A data hierarchical storage optimization method based on intelligent scheduling under cloud network integration

Abstract

This invention discloses a data hierarchical storage optimization method based on intelligent scheduling under cloud-network convergence, relating to the field of input / output data processing technology. It includes a data dynamic heat calculation module, a data processing module, a hard disk module, and a dynamic decay factor calculation module. The data dynamic heat calculation module collects data access characteristics in real time and constructs a corresponding dynamic heat model. The dynamic heat model calculates the heat value of an object. When a data object is accessed, its heat value is calculated using the heat value calculation formula. Based on the obtained heat value, the data is stored in the optimal hard disk location through the data storage location allocation unit within the hard disk module. This invention, by constructing a multi-dimensional dynamic heat model, achieves intelligent and adaptive optimization of data storage hierarchical storage, ensuring efficient utilization of storage resources and extending hardware lifespan.

Description

Technical Field

[0001] This invention relates to the field of input / output data processing technology, specifically to a data hierarchical storage optimization method based on intelligent scheduling under cloud-network convergence. Background Technology

[0002] When extracting data, the data is usually temporarily stored in a cache after the command is issued, and then processed by components such as information processors. The process of extracting the data is reversed.

[0003] Currently, data storage is generally done on specific hard drives according to system requirements. Various types of data are scattered across several hard drive partitions. This storage method causes frequently accessed data to be stored on older hard drives or mechanical hard drives with slow read speeds, which increases I / O latency and gives users a slow system processing experience. Summary of the Invention

[0004] To address the problem mentioned in the background art that frequently accessed data is stored on slow hard drives, the present invention aims to provide a data hierarchical storage optimization method based on intelligent scheduling under cloud-network convergence.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a data hierarchical storage optimization method based on intelligent scheduling under cloud-network convergence, comprising a data dynamic heat calculation module, a data processing module, a hard disk module and a dynamic decay factor calculation module; The data dynamic popularity calculation module collects data access characteristics in real time and constructs a corresponding dynamic popularity model. The popularity value of object d is then calculated using the dynamic popularity model. Its heat value The calculation formula is as follows: (1) In the above formula, This represents the popularity value of data object d at time t, and its value ranges from 0 to ∞. Let be the access frequency (times / second) of data object d during the i-th access. This is a data type weighting factor, which is based on the numerical value according to the category. Decide; The timestamp of the i-th visit; The weighting coefficients for semantic similarity; For computers to timestamp Time-allocated computing power; For data d and the current hot dataset semantic similarity, It is a dynamic decay factor; The hard disk module includes a hard disk data storage level calibration unit, a data storage location allocation unit, a hard disk identification unit, a hard disk bad sector calibration unit, and a hard disk verification unit; When data object d is accessed, it is determined by its popularity value. The calculation formula calculates the heat value of data object d, and based on the obtained heat value of data object d, the data storage location allocation unit in the hard disk module saves data d in the optimal hard disk.

[0006] Furthermore, The numerical video stream is 1.5, the structured data is 1.3, the log file is 1.0, and other data is 0.75.

[0007] Furthermore, in the formula (1) The dynamic attenuation factor is calculated as follows: (2) In the above formula, the values ​​are set according to business needs (video stream and structured data are calibrated to 0.1, log files are calibrated to 0.08, and other data are calibrated to 0.05). This represents the percentage of remaining hard drive capacity. This is the hard drive capacity alarm threshold; In the above formula (2) The basic attenuation rate is calculated using the following formula: (3) In the above formula, This represents the remaining lifespan of the hard drive. This refers to the estimated lifespan of the hard drive. When a user requests data object d, the real-time data acquisition unit and the access frequency recording unit will... , , , Data is collected from the resource monitoring system via the computing power allocation extraction unit. (Dynamically adjust the hard drive's IRQ interrupt frequency based on the GPU computing power allocation status, reducing the number of interrupt requests and I / O latency during high-computing-power tasks), and use the semantic similarity calculation unit to... The values ​​are used for calculation.

