A method and system for synchronous read and write of multi-path devices

Abstract

The application provides a multi-path device synchronous read-write method and system, relates to the technical field of multi-path synchronous read-write operation monitoring, and realizes intelligent identification and intervention on device read-write conflicts by array arrangement on the multi-path device, collection of first state data sets and second state data sets of the device, calculation of addressing interference coefficients Xzr i of the i-th device and array conflict scheduling cost coefficients Cb i , and accurate evaluation of synchronous scheduling risks. After the execution of the scheduling strategy, the system further realizes real-time perception of fine-grained data interference based on page block heat tracking and adjacent page disturbance analysis, and dynamically triggers a hot data migration strategy, so that the read disturbance between devices and bus contention are effectively reduced, and the overall read-write efficiency and data consistency are improved. The system is particularly suitable for high-concurrency, distributed storage and high-reliability scenarios.

Description

Technical Field

[0001] This invention relates to the field of multi-channel synchronous read / write operation monitoring technology, specifically a method and system for multi-channel device synchronous read / write. Background Technology

[0002] With the rapid development of technologies such as big data, artificial intelligence, and edge computing, scenarios involving high-speed reading and writing of massive amounts of data are increasing, especially in distributed storage systems, database clusters, high-performance computing platforms (HPC), data centers (IDC), video surveillance storage, network log storage, and industrial control systems. These systems commonly employ array read / write structures composed of multiple parallel devices to improve overall system throughput and fault tolerance. For example, common RAID (Redundant Array of Independent Disks) systems, distributed SSD storage arrays, NAS (Network Attached Storage) clusters, and cloud block storage services all achieve parallel data reading and writing and high-availability storage through the collaborative work of multiple storage devices.

[0003] However, in these multi-device array structures, because multiple devices need to simultaneously respond to concurrent tasks from the upper-layer scheduler, the addresses requesting read and write operations may be highly concentrated, leading to frequent head seeking behavior. This is especially true in mechanical hard drive scenarios, where the head repeatedly switches between multiple physical regions, easily causing addressing conflicts and response delays. In solid-state drive arrays, frequent concurrent writes can also lead to page block aging, write amplification, and adjacent page corruption (i.e., the "read perturbation" effect). Furthermore, in synchronous write scenarios, differences in cache hits between devices, inconsistent scheduling priorities, and write failure retries can also cause decreased synchronization efficiency and performance jitter.

[0004] Existing technologies mostly focus on optimizing I / O scheduling for single devices, or simply perform load balancing at the storage controller layer. They lack systematic analysis and early warning mechanisms for parameters such as addressing behavior, conflict level, cache performance, and write stability of multi-channel array devices under high-concurrency synchronous read and write operations. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for synchronous reading and writing of multiple devices to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A method for synchronous read / write of multiple devices, comprising the following steps:

[0008] Step 1: Arrange the multi-channel devices in an array, and collect the current head addressing position sequence of the i-th device and the read / write frequency of the i-th device within the time interval t. Read / write access conflict count The most recent write failure rate Xr of the i-th device i and the current cache hit rate Hc i Generate a first state dataset; and make a preliminary judgment on the synchronization scheduling status in the device array based on the first state dataset. If it is determined that there is a high-frequency addressing or concurrent write conflict, issue a corresponding warning instruction and construct an abnormal synchronous read and write device group.

[0009] Step 2: After receiving the abnormal synchronous read / write device group, collect the operating environment factors of the i-th device in real time to generate the second state dataset;

[0010] Step 3: Construct a multi-device conflict identification model based on a convolutional neural network, and train it using the first state dataset and the second state dataset as input to calculate the addressing interference coefficient Xzr of the i-th device. i And array conflict scheduling cost coefficient Cb i And a preset interference tolerance threshold X is set. When any item exceeds the interference tolerance threshold X, a first scheduling strategy or a second scheduling strategy is generated.

[0011] Step 4: After executing the first or second scheduling strategy, track and analyze the page block access heat during the execution process, record the long-term active page block regions, and construct the neighboring page disturbance coefficient Rd for the j-th page block of the i-th device. i,j A preset safety threshold Y is used to determine the neighboring page disturbance coefficient Rd of the j-th page block of the i-th device. i,j Does the j-th page block in the i-th device experience interference from other devices? If so, trigger the third scheduling strategy.

[0012] A multi-device synchronous read / write system, comprising:

[0013] The equipment array step unit arranges multiple devices in an array and marks several devices.

[0014] The first acquisition unit is used to acquire the current head addressing position sequence of the i-th device and the read / write frequency of the i-th device within the time interval t, using an integrated I / O performance monitoring module, a head addressing tracker, a write check logger, a cache hit detector, and a synchronization clock module. Read / write access conflict count The most recent write failure rate Xr of the i-th device i and the current cache hit rate Hc i Generate the first state dataset;

[0015] The first state determination unit is used to make a preliminary judgment on the synchronous scheduling state in the device array based on the first state dataset. If it is determined that there is a high-frequency addressing or concurrent write conflict, it issues a corresponding warning instruction and constructs an abnormal synchronous read and write device group.

[0016] The second acquisition unit, upon receiving a group of abnormal synchronous read / write devices, collects real-time environmental factors of the i-th device, including the voltage fluctuation amplitude of the i-th device. Array bus utilization i Controller delay fluctuation coefficient CtrlVar i Equipment heat growth rate (GeatRate) i and I / O contention density i Generate the second state dataset;

[0017] The model building unit constructs a multi-device conflict recognition model based on a convolutional neural network, and trains it using the first-state dataset and the second-state dataset as input to calculate the addressing interference coefficient Xzr of the i-th device. i And array conflict scheduling cost coefficient Cb i ;

[0018] The second conflict determination unit is used to preset the interference tolerance threshold X. When any one of them exceeds the interference tolerance threshold X, it generates the first scheduling strategy or the second scheduling strategy.

