synchronously collect energy consumption data such as voltage / current / power of QTL1999 and performance data such as IOPS / latency / bandwidth of fio via the pyquarch interface, ensuring accurate alignment of the timelines of the two types of data. The script automatically filters outliers and preprocesses the data before uploading it to a MySQL database on a Windows server. 3. Data Query and Report Generation: Users can obtain analysis results or generate reports in two ways: ① Manually filter conditions such as SSD model, SN, and test model, and the system automatically generates visual charts and data tables; ② Input natural language queries (such as "compare the 1M sequential energy consumption ratio of models A and C"), and the AI ​​will parse and output structured conclusions; Select modules such as data overview, comparison charts, and anomaly analysis, and the system will automatically generate an HTML report containing horizontal and vertical comparison charts of 50 SNs, energy consumption ratio ranking, and anomaly data annotations. The overall export time is ≤25 seconds. According to tests, the time for screening the consistency of 100 SNs of the same model has been reduced from the traditional 2 hours to 4.8 minutes, an efficiency improvement of 25 times.

[0065] The above description is merely a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and environments, and can be altered within the scope of the concept described herein through the above teachings or related technologies or knowledge. Modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the protection scope of the appended claims.

Claims

1. An intelligent management system for SSD power consumption testing based on QTL1999, characterized in that: The system adopts a layered and modular architecture, including a test execution layer, which is connected to a data processing layer, which is connected to a storage layer, which is connected to an application layer, and the application layer is connected to a user terminal. The test execution layer is based on QTL1999 devices and a Linux test platform. It builds standardized scripts through the pyquarch interface, configures business models using configuration files, and automatically and synchronously collects SSD energy consumption and performance data. The data processing layer is used to integrate multi-source test data and preprocess it through a unified data association rule. The data association rule uses the SN number, test timestamp, and model label as unique identifiers. The storage layer is used to store test equipment information, test records of the test disk at various time points, power consumption data and performance data corresponding to each test record; The application layer has data retrieval, comparison analysis, and report generation functions, including a multi-dimensional manual comparison module and an interactive AI question-and-answer module. The multi-dimensional manual comparison module is used to perform batch retrieval, horizontal and vertical comparison, and consistency screening based on model, SN data, test model, and time conditions, and automatically generate visual charts. The interactive AI question-and-answer module parses user intent through natural language queries, batch retrieves and integrates underlying multi-source data, and finally outputs structured analysis results, realizing integrated management of SSD energy consumption test data.

2. The intelligent management system for SSD power consumption testing based on QTL1999 according to claim 1, characterized in that: The configuration file uses a lightweight YAML format and defines the test process through tag-based encapsulation and modular configuration, replacing manual script writing. The tag-based encapsulation includes preprocessing condition tagging and test model tagging. The preprocessing condition tagging encapsulates the pre-operations before testing, and the test model tagging encapsulates the performance-energy consumption test scenarios for different IO modes.

3. The intelligent management system for SSD power consumption testing based on QTL1999 according to claim 1, characterized in that: The data association rules use a triple unique identifier as an index to establish a hierarchical association relationship of "SSD basic information - test environment information - raw collected data", thereby realizing data hierarchical binding traceability and batch retrieval. The triple unique identifier includes: The SSD basic information is associated with the SN number as a unique identifier, which includes the model, capacity and firmware version; The test environment information is associated with the SN number and the test timestamp as unique identifiers. The test environment information includes the test configuration, host CPU model, memory capacity, and operating system version. The original acquired data is associated with the serial number (SN), test timestamp, and model label as unique identifiers. The original acquired data includes voltage, current, power, IOPS, bandwidth, and latency metrics with different model labels.

4. The intelligent management system for SSD power consumption testing based on QTL1999 according to claim 1, characterized in that: The horizontal and vertical comparisons of the aforementioned multi-dimensional manual comparison module include: Horizontal comparison is used to compare the energy efficiency differences of SSDs of different models or serial numbers under the same test model, and automatically generate bar charts or line charts. Longitudinal comparison is used to analyze the consistency of SSDs of the same model in different batches or at different test times, and automatically generates data comparison tables and anomaly annotations.

5. The intelligent management system for SSD power consumption testing based on QTL1999 according to claim 1, characterized in that: The interactive AI question-answering module is based on an open-source large model and has been customized and optimized for SSD power consumption testing, including: Configure domain terminology dictionaries and semantic parsing rules to improve the recognition accuracy of professional queries for SSD power consumption testing; Convert natural language queries into standardized database commands; Configure multi-turn dialogue context awareness function so that users do not need to repeat basic conditions when asking follow-up questions; After responding to a query, the interactive AI question-answering module automatically outputs visual charts and structured conclusions.

6. The intelligent management system for SSD power consumption testing based on QTL1999 according to claim 5, characterized in that: The open-source large model mentioned is DeepSeek-V3.1.

Images