[0008] Furthermore, the data dynamic heat calculation module includes a real-time data acquisition unit, a data communication unit, a data similarity weight coefficient unit, a computing power allocation extraction unit, a semantic similarity calculation unit, and an access frequency recording unit. The data transmission processing unit is communicatively connected to the real-time data acquisition unit, the data transmission processing unit is communicatively connected to the data communication unit, the data communication unit is communicatively connected to the dynamic decay factor calculation module, the data transmission processing unit is communicatively connected to the data processing module, the data transmission processing unit is communicatively connected to the data similarity weight coefficient unit, the data transmission processing unit is communicatively connected to the computing power allocation extraction unit, and the data transmission processing unit is communicatively connected to the access frequency recording unit. The real-time data acquisition unit captures user-triggered data access commands through the input interface of the cloud network terminal and parses them into API call events in real time.

[0009] Furthermore, the dynamic attenuation factor calculation module includes a hard disk data communication unit, a hard disk attenuation calculation unit, and an alarm unit. The hard disk data communication unit is communicatively connected to the hard disk attenuation calculation unit, the hard disk attenuation calculation unit is communicatively connected to the alarm unit, and the hard disk data communication unit is communicatively connected to the hard disk module. The hard disk verification unit is communicatively connected to the hard disk bad sector calibration unit, the hard disk bad sector calibration unit is communicatively connected to the hard disk identification unit, the hard disk identification unit is communicatively connected to the hard disk data storage level calibration unit, and the hard disk data storage level calibration unit is communicatively connected to the hard disk data storage location allocation unit.

[0010] Furthermore, the hard drive verification unit periodically scans physical sectors and marks areas with excessive read / write error rates; Hard drive bad sector identification unit: Adds faulty sectors to a blacklist to prevent data from being written; Hard disk identification unit: used for hard disks containing specific data; Hard disk data storage class calibration unit: used to calibrate the storage class of each hard disk; Hard disk data storage location allocation unit: used to allocate and store data for each heat value.

[0011] Furthermore, the dynamic attenuation factor calculation module is mainly used for dynamic attenuation factor calculation. The calculation first identifies the remaining capacity, total capacity, design life and used life of each hard drive through the hard drive identification unit, and then issues an alarm through the alarm unit when the remaining capacity or remaining life is too low. Based on the information identified by the hard drive identification unit, each hard drive is classified into different levels by the hard drive data storage level calibration unit. The classification includes: Category 1, Category 2, and Category 3 (to be migrated out). Hard drives in the Category 3 (to be migrated out) indicate that they are no longer suitable for data storage, and an alarm is issued through the alarm unit. When the I / O queue depth of a Category 1 hard drive exceeds the threshold, some high-frequency data is automatically migrated to a Category 2 hard drive, and the priority of the DMA transfer channel of the corresponding interface is adjusted. Data with a popularity value greater than or equal to a predetermined value X is stored in a Category 1 hard drive, and data with a popularity value less than X is stored in a Category 2 hard drive, thus achieving tiered storage.

[0012] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. When the I / O queue depth of a type 1 hard drive exceeds a threshold, this invention automatically migrates some high-frequency data to a type 2 hard drive and adjusts the DMA transfer channel priority of the corresponding interface, greatly increasing the security of stored data; it dynamically adjusts the IRQ interrupt frequency of the hard drive according to the GPU computing power allocation status (e.g., reducing the number of interrupt requests during high-computing-power tasks to reduce I / O latency); and by constructing a multi-dimensional dynamic heat model, it realizes intelligent data storage and adaptive optimization of data tiered storage optimization. The heat value calculation not only includes access frequency and semantic similarity, but also introduces data type weighting factors to accurately identify high-value data; it introduces a dynamic decay factor linked to the hard drive hardware status: when the remaining capacity of the hard drive is insufficient or its lifespan is nearing a threshold, the system automatically accelerates the elimination of cold data and triggers an alarm.

[0013] 2. This invention improves the overall efficiency of data processing through semantic relevance enhancement and automated tiering strategies. The introduction of semantic similarity weights enables the system to intelligently identify data strongly related to business (such as associated logs or collaborative analysis files). Even if the access frequency is low, it may still be retained in the performance layer to prevent critical data from being mistakenly migrated due to fluctuations in access frequency. The hard disk health management mechanism (bad sector labeling and graded labeling) ensures storage reliability at the physical level. Faulty sectors are actively isolated, high-health hard disks (Category I) and medium-health hard disks (Category II) are dedicated to storing hot data, and low-health hard disks (pending migration) are gradually phased out, forming a closed-loop fault-tolerant system. The fully automated tiered storage strategy reduces operation and maintenance costs, while the integration of computing power allocation factors ensures fast access to data required for high-computing tasks (such as AI inference) and avoids idle computing resources.