[0019] The third acquisition unit is used to track and analyze the page block access heat during the execution process after executing the first or second scheduling strategy, record the long-term active page block region, and construct the neighboring page disturbance coefficient Rd of the j-th page block of the i-th device. i,j ;

[0020] The third interference determination unit presets a safety threshold Y to determine the neighboring page disturbance coefficient Rd of the j-th page block of the i-th device. i,j Does the j-th page block in the i-th device experience interference from other devices? If so, trigger the third scheduling strategy.

[0021] The control unit is used to execute the first scheduling strategy, the second scheduling strategy, and the third scheduling strategy.

[0022] Compared with the prior art, the beneficial effects of the present invention are:

[0023] This invention collects multi-dimensional operational characteristics of devices, such as addressing sequences, read / write frequencies, access conflict counts, write failure rates, and cache hit rates, and combines these with operational environment information to form a precise state characterization capability, improving the accuracy of identifying abnormal synchronization states. The conflict identification model built using a convolutional neural network can distinguish between addressing interference and scheduling conflict types. By calculating the interference tolerance threshold X and the safety threshold Y, it achieves intelligent hierarchical processing of abnormal states, avoiding resource waste or insufficient processing caused by a uniform strategy. By generating first and second scheduling strategies, it effectively diverts hot data or conflicting tasks, reducing head oscillation and concentrated page block writes caused by high-frequency addressing, improving the collaborative efficiency and lifespan of mechanical hard drives or solid-state drive arrays. Using the neighboring page perturbation coefficient index to track and analyze abnormal access behavior of specific page blocks, it can accurately determine the degree of interference of other devices' read / write activities on the current device's page blocks, making it particularly suitable for the dynamic monitoring and intervention of "read perturbation" problems in solid-state drive arrays. When the disturbance coefficient of a neighboring page is detected to be higher than the safety threshold Y, the system can promptly trigger the third scheduling strategy. Through task reallocation, hot page block migration, cache isolation, or load adjustment, cross-device interference suppression is achieved, preventing data corruption or write amplification effect from repeated access to a page block by multiple devices. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the overall method steps of the present invention.

[0025] Figure 2 This is a schematic diagram of the system flow of the present invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0027] Example 1:

[0028] Please see Figure 1 This invention provides a technical solution: a method for synchronous reading and writing of multiple devices, the specific steps of which include:

[0029] Step 1: Arrange the multi-channel devices in an array, and collect the current head addressing position sequence of the i-th device and the read / write frequency of the i-th device within the time interval t. Read / write access conflict count The most recent write failure rate Xr of the i-th device i and the current cache hit rate Hc iGenerate a first state dataset; and make a preliminary judgment on the synchronization scheduling status in the device array based on the first state dataset. If it is determined that there is a high-frequency addressing or concurrent write conflict, issue a corresponding warning instruction and construct an abnormal synchronous read and write device group.

[0030] Step 2: After receiving the abnormal synchronous read / write device group, collect the operating environment factors of the i-th device in real time to generate the second state dataset;

[0031] Step 3: Construct a multi-device conflict identification model based on a convolutional neural network, and train it using the first state dataset and the second state dataset as input to calculate the addressing interference coefficient Xzr of the i-th device. i And array conflict scheduling cost coefficient Cb i And a preset interference tolerance threshold X is set. When any item exceeds the interference tolerance threshold X, a first scheduling strategy or a second scheduling strategy is generated.

[0032] Step 4: After executing the first or second scheduling strategy, track and analyze the page block access heat during the execution process, record the long-term active page block regions, and construct the neighboring page disturbance coefficient Rd for the j-th page block of the i-th device. i,j A preset safety threshold Y is used to determine the neighboring page disturbance coefficient Rd of the j-th page block of the i-th device. i,j Does the j-th page block in the i-th device experience interference from other devices? If so, trigger the third scheduling strategy.

[0033] In this embodiment, the method collects multi-dimensional operational characteristics of the device, such as addressing sequence, read / write frequency, access conflict count, write failure rate, and cache hit rate, and combines them with operational environment information to form a precise state characterization capability, thereby improving the accuracy of identifying abnormal synchronization states. The conflict identification model built using a convolutional neural network can distinguish between addressing interference and scheduling conflict types, and by calculating the interference tolerance threshold (X) and disturbance safety threshold (Y), it achieves intelligent hierarchical processing of abnormal states, avoiding resource waste or insufficient processing caused by a uniform strategy. By generating first and second scheduling strategies, it effectively diverts hot data or conflicting tasks, reducing head oscillation and concentrated page block writing problems caused by high-frequency addressing, and improving the collaborative efficiency and lifespan of mechanical hard drives or solid-state drive arrays. Using the neighboring page disturbance coefficient index to track and analyze abnormal access behavior of specific page blocks, it can accurately determine the degree of interference of other device read / write activities on the current device's page blocks, which is particularly suitable for the dynamic monitoring and intervention of "read disturbance" problems in solid-state drive arrays. When the disturbance coefficient of a neighboring page is detected to be higher than the safety threshold Y, the system can promptly trigger the third scheduling strategy. Through task reallocation, hot page block migration, cache isolation, or load adjustment, cross-device interference suppression is achieved, preventing data corruption or write amplification effect from repeated access to a page block by multiple devices.

[0034] Example 2,

[0035] Step one includes:

[0036] S101. Using an integrated I / O performance monitoring module, head addressing tracker, write check logger, cache hit detector, and synchronization clock module, real-time information is collected from the i-th device to obtain the following metrics:

[0037] The current head addressing position sequence of the i-th device;

[0038] The read / write frequency of the i-th device within the time interval t Specifically: Set a time interval t, read the total number of read and write requests within that time interval, and obtain the number of read requests. and number of write requests Then the number of read requests and number of write requests Add them together to get the total number of visits Divide the total number of accesses by the length of the time interval to obtain the read / write frequency.

[0039] The number of read / write access conflicts of the i-th device within the time interval t. Specifically: record the conflict time of the i-th device, obtain the number of conflicts within the time interval t, and calculate it using the following formula:

[0040]

[0041] in, This represents the number of times each read / write request of the i-th device conflicts with other requests within the time interval t, and N represents the total number of devices participating in the statistics in the system, i.e., from the 1st device to the Nth device;

[0042] The most recent write failure rate Xr of the i-th device i Specifically: Xr i =xrsbcs / Total Write ;

[0043] Where xrsbcs represents the number of write failures, and Total... Write This indicates the total number of write attempts, obtained from the device logs;

[0044] The current cache hit rate Hc of the i-th device i Specifically: Hc i =Cache Hits / Total Cache ;

[0045] Where Cache Hits represents the number of cache hits, and Total... Cache This indicates the total number of cache requests;

[0046] S102, Collect the current head addressing position sequence of the i-th device and the read / write frequency of the i-th device within the time interval t. Read / write access conflict count The most recent write failure rate Xr of the i-th device i and the current cache hit rate Hc i It is updated in real time and stored as the first state dataset for subsequent judgment and early warning.

[0047] Furthermore, step one above also includes:

[0048] S103. Determine the operating status of the i-th device based on the first state dataset and issue a corresponding alarm command, including:

[0049] If the addressing jump frequency per unit time in the current head addressing position sequence of the i-th device is significantly higher than twice the set average addressing frequency of the device, it indicates that the device has high-frequency fragment access or scheduling abnormality, and the first addressing abnormality warning instruction is generated; if the addressing jump frequency per unit time is not higher than twice the set average addressing frequency of the device, it indicates that there is no high-frequency access abnormality, and monitoring continues.

[0050] When the read / write frequency of the i-th device within the time interval t If the frequency exceeds 200 times / second, it indicates that the device is under abnormally high load, generating the first high load warning command; if the read / write frequency of the i-th device within the time interval t... If the frequency does not exceed 200 times per second, it indicates that there is no abnormal load, and monitoring should continue.

[0051] The number of read / write access conflicts of the i-th device within the time interval t If the number of read / write access conflicts exceeds 10% of the total number of accesses, it indicates that the device is at risk of frequent resource contention during access scheduling, and a first access conflict warning instruction is generated; if the number of read / write access conflicts of i devices within the time interval t... If the number of accesses does not exceed 10% of the total number of accesses, it means that the access conflict of the device is within the expected range, and continuous monitoring will continue.

[0052] When the write failure rate of the most recent task on the i-th device is Xr i If the write failure rate exceeds 10%, it indicates a risk of unstable write operations, disk corruption, or control errors, generating a first write failure warning command; if Xr i If the percentage is below 10%, it is considered normal; continued monitoring is necessary.

[0053] When the current cache hit rate Hc of the i-th device i A value greater than 0.6 indicates an abnormal caching mechanism on the device, leading to frequent accesses to main memory and generating a first cache invalidation warning instruction; if Xr i A value ≤0.6 indicates that the cache is functioning normally; continuous monitoring is required.

[0054] S104. Construct an abnormal synchronous read / write device group by aggregating the first addressing anomaly warning instruction, the first high load warning instruction, the first access conflict warning instruction, the first write failure warning instruction, and the first cache failure warning instruction.

[0055] In this embodiment, step one of the present invention achieves refined and real-time monitoring of the operating status of each sub-device in a multi-channel system by constructing an integrated I / O performance monitoring module, a head addressing tracker, a write check logger, a cache hit detector, and a synchronization clock module. Unlike the coarse-grained methods of traditional systems that focus on overall scheduling performance or simple latency indicators, this approach not only considers conventional indicators such as read / write frequency and conflict count, but also introduces multi-dimensional state variables such as address jump frequency, cache hit rate, and write failure rate, significantly enhancing the ability to detect abnormal device behavior in advance. By introducing configurable thresholds such as "more than twice the average addressing frequency," "read / write frequency exceeding 200 times / second," and "access conflict rate exceeding 10%," dynamic adaptation to different business scenarios, load characteristics, and device types is achieved. Through the integrated identification of multiple types of early warning signals, not only can target devices participating in abnormal scheduling behavior be accurately screened, but high-quality input can also be provided for subsequent conflict identification modeling and scheduling strategy formulation.

[0056] Example 3

[0057] Step two includes:

[0058] S201. Based on the unique device identifier carried in each of the first addressing anomaly warning instruction, first high load warning instruction, first access conflict warning instruction, first write failure warning instruction, and first cache failure warning instruction of the abnormal synchronous read / write device group, including device ID, logical address, or array location information; use the device management table or the device mapping table of the centralized controller to locate the device and mark the physical location, network address, and subsystem information of each target device in the abnormal synchronous read / write device group;

[0059] S202, integrating voltage sensor, communication sensor, temperature sensor, controller response monitoring device, and I / O queue monitoring device, performs real-time operating environment information acquisition on the i-th device to obtain...

[0060] Includes: the voltage fluctuation amplitude of the i-th device Array bus utilization i Controller delay fluctuation coefficient CtrlVar i Equipment heat growth rate (GeatRate) i and I / O contention density i Generate the second state dataset.

[0061] Furthermore, S202 includes:

[0062] S2021. A voltage sensor is used to record the maximum value of the power supply voltage of the i-th device within the time interval t. and minimum value Calculate the difference between the two values ​​to obtain the voltage fluctuation amplitude of the i-th device.