[0014] The parts of the device not covered herein are the same as or can be implemented using existing technologies. Attached Figure Description

[0015] Figure 1 This is a system block diagram of a data hierarchical storage optimization method based on intelligent scheduling under cloud-network convergence according to the present invention.

[0016] Figure 2This is a simplified flowchart of a data hierarchical storage optimization method based on intelligent scheduling under cloud-network convergence according to the present invention. Detailed Implementation

[0017] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0018] like Figure 1 As shown, the present invention provides a data hierarchical storage optimization method based on intelligent scheduling under cloud-network convergence, including a data dynamic heat calculation module, a data processing module, a hard disk module and a dynamic decay factor calculation module; The data access characteristics are collected in real time by the dynamic heat calculation module, and a corresponding dynamic heat model is constructed. The heat value of object d is then calculated using the dynamic heat model. Its heat value The calculation formula is as follows: (1) In the above formula, This represents the popularity value of data object d at time t, and its value ranges from 0 to ∞. Let be the access frequency (times / second) of data object d during the i-th access. This is a data type weighting factor, which is based on the numerical value according to the category. Decision (video stream 1.5, structured data 1.3, log file 1.0, other data 0.75); The timestamp of the i-th visit; The weighting coefficients for semantic similarity; For computers to timestamp Time-allocated computing power; For data d and the current hot dataset semantic similarity; In the above formula (1) The dynamic attenuation factor is calculated as follows: (2) In the above formula, the values ​​are set according to business needs (video stream and structured data are calibrated to 0.1, log files are calibrated to 0.08, and other data are calibrated to 0.05). This represents the percentage of remaining hard drive capacity. This is the hard drive capacity alarm threshold; In the above formula (2) The basic attenuation rate is calculated using the following formula: (3) In the above formula, This represents the remaining lifespan of the hard drive. This refers to the estimated lifespan of the hard drive. The hard disk module includes a hard disk data storage level calibration unit, a data storage location allocation unit, a hard disk identification unit, a hard disk bad sector calibration unit, and a hard disk verification unit; When data object d is accessed, it is determined by its popularity value. The calculation formula calculates the heat value of data object d, and based on the obtained heat value of data object d, the data storage location allocation unit in the hard disk module saves data d in the optimal storage.

[0019] The data dynamic popularity calculation module includes a real-time data acquisition unit, a data communication unit, a data similarity weight coefficient unit, a computing power allocation extraction unit, a semantic similarity calculation unit, and an access frequency recording unit.

[0020] The data transmission processing unit is connected to the real-time data acquisition unit, the data transmission processing unit is connected to the data communication unit, the data communication unit is connected to the dynamic attenuation factor calculation module, the data transmission processing unit is connected to the data processing module, the data transmission processing unit is connected to the data similarity weight coefficient unit, the data transmission processing unit is connected to the computing power allocation extraction unit, and the data transmission processing unit is connected to the access frequency recording unit. Real-time data acquisition unit: captures data access characteristics (access timestamp, frequency, computing power allocation) in real time. Input: User request events (such as API calls, file access).

[0021] Information collection logic: Each access generates a record containing: {Data ID, Timestamp, Computing Power Allocation}.

[0022] Triggering condition: Data acquisition is completed within 50ms after the data access operation is completed.

[0023] Output: The raw access feature data stream is pushed to the data processing module.

[0024] Access frequency recording unit: Statistical data on the access frequency of object d (times / second); Input: Access frequency of a specific object.

[0025] Statistical rules: The count is calculated by sliding the count in time windows (default 1 second) to determine the number of accesses per second.

[0026] Smoothing is applied to sudden traffic spikes: (j) i For the current count, j i-1 (This is the average of the previous window).

[0027] Computing power allocation extraction unit: Obtains the current computing power allocation value from the resource monitoring system.