[0063]

[0064] S2022. Use a communication sensor to record the total time of data frames communicated by the i-th device on the bus, and divide it by the time interval t to obtain the array bus utilization rate BusUti of the i-th device. i :

[0065] S2023. Using a controller response monitoring device, record the response time of multiple control commands, calculate the ratio of its standard deviation to the mean, and obtain the controller delay fluctuation coefficient CtrlVar for the i-th device. i :

[0066] S024. Use a temperature sensor module to periodically read the temperature data of the i-th device and calculate the device heat growth rate (GeatRate). i The expression is:

[0067]

[0068] Where Wd(t2) is the temperature of the i-th device at the second time point t2; Wd(t1) is the temperature of the i-th device at the first time point t1; t2-t1 represents the interval between the second time point and the first time point;

[0069] S2025. Using an I / O queue monitoring device, record the ratio of the total queuing time for device I / O requests to the number of requests within a time interval t, and obtain the I / O contention density (IODensity) of the i-th device. i .

[0070] In this embodiment, by using device ID, logical address, array location information in conjunction with device management table and controller mapping table, the physical / network location of the device in the system and its module can be quickly identified, effectively improving the accuracy of subsequent intervention or hot migration strategies; voltage fluctuation amplitude reflects power supply stability, helping to identify performance jitter caused by power supply interference;

[0071] The array bus utilization rate can indicate the current bus bottleneck situation, making it easier to determine whether it is caused by communication resource congestion;

[0072] The controller latency fluctuation coefficient can reveal problems at the underlying instruction processing level, such as control path interference or cache conflicts.

[0073] The heat growth rate is associated with potential equipment self-heating or heat dissipation anomalies, which helps identify performance degradation caused by temperature control imbalances;

[0074] I / O contention density is directly related to the system resource scheduling status and is an important support for analyzing access conflicts and load concentration.

[0075] Example 4

[0076] Step three includes:

[0077] S301. Using a convolutional neural network (CNN), construct an initial CNN model. Train and test the initial CNN model using a first-state dataset and a second-state dataset. Use the trained initial CNN model as a multi-device conflict identification model. Simultaneously, use the intermediate layer output of the device operation state model as a feature vector to identify feature information. Train and test the multi-device conflict identification model using the acquired feature information. Use the trained multi-device conflict identification model as a data operation prediction model to analyze and obtain the addressing interference coefficient Xzr of the i-th device. i And array conflict scheduling cost coefficient Cb i ;

[0078] S302, Addressing interference coefficient Xzr of the i-th device i The method of obtaining it is:

[0079] Extract the total number of accesses of the i-th device in the first state dataset within the time interval t. Read / write access conflict count and the current cache hit rate Hc i And the controller delay fluctuation coefficient CtrlVar in the second state dataset. i After dimensionless processing, the addressing interference coefficient Xzr of the i-th device is calculated using the following formula. i :

[0080]

[0081] In the formula, α1, α2, and α3 are weights, and α1 + α2 + α3 = 1, where Xzr is the addressing interference coefficient of the i-th device. i The higher the value, the more severe the interference with the current head addressing of the device, which is more likely to cause performance degradation or warnings.

[0082] S303, the array conflict scheduling cost coefficient Cb of the i-th device. i The method of obtaining it is:

[0083] Extract the array bus utilization (BusUti) of the i-th device in the second dataset. i Equipment heat growth rate GeagRate i I / O contention density i and voltage fluctuation amplitude After dimensionless processing, the array conflict scheduling cost coefficient Cb of the i-th device is calculated using the following formula. i :

[0084]

[0085] In the formula, α4, α5, α6, and α7 are weights, and α4 + α5 + α6 + α7 = 1, where Cb is the array conflict scheduling cost coefficient of the i-th device. i The higher the value, the more system resource contention or risk will occur if the device is scheduled under the current state.

[0086] S304. Preset the interference tolerance threshold X, and set the addressing interference coefficient Xzr of the i-th device. i and the array conflict scheduling cost coefficient Cb for the i-th device. i By comparing with the interference tolerance threshold X, the following judgment results are obtained, including:

[0087] When the addressing interference coefficient Xzr of the i-th device i and the array conflict scheduling cost coefficient Cb i If none of the values ​​exceed the interference tolerance threshold X, it means that the device is deemed to be in a qualified operating state, and the original scheduling strategy will be maintained so that it can continue to participate in synchronous read and write tasks and be marked as a valid scheduling device.

[0088] When the addressing interference coefficient Xzr of the i-th device i The interference tolerance threshold X is exceeded, but the array conflict scheduling cost coefficient Cb i If the interference tolerance threshold X is not exceeded, it is determined that the device's head addressing is abnormally affected by external interference. A first scheduling strategy is generated, including: reducing the scheduling priority of the device by 20%-30%, reducing the task scheduling of the device by 15%-20%, and pre-allocating 30%-50% of the write request cache of the device to other low-load devices.

[0089] When the addressing interference coefficient Xzr of the i-th device i The interference tolerance threshold X was not exceeded, but the array conflict scheduling cost coefficient Cb i If the interference tolerance threshold X is exceeded, it is determined that the device has a significant abnormal risk in terms of resource contention, and a second scheduling strategy is generated, including: reducing the scheduling priority of the device by 31%-40%, reducing the task scheduling of the device by 21%-30%, reducing the array bus usage of the device by 10%-20%, and reassigning some write tasks to other available devices.

[0090] When the addressing interference coefficient Xzr of the i-th device i and the array conflict scheduling cost coefficient Cb i If all exceed the interference tolerance threshold X, it indicates that the device has an abnormal fault, is marked as a "high-risk conflict device", and a read / write prohibition instruction is generated, including: adding it to the fault monitoring queue, prohibiting the device from participating in all real-time write tasks, starting the redundancy backup strategy and data hot migration mechanism, and notifying the operation and maintenance system to perform in-depth diagnosis or replacement.

[0091] In this embodiment, step three, based on the first and second state datasets, integrates a deep learning model and a multi-dimensional interference modeling strategy. Through device behavior feature extraction and conflict scheduling coefficient construction, dynamic scheduling optimization of abnormal synchronous read / write device groups is achieved. Innovatively, device access behavior parameters and operating environment parameters are integrated into two key coefficients to measure head addressing stability and resource scheduling conflict cost, respectively. These coefficients balance hardware-level interference and system-level contention, providing clear quantitative indicators for conflict classification and strategy determination. By combining the addressing interference coefficient and scheduling cost coefficient with preset thresholds, four scheduling scenarios are formed, covering normal state, mild addressing interference, prominent resource conflict, and high-risk conflict devices. This ensures dynamic and refined scheduling strategies, effectively balancing system performance and stability. For devices with mild conflicts, flexible adjustments are made through methods such as reducing scheduling priority, controlling task proportion, and pre-allocating cache. For high-risk conflict devices, read / write prohibition, task transfer, and data hot migration are directly implemented to ensure the stability and reliability of critical data write paths. This step not only completes a closed-loop process from detection and analysis to judgment and execution, but also includes redundant backup, fault monitoring and hot migration modules, forming a device-level dynamic resource management system that takes into account real-time performance, predictability and stability.

[0092] Example 5

[0093] Step four includes:

[0094] S401. After executing the first and second scheduling policies generated in S304, the page block access heat during the execution process is tracked and analyzed, the long-term active page block regions are recorded, and the neighboring page disturbance coefficient Rd of the j-th page block of the i-th device is constructed. i,j Specifically:

[0095] S4011. The i-th device and the neighboring device m each have several page blocks. The number of accesses and access timestamps of each page block are recorded in real time. The access popularity of each page block is extracted. Based on the access popularity of the page blocks, the page block areas that have been activated for a long time are identified, and it is determined whether the data in these page block areas is hot data.

[0096] S4012. Calculate the neighbor page disturbance coefficient Rd of the j-th page block of the i-th device using the following formula. i,j :

[0097]

[0098] Among them, P i,j P represents the physical location of the j-th page block of the i-th device. m,l H represents the physical location of the l-th page block of the neighboring device m. i,jH represents the access frequency of the j-th page block of the i-th device. m,l This represents the access frequency of the l-th page block of the neighboring device m, |H i,j -H m,l The calculation represents the sum of the differences in access frequency between the j-th page block of the i-th device and all other page blocks in the system; |P i,j -P m,l The calculation is the difference in physical location distance between the j-th page block of the i-th device and the l-th page block of other devices m, where M represents the total number of all page blocks of all devices in the system;

[0099] S402. Set the security threshold Y, and set the neighboring page disturbance coefficient Rd of the j-th page block of the i-th device. i,j The data is compared with a security threshold Y to determine whether the j-th page block in the i-th device is being interfered with by other devices, including:

[0100] When Rd i,j When the value exceeds the safety threshold Y, it indicates abnormal page block interference between devices, and a third scheduling strategy is generated, including: migrating the hot data of the j-th page block of the i-th device to the idle device after addressing path optimization, with a migration ratio of 10%-30%, to reduce the risk of read disturbance;

[0101] When Rd i,j If the value is not higher than the safety threshold Y, it indicates that the page block interference between devices is normal, and the first scheduling strategy and the second scheduling strategy will continue to be executed.

[0102] This step, based on long-term tracking of device page block access popularity and proximity disturbance modeling, further enhances the system's ability to dynamically identify page-level data interference risks, realizes hot data migration optimization and continuous refinement of scheduling strategies, calculates disturbance coefficients combining page block access popularity and physical location distance, measures the potential interference impact between page blocks from two dimensions of spatial distribution and access intensity, fills the accuracy gap of traditional device-level or bus-level conflict identification, and achieves more granular data-level conflict detection.

[0103] By recording and analyzing the access frequency and timestamps of each page block, data page blocks with continuously active characteristics are effectively extracted, supporting accurate identification of hot data and providing clear triggering criteria and target selection conditions for hot data migration. The system calculates the differences in heat and location distribution between page blocks across devices to determine if there is a trend of multiple devices competing to access neighboring physical page blocks, which is particularly suitable for high-density deployment environments in array devices, shared storage, or virtualized resource scenarios. For device page blocks with a disturbance coefficient higher than the safety threshold Y, a third scheduling strategy is executed to migrate hot page blocks to other devices with better addressing paths and currently idle status, thereby reducing the risk of read disturbances and cache hit drops caused by data access conflicts and improving the overall read and write performance of the system.

[0104] Example 6

[0105] Please refer to Figure 2 A multi-device synchronous read / write system, comprising:

[0106] The equipment array step unit arranges multiple devices in an array and marks several devices.

[0107] The first acquisition unit is used to acquire the current head addressing position sequence of the i-th device and the read / write frequency of the i-th device within the time interval t, using an integrated I / O performance monitoring module, a head addressing tracker, a write check logger, a cache hit detector, and a synchronization clock module. Read / write access conflict count The most recent write failure rate Xr of the i-th device i and the current cache hit rate Hc i Generate the first state dataset;

[0108] The first state determination unit is used to make a preliminary judgment on the synchronous scheduling state in the device array based on the first state dataset. If it is determined that there is a high-frequency addressing or concurrent write conflict, it issues a corresponding warning instruction and constructs an abnormal synchronous read and write device group.

[0109] The second acquisition unit, upon receiving a group of abnormal synchronous read / write devices, collects real-time environmental factors of the i-th device, including the voltage fluctuation amplitude of the i-th device. Array bus utilization i Controller delay fluctuation coefficient CtrlVar i Equipment heat growth rate (GeatRate) i and I / O contention density i Generate the second state dataset;

[0110] The model building unit constructs a multi-device conflict recognition model based on a convolutional neural network, and trains it using the first-state dataset and the second-state dataset as input to calculate the addressing interference coefficient Xzr of the i-th device. i And array conflict scheduling cost coefficient Cb i ;

[0111] The second conflict determination unit is used to preset the interference tolerance threshold X. When any one of them exceeds the interference tolerance threshold X, it generates the first scheduling strategy or the second scheduling strategy.