[0028] Output: The system retrieves the current computing power allocation value. Information collection logic: Obtain CPU / GPU allocation from the monitoring agent of the Kubernetes cluster or physical machine.

[0029] Output: CPU / GPU allocation ratio values.

[0030] Semantic similarity calculation unit: calculates the semantic relevance between data d and the current hot dataset.

[0031] Data similarity weight coefficient unit: Sets the semantic similarity weight coefficient.

[0032] Both the semantic similarity calculation unit and the data similarity weight coefficient unit adopt the following logic.

[0033] Input: Semantic similarity calculation request.

[0034] Hot dataset definition: A collection of data containing the top 100 most popular data points over the past 24 hours.

[0035] Text similarity judgment logic: Based on TF-IDF vectorization (converting text into numerical vectors for easier machine learning processing), cosine similarity is calculated.

[0036] For non-text image data (such as video): extract metadata tags (resolution, subject category) for matching.

[0037] Output: 0 or 1, where 0 indicates no association and 1 indicates full association.

[0038] Then, substitute the various data obtained above into formula (1) to calculate the heat value of a specific data object d.

[0039] The dynamic attenuation factor calculation module includes a hard disk data communication unit, a hard disk attenuation calculation unit, and an alarm unit. The hard disk data communication unit is communicatively connected to the hard disk attenuation calculation unit, the hard disk attenuation calculation unit is communicatively connected to the alarm unit, and the hard disk data communication unit is communicatively connected to the hard disk module.

[0040] The hard disk verification unit is connected to the hard disk bad sector calibration unit, the hard disk bad sector calibration unit is connected to the hard disk identification unit, the hard disk identification unit is connected to the hard disk data storage level calibration unit, and the hard disk data storage level calibration unit is connected to the hard disk data storage location allocation unit.

[0041] Hard drive verification unit: Periodically scans physical sectors and marks areas with excessive read / write error rates; Hard drive bad sector identification unit: Adds faulty sectors to a blacklist to prevent data from being written; Hard disk identification unit: used for hard disks containing specific data; Hard disk data storage class calibration unit: used to calibrate the storage class of each hard disk; Hard disk data storage location allocation unit: used to allocate and store data for each heat value.

[0042] The dynamic attenuation factor calculation module is mainly used for dynamic attenuation factor calculation. The calculation first identifies the remaining capacity, total capacity, design life and used life of each hard drive through the hard drive identification unit, and then issues an alarm through the alarm unit when the remaining capacity or remaining life is too low. Based on the information identified by the hard drive identification unit, each hard drive is classified into different levels by the hard drive data storage level calibration unit. The classification includes: Category 1, Category 2, and Category 3 (to be moved out). Hard drives in the Category 3 (to be moved out) indicate that they are no longer suitable for data storage. An alarm is issued through the alarm unit, and data with a heat value greater than or equal to a predetermined value X is stored in Category 1 hard drives, while data with a heat value less than X is stored in Category 2 hard drives, thus achieving tiered storage.

[0043] It should be noted that Category 1 hard drives are defined as: NVMe SSDs with a health score >80 and support for PCIe 4.0 x4 interface; Category 2 hard drives are defined as: SATA SSDs / HDDs with a health score of 60-80 and support for hot-swapping.

[0044] Specifically, this embodiment adopts an intelligent storage resource scheduling mechanism and constructs a hierarchical response system for different types of storage media. When the I / O queue depth of a type of high-speed hard drive exceeds a preset threshold (e.g., depth > 100), the storage controller will activate a heat perception algorithm. By analyzing the correlation between LBA access frequency and data blocks in real time, it will identify a subset of high-heat data and use a double buffering mechanism to gradually migrate it to a designated high-speed storage area of ​​a type of large-capacity hard drive. At the same time, the system synchronously optimizes the DMA transmission architecture, dynamically allocates transmission bandwidth based on the load status of the PCIe channel, and implements a priority improvement strategy for logical channels involving data migration to ensure that the transmission latency of critical data is controlled within 200 microseconds.