[0112] The third acquisition unit is used to track and analyze the page block access heat during the execution process after executing the first or second scheduling strategy, record the long-term active page block region, and construct the neighboring page perturbation coefficient Rd of the j-th page block of the i-th device. i,j ;

[0113] The third interference determination unit presets a safety threshold Y to determine the neighboring page disturbance coefficient Rd of the j-th page block of the i-th device. i,j Does the j-th page block in the i-th device experience interference from other devices? If so, trigger the third scheduling strategy.

[0114] The control unit is used to execute the first scheduling strategy, the second scheduling strategy, and the third scheduling strategy.

[0115] A multi-device synchronous read / write system integrates multi-dimensional acquisition modules such as I / O performance monitoring, address tracking, and cache detection. Combined with a deep learning model, it achieves intelligent identification and intervention for device read / write conflicts, resulting in significant benefits. The system can perform layered detection and dynamic modeling of issues such as addressing anomalies, access conflicts, and cache failures in the device array. By calculating the addressing interference coefficient and conflict scheduling cost coefficient, it accurately assesses the risk of synchronous scheduling. After executing the scheduling strategy, the system further achieves real-time perception of fine-grained data interference based on page block heat tracking and adjacent page disturbance analysis, and dynamically triggers hot data migration strategies. This effectively reduces read disturbances and bus contention between devices, improving overall read / write efficiency and data consistency. This system is particularly suitable for high-concurrency, distributed storage, and high-reliability scenarios, possessing good scalability and intelligent adaptability.

[0116] It should be noted that all calculation formulas in this application employ, but are not limited to, regression analysis from machine learning algorithms to deeply analyze the collected parameters and identify their natural trends and interrelationships. Specialized software, such as Python's Scikit-learn library or the R language, is used to automatically generate mathematical models that match the data. Then, cross-validation and other methods are used to objectively evaluate the model performance, and continuous feedback and optimization are combined to ensure that the created formulas truly reflect the inherent laws of the data, thereby guaranteeing their validity and accuracy, and ensuring that the calculation process conforms to the constraints of natural laws, rather than being based on artificially set rules.

[0117] The technical solution of this invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of this invention.

[0118] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0119] It should be noted that the above 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 preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for synchronous reading and writing of multiple devices, characterized in that, The specific steps include: Step 1: Arrange the multi-channel devices in an array, and collect the current head addressing position sequence of the i-th device and the read / write frequency of the i-th device within the time interval t. Number of read / write access conflicts The most recent write failure rate of the i-th device and current cache hit rate Generate a first state dataset; and make a preliminary judgment on the synchronization scheduling status in the device array based on the first state dataset. If it is determined that there is a high-frequency addressing or concurrent write conflict, issue a corresponding warning instruction and construct an abnormal synchronous read and write device group. Step 2: After receiving the abnormal synchronous read / write device group, collect the operating environment factors of the i-th device in real time to generate the second state dataset; Step 3: Construct a multi-device conflict identification model based on a convolutional neural network, and train it using the first state dataset and the second state dataset as input to calculate the addressing interference coefficient of the i-th device. Array conflict scheduling cost coefficient And a preset interference tolerance threshold X is set. When any one of them exceeds the interference tolerance threshold X, a first scheduling policy or a second scheduling policy is generated; including: S301. Using a convolutional neural network (CNN), construct an initial CNN model. Train and test the initial CNN model using a first-state dataset and a second-state dataset. Use the trained initial CNN model as a multi-device conflict identification model. Simultaneously, use the intermediate layer output of the device operation state model as a feature vector to identify feature information. Train and test the multi-device conflict identification model using the acquired feature information. Use the trained multi-device conflict identification model as a data operation prediction model to analyze and obtain the addressing interference coefficient of the i-th device. Array conflict scheduling cost coefficient ; S302, Addressing interference coefficient of the i-th device The method of obtaining it is: Extract the total number of accesses of the i-th device in the first state dataset within the time interval t. Number of read / write access conflicts and current cache hit rate and the controller delay fluctuation coefficient in the second state dataset. After dimensionless processing, the addressing interference coefficient of the i-th device is calculated using the following formula. : ； In the formula, , and As weight, and The addressing interference coefficient of the i-th device The higher the value, the more severe the interference with the current head addressing of the device, which is more likely to cause performance degradation or warnings. S303, Array conflict scheduling cost coefficient of the i-th device The method of obtaining it is: Extract the array bus utilization of the i-th device in the second dataset. Equipment heat growth rate I / O contention density and voltage fluctuation amplitude After dimensionless processing, the array conflict scheduling cost coefficient of the i-th device is calculated using the following formula. : ； In the formula, , , and As weight, and The array conflict scheduling cost coefficient of the i-th device The higher the value, the more system resource contention or risk will occur if the device is scheduled under the current state. When the addressing interference coefficient of the i-th device Exceeding the interference tolerance threshold X, but the array conflict scheduling cost coefficient If the interference tolerance threshold X is not exceeded, it is determined that the device's head addressing is abnormally affected by external interference, and the first scheduling strategy is generated. When the addressing interference coefficient of the i-th device The interference tolerance threshold X was not exceeded, but the array conflict scheduling cost coefficient was... If the interference tolerance threshold X is exceeded, it is determined that the device has a significant abnormal risk in terms of resource contention, and a second scheduling strategy is generated. Step 4: After executing the first or second scheduling strategy, track and analyze the page block access heat during the execution process, record the long-term active page block regions, and construct the neighboring page perturbation coefficient of the j-th page block of the i-th device. A preset security threshold Y is used to determine the neighboring page disturbance coefficient of the j-th page block of the i-th device. Does the j-th page block in the i-th device experience interference from other devices? If so, trigger the third scheduling strategy.

2. The method for synchronous reading and writing of multiple devices according to claim 1, characterized in that: Step one includes: S101. Using an integrated I / O performance monitoring module, head addressing tracker, write check logger, cache hit detector, and synchronization clock module, real-time information is collected from the i-th device to obtain the following metrics: The current head addressing position sequence of the i-th device; The read / write frequency of the i-th device within the time interval t Specifically, it involves setting a time interval t, reading the total number of read and write requests within that time interval, and obtaining the number of read requests. and number of write requests Then, the number of read requests and number of write requests Add them together to get the total number of visits Divide the total number of accesses by the length of the time interval to obtain the read / write frequency. ; The number of read / write access conflicts for the i-th device within the time interval t. Specifically: record the conflict time of the i-th device, obtain the number of conflicts within the time interval t, and calculate it using the following formula: ； in, This represents the number of times each read / write request of the i-th device conflicts with other requests within the time interval t, and N represents the total number of devices participating in the statistics in the system, i.e., from the 1st device to the Nth device; The most recent write failure rate of the i-th device Specifically: ; in, This indicates the number of write failures. This indicates the total number of write attempts, obtained from the device logs; Current cache hit rate of the i-th device Specifically: ; in, Indicates the number of cache hits. This indicates the total number of cache requests; S102, Collect the current head addressing position sequence of the i-th device and the read / write frequency of the i-th device within the time interval t. Number of read / write access conflicts The most recent write failure rate of the i-th device and current cache hit rate It is updated in real time and stored as the first state dataset for subsequent judgment and early warning.

3. The method for synchronous reading and writing of multiple devices according to claim 2, characterized in that: Step one also includes: S103. Determine the operating status of the i-th device based on the first state dataset and issue a corresponding alarm command, including: If the addressing jump frequency per unit time in the current head addressing position sequence of the i-th device is significantly higher than twice the set average addressing frequency of the device, it indicates that the device has high-frequency fragment access or scheduling abnormality, and the first addressing abnormality warning instruction is generated; if the addressing jump frequency per unit time is not higher than twice the set average addressing frequency of the device, it indicates that there is no high-frequency access abnormality, and monitoring continues. When the read / write frequency of the i-th device within the time interval t If the frequency exceeds 200 times / second, it indicates that the device is under abnormally high load, generating the first high load warning command; if the read / write frequency of the i-th device within the time interval t... If the frequency does not exceed 200 times per second, it indicates that there is no abnormal load, and monitoring should continue. The number of read / write access conflicts of the i-th device within the time interval t If the number of read / write access conflicts exceeds 10% of the total number of accesses, it indicates that the device is at risk of frequent resource contention during access scheduling, and a first access conflict warning instruction is generated; if the number of read / write access conflicts of i devices within the time interval t... If the number of accesses does not exceed 10% of the total number of accesses, it means that the access conflict of the device is within the expected range, and continuous monitoring will continue. When the most recent write failure rate of the i-th device If the write failure rate exceeds 10%, it indicates a risk of unstable write operations, disk damage, or control errors, generating a first write failure warning command; if If the percentage is below 10%, it is considered normal; continued monitoring is necessary. When the current cache hit rate of the i-th device This indicates that the device's caching mechanism is malfunctioning, leading to frequent accesses to main memory and generating a first cache invalidation warning; if This indicates that the cache is working properly and will continue to be monitored. S104. Construct an abnormal synchronous read / write device group by aggregating the first addressing anomaly warning instruction, the first high load warning instruction, the first access conflict warning instruction, the first write failure warning instruction, and the first cache failure warning instruction.

4. The method for synchronous reading and writing of multiple devices according to claim 3, characterized in that: Step two includes: S201. Based on the unique device identifier carried in each of the first addressing anomaly warning instruction, first high load warning instruction, first access conflict warning instruction, first write failure warning instruction, and first cache failure warning instruction of the abnormal synchronous read / write device group, including device ID, logical address, or array location information; use the device management table or the device mapping table of the centralized controller to locate the device and mark the physical location, network address, and subsystem information of each target device in the abnormal synchronous read / write device group; S202, integrating voltage sensor, communication sensor, temperature sensor, controller response monitoring device, and I / O queue monitoring device, performs real-time operating environment information acquisition on the i-th device to obtain... Includes: the voltage fluctuation amplitude of the i-th device Array bus utilization Controller delay fluctuation coefficient Equipment heat growth rate and I / O contention density Generate the second state dataset.

5. The method for synchronous reading and writing of multiple devices according to claim 4, characterized in that: S202 includes: S2021. A voltage sensor is used to record the maximum value of the power supply voltage of the i-th device within the time interval t. and minimum value Calculate the difference between the two values ​​to obtain the voltage fluctuation amplitude of the i-th device. : ; S2022. Use a communication sensor to record the total time of data frames communicated by the i-th device on the bus, and divide it by the time interval t to obtain the array bus utilization rate of the i-th device. ; S2023. Using a controller response monitoring device, record the response time of multiple control commands, calculate the ratio of its standard deviation to the mean, and obtain the controller delay fluctuation coefficient of the i-th device. ; S2024. Use a temperature sensor module to periodically read the temperature data of the i-th device and calculate the heat growth rate of the device. The expression is: ； in, It is at the second time point The temperature of the i-th device at that time; It was at the first moment The temperature of the i-th device at that time; This indicates the interval between the second time point and the first time point; S2025. Using an I / O queue monitoring device, record the ratio of the total queuing time for device I / O requests to the number of requests within a time interval t, to obtain the I / O contention density of the i-th device. .