[0045] At the computing power collaboration level, the system establishes a computing-storage joint optimization model through deep interaction with the GPU scheduler. When a high computing power task with CUDA core utilization exceeding 75% is detected, the storage subsystem will start a low-interference mode: based on the characteristics of the NUMA architecture, the IRQ interrupt mapping table is reconstructed, the interrupt request frequency is reduced, and the batch completion queue mechanism is enabled to increase the number of I / O requests processed in a single interrupt. This adaptive adjustment strategy reduces the arbitration conflict rate of the PCIe bus during GPU-intensive computing, while maintaining the storage access latency within the 15ms threshold agreed upon by the SLA, thus achieving an elastic balance between computing and I / O resources.

[0046] It should be noted that in this embodiment, the system dynamically triggers multi-dimensional performance optimization strategies by monitoring the I / O queue depth of the first type of hard drive in real time. When the concurrent request queue of the NVMe hard drive exceeds a preset threshold (e.g., 100 incomplete requests), the system first selects high-load data blocks with long queue times and low popularity values ​​from the first type of hard drive based on a comprehensive evaluation of data popularity and access latency. These high-load data blocks are then migrated to the fast access area of ​​the second type of hard drive. The migration decision not only considers the historical access frequency of the data but also dynamically adjusts the priority based on the current queue pressure to ensure that core high-frequency data always resides in the high-performance storage layer. At the same time, it alleviates the congestion of the first type of hard drive. To balance the efficiency of cross-level data transfer, the system synchronously adjusts the DMA channel priority of the hardware interface by reducing the DMA bandwidth allocation weight of the first type of hard drive and alleviating the congestion of the second type of hard drive. The system temporarily increases transmission quotas for hard drives to redistribute traffic between storage layers. During this process, the system reserves a minimum guaranteed bandwidth for Class I hard drives to prevent a sudden drop in access performance for critical business data. At the same time, the system dynamically optimizes the hard drive interrupt request mechanism for the real-time status of GPU computing-intensive tasks (such as AI model training): during peak GPU computing load periods, interrupt merging technology is used to extend the response cycle and reduce CPU resource contention caused by frequent I / O interrupt processing; when GPU load decreases, it switches to low-latency interrupt mode to ensure that the storage system responds quickly to data requests. This hardware and software coordinated control mechanism reduces I / O latency fluctuations by 60% during high computing load tasks, stabilizes GPU resource utilization at over 90%, and increases the throughput of Class II hard drives by 40%, forming a load-balanced and elastically scalable storage resource pool.

[0047] In this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, without necessarily requiring or implying any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0048] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A data hierarchical storage optimization method based on intelligent scheduling under cloud-network convergence, characterized in that: It includes a data dynamic heat calculation module, a data processing module, a hard disk module, and a dynamic decay factor calculation module; The data dynamic popularity calculation module collects data access characteristics in real time and constructs a corresponding dynamic popularity model. The popularity value of object d is then calculated using the dynamic popularity model. Its heat value The calculation formula is as follows: （1） In the above formula, This represents the popularity value of data object d at time t, and its value ranges from 0 to ∞. Let be the access frequency (times / second) of data object d during the i-th access. This is a data type weighting factor, which is based on the numerical value according to the category. Decide; The timestamp of the i-th visit; The weighting coefficients for semantic similarity; For computers to timestamp Time-allocated computing power; For data d and the current hot dataset semantic similarity, It is a dynamic decay factor; The hard disk module includes a hard disk data storage level calibration unit, a data storage location allocation unit, a hard disk identification unit, a hard disk bad sector calibration unit, and a hard disk verification unit; When data object d is accessed, it is determined by its popularity value. The calculation formula calculates the heat value of data object d, and based on the obtained heat value of data object d, the data storage location allocation unit in the hard disk module saves data d in the optimal hard disk.

2. The data hierarchical storage optimization method based on intelligent scheduling under cloud-network convergence according to claim 1, characterized in that: The numerical video stream is 1.5, the structured data is 1.3, the log file is 1.0, and other data is 0.

75.