6. The method for synchronous reading and writing of multiple devices according to claim 5, characterized in that: Step three also includes: S304. Preset the interference tolerance threshold X, and set the addressing interference coefficient of the i-th device. and the array conflict scheduling cost coefficient of the i-th device By comparing with the interference tolerance threshold X, the following judgment results are obtained, including: When the addressing interference coefficient of the i-th device and array conflict scheduling cost coefficient If none of the values ​​exceed the interference tolerance threshold X, it means that the device is deemed to be in a qualified operating state, and the original scheduling strategy will be maintained so that it can continue to participate in synchronous read and write tasks and be marked as a valid scheduling device. When the addressing interference coefficient of the i-th device Exceeding the interference tolerance threshold X, but the array conflict scheduling cost coefficient If the interference tolerance threshold X is not exceeded, it is determined that the device's head addressing is abnormally affected by external interference. A first scheduling strategy is generated, including: reducing the scheduling priority of the device by 20%-30%, reducing the task scheduling of the device by 15%-20%, and pre-allocating 30%-50% of the write request cache of the device to other low-load devices. When the addressing interference coefficient of the i-th device The interference tolerance threshold X was not exceeded, but the array conflict scheduling cost coefficient was... If the interference tolerance threshold X is exceeded, it is determined that the device has a significant abnormal risk in terms of resource contention, and a second scheduling strategy is generated, including: reducing the scheduling priority of the device by 31%-40%, reducing the task scheduling of the device by 21%-30%, reducing the array bus usage of the device by 10%-20%, and reassigning some write tasks to other available devices. When the addressing interference coefficient of the i-th device and array conflict scheduling cost coefficient If all exceed the interference tolerance threshold X, it indicates that the device has an abnormal fault, is marked as a "high-risk conflict device", and a read / write prohibition instruction is generated, including: adding it to the fault monitoring queue, prohibiting the device from participating in all real-time write tasks, starting the redundancy backup strategy and data hot migration mechanism, and notifying the operation and maintenance system to perform in-depth diagnosis or replacement.

7. The method for synchronous reading and writing of multiple devices according to claim 6, characterized in that: Step four includes: S401. After executing the first and second scheduling policies generated in S304, the page block access heat during the execution process is tracked and analyzed, the long-term active page block regions are recorded, and the neighboring page disturbance coefficient of the j-th page block of the i-th device is constructed. Specifically: S4011. The i-th device and the neighboring device m each have several page blocks. The number of accesses and access timestamps of each page block are recorded in real time. The access popularity of each page block is extracted. Based on the access popularity of the page blocks, the page block areas that have been activated for a long time are identified, and it is determined whether the data in these page block areas is hot data. S4012. Calculate the neighboring page perturbation coefficient of the j-th page block of the i-th device using the following formula. : ； in, This represents the physical location of the j-th page block of the i-th device. Indicates the neighboring device m. The physical location of each page block This represents the access frequency of the j-th page block on the i-th device. Indicates the neighboring device m. The access popularity of each page block It calculates the sum of the differences in access frequency between the j-th page block of the i-th device and all other page blocks in the system; The calculation involves comparing the j-th page block of the i-th device with the page block of the m-th device. The difference in physical location distance between individual page blocks This represents the total number of page blocks across all devices in the system. S402. Set a security threshold Y, and set the neighboring page disturbance coefficient of the j-th page block of the i-th device. The data is compared with a security threshold Y to determine whether the j-th page block in the i-th device is being interfered with by other devices, including: when When the value exceeds the safety threshold Y, it indicates abnormal page block interference between devices, and a third scheduling strategy is generated, including: migrating the hot data of the j-th page block of the i-th device to the idle device after addressing path optimization, with a migration ratio of 10%-30%, to reduce the risk of read disturbance; when If the value is not higher than the safety threshold Y, it indicates that the page block interference between devices is normal, and the first scheduling strategy and the second scheduling strategy will continue to be executed.

8. A multi-device synchronous read / write system, applied to the multi-device synchronous read / write method according to any one of claims 1-7, characterized in that: include: The equipment array step unit arranges multiple devices in an array and marks several devices. The first acquisition unit is used to acquire the current head addressing position sequence of the i-th device and the read / write frequency of the i-th device within the time interval t, using an integrated I / O performance monitoring module, a head addressing tracker, a write check logger, a cache hit detector, and a synchronization clock module. Number of read / write access conflicts The most recent write failure rate of the i-th device and current cache hit rate Generate the first state dataset; The first state determination unit is used to make a preliminary judgment on the synchronous scheduling state in the device array based on the first state dataset. If it is determined that there is a high-frequency addressing or concurrent write conflict, it issues a corresponding warning instruction and constructs an abnormal synchronous read and write device group. The second acquisition unit, upon receiving a group of abnormal synchronous read / write devices, collects real-time environmental factors of the i-th device, including the voltage fluctuation amplitude of the i-th device. Array bus utilization Controller delay fluctuation coefficient Equipment heat growth rate and I / O contention density Generate the second state dataset; The model building unit constructs a multi-device conflict recognition model based on a convolutional neural network, and trains it using the first-state dataset and the second-state dataset as input to calculate the addressing interference coefficient of the i-th device. Array conflict scheduling cost coefficient ; The second conflict determination unit is used to preset the interference tolerance threshold X. When any one of them exceeds the interference tolerance threshold X, it generates the first scheduling strategy or the second scheduling strategy. The third acquisition unit is used to track and analyze the page block access heat during the execution process after executing the first or second scheduling strategy, record the long-term active page block regions, and construct the neighboring page perturbation coefficient of the j-th page block of the i-th device. ; The third interference determination unit presets a safety threshold Y to determine the neighboring page disturbance coefficient of the j-th page block of the i-th device. Does the j-th page block in the i-th device experience interference from other devices? If so, trigger the third scheduling strategy. The control unit is used to execute the first scheduling strategy, the second scheduling strategy, and the third scheduling strategy.

Images