3. The data hierarchical storage optimization method based on intelligent scheduling under cloud-network convergence according to claim 2, characterized in that: In the formula (1) The dynamic attenuation factor is calculated as follows: （2） In the above formula, the values ​​are set according to business needs (video stream and structured data are calibrated to 0.1, log files are calibrated to 0.08, and other data are calibrated to 0.05). This represents the percentage of remaining hard drive capacity. This is the hard drive capacity alarm threshold; In the above formula (2) The basic attenuation rate is calculated using the following formula: （3） In the above formula, This represents the remaining lifespan of the hard drive. This refers to the estimated lifespan of the hard drive. When a user requests data object d, the real-time data acquisition unit and the access frequency recording unit will... , , , Data is collected from the resource monitoring system via the computing power allocation extraction unit. (Dynamically adjust the hard drive's IRQ interrupt frequency based on the GPU computing power allocation status, reducing the number of interrupt requests and I / O latency during high-computing-power tasks), and use the semantic similarity calculation unit to... The values ​​are used for calculation.

4. The data hierarchical storage optimization method based on intelligent scheduling under cloud-network convergence according to claim 3, characterized in that: The dynamic data popularity calculation module includes a real-time data acquisition unit, a data communication unit, a data similarity weight coefficient unit, a computing power allocation extraction unit, a semantic similarity calculation unit, and an access frequency recording unit. The data transmission processing unit is communicatively connected to the real-time data acquisition unit, the data transmission processing unit is communicatively connected to the data communication unit, the data communication unit is communicatively connected to the dynamic decay factor calculation module, the data transmission processing unit is communicatively connected to the data processing module, the data transmission processing unit is communicatively connected to the data similarity weight coefficient unit, the data transmission processing unit is communicatively connected to the computing power allocation extraction unit, and the data transmission processing unit is communicatively connected to the access frequency recording unit. The real-time data acquisition unit captures user-triggered data access commands through the input interface of the cloud network terminal and parses them into API call events in real time.

5. The data hierarchical storage optimization method based on intelligent scheduling under cloud-network convergence according to claim 4, characterized in that: The dynamic attenuation factor calculation module includes a hard disk data communication unit, a hard disk attenuation calculation unit, and an alarm unit. The hard disk data communication unit is communicatively connected to the hard disk attenuation calculation unit, the hard disk attenuation calculation unit is communicatively connected to the alarm unit, and the hard disk data communication unit is communicatively connected to the hard disk module. The hard disk verification unit is communicatively connected to the hard disk bad sector calibration unit, the hard disk bad sector calibration unit is communicatively connected to the hard disk identification unit, the hard disk identification unit is communicatively connected to the hard disk data storage level calibration unit, and the hard disk data storage level calibration unit is communicatively connected to the hard disk data storage location allocation unit.

6. The data hierarchical storage optimization method based on intelligent scheduling under cloud-network convergence according to claim 5, characterized in that: Hard drive verification unit: Periodically scans physical sectors and marks areas with excessive read / write error rates; Hard drive bad sector identification unit: Adds faulty sectors to a blacklist to prevent data from being written; Hard disk identification unit: used for hard disks containing specific data; Hard disk data storage class calibration unit: used to calibrate the storage class of each hard disk; Hard disk data storage location allocation unit: used to allocate and store data for each heat value.

7. The data hierarchical storage optimization method based on intelligent scheduling under cloud-network convergence according to claim 6, characterized in that: The dynamic attenuation factor calculation module is mainly used for dynamic attenuation factor calculation. The calculation first identifies the remaining capacity, total capacity, design life and used life of each hard drive through the hard drive identification unit, and then issues an alarm through the alarm unit when the remaining capacity or remaining life is too low. Based on the information identified by the hard drive identification unit, each hard drive is classified into different levels by the hard drive data storage level calibration unit. The classification includes: Category 1, Category 2, and Category 3 (to be migrated out). Hard drives in the Category 3 (to be migrated out) indicate that they are no longer suitable for data storage, and an alarm is issued through the alarm unit. When the I / O queue depth of a Category 1 hard drive exceeds the threshold, some high-frequency data is automatically migrated to a Category 2 hard drive, and the priority of the DMA transfer channel of the corresponding interface is adjusted. Data with a popularity value greater than or equal to a predetermined value X is stored in a Category 1 hard drive, and data with a popularity value less than X is stored in a Category 2 hard drive, thus achieving tiered storage.

Images