WO2021088404A1 - 一种数据处理方法、装置、设备及可读存储介质 - Google Patents

一种数据处理方法、装置、设备及可读存储介质 Download PDF

Info

Publication number
WO2021088404A1
WO2021088404A1 PCT/CN2020/102020 CN2020102020W WO2021088404A1 WO 2021088404 A1 WO2021088404 A1 WO 2021088404A1 CN 2020102020 W CN2020102020 W CN 2020102020W WO 2021088404 A1 WO2021088404 A1 WO 2021088404A1
Authority
WO
WIPO (PCT)
Prior art keywords
data
state drive
period
solid state
solid
Prior art date
Application number
PCT/CN2020/102020
Other languages
English (en)
French (fr)
Inventor
王岩
李卫军
Original Assignee
深圳大普微电子科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 深圳大普微电子科技有限公司 filed Critical 深圳大普微电子科技有限公司
Publication of WO2021088404A1 publication Critical patent/WO2021088404A1/zh
Priority to US17/733,225 priority Critical patent/US20220253214A1/en

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
    • G06F3/06—Digital input from, or digital output to, record carriers, e.g. RAID, emulated record carriers or networked record carriers
    • G06F3/0601—Interfaces specially adapted for storage systems
    • G06F3/0602—Interfaces specially adapted for storage systems specifically adapted to achieve a particular effect
    • G06F3/061—Improving I/O performance
    • G06F3/0611—Improving I/O performance in relation to response time
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
    • G06F3/06—Digital input from, or digital output to, record carriers, e.g. RAID, emulated record carriers or networked record carriers
    • G06F3/0601—Interfaces specially adapted for storage systems
    • G06F3/0602—Interfaces specially adapted for storage systems specifically adapted to achieve a particular effect
    • G06F3/061—Improving I/O performance
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
    • G06F3/06—Digital input from, or digital output to, record carriers, e.g. RAID, emulated record carriers or networked record carriers
    • G06F3/0601—Interfaces specially adapted for storage systems
    • G06F3/0628—Interfaces specially adapted for storage systems making use of a particular technique
    • G06F3/0629—Configuration or reconfiguration of storage systems
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
    • G06F3/06—Digital input from, or digital output to, record carriers, e.g. RAID, emulated record carriers or networked record carriers
    • G06F3/0601—Interfaces specially adapted for storage systems
    • G06F3/0628—Interfaces specially adapted for storage systems making use of a particular technique
    • G06F3/0653—Monitoring storage devices or systems
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
    • G06F3/06—Digital input from, or digital output to, record carriers, e.g. RAID, emulated record carriers or networked record carriers
    • G06F3/0601—Interfaces specially adapted for storage systems
    • G06F3/0628—Interfaces specially adapted for storage systems making use of a particular technique
    • G06F3/0655—Vertical data movement, i.e. input-output transfer; data movement between one or more hosts and one or more storage devices
    • G06F3/0659—Command handling arrangements, e.g. command buffers, queues, command scheduling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
    • G06F3/06—Digital input from, or digital output to, record carriers, e.g. RAID, emulated record carriers or networked record carriers
    • G06F3/0601—Interfaces specially adapted for storage systems
    • G06F3/0668—Interfaces specially adapted for storage systems adopting a particular infrastructure
    • G06F3/0671—In-line storage system
    • G06F3/0673—Single storage device
    • G06F3/0679—Non-volatile semiconductor memory device, e.g. flash memory, one time programmable memory [OTP]

Definitions

  • This application relates to the field of computer technology, and in particular to a data processing method, device, equipment, and readable storage medium.
  • the management of solid-state drives is mostly determined by technical personnel based on their own experience. Therefore, the management of solid-state drives relies on the working experience and professional knowledge of technical personnel, and requires high technical personnel. At the same time, because the technical staff has more subjective opinions based on the management strategy summarized by experience, the management strategy may not be accurate enough. When there is a deviation in the management strategy, the management operation of the solid state drive will affect the external services provided by the solid state drive, which will reduce the read and write performance of the solid state drive, extend the response time to user operations, and reduce the service capability of the low solid state drive .
  • the purpose of this application is to provide a data processing method, device, device, and readable storage medium, so as to realize effective management of the solid state drive and improve the read and write performance of the solid state drive.
  • the specific plan is as follows:
  • this application provides a data processing method applied to a solid state drive, including:
  • historical I/O data is the data accessed by the solid state drive in a preset time period
  • the predictive results include: the data intensity of the SSD to be accessed in the future window period, and the future window period is determined according to the period during which the SSD is accessed;
  • the process of determining the future window period includes:
  • the period during which the solid state drive is accessed by the current business is determined as the future window period.
  • the process of determining the future window period includes:
  • the future window period is determined by the least common multiple, including:
  • it also includes:
  • using a predictive model to learn historical I/O data to obtain predictive results includes:
  • managing the solid state drive according to the prediction result includes:
  • this application provides a data processing device applied to a solid state drive, including:
  • the acquisition module is used to acquire historical I/O data;
  • the historical I/O data is the data accessed by the solid state drive in a preset time period;
  • the prediction module is used to learn historical I/O data using the prediction model to obtain prediction results;
  • the prediction results include: the data intensity of the solid state drive during the future window period, and the future window period is determined according to the period during which the solid state drive is accessed;
  • the management module is used to manage the solid state drive according to the prediction result.
  • this application provides a data processing device, including:
  • Memory used to store computer programs
  • the processor is used to execute a computer program to implement the data processing method disclosed above.
  • the present application provides a readable storage medium for storing a computer program, where the computer program is executed by a processor to implement the data processing method disclosed above.
  • this application provides a data processing method applied to a solid state drive, including: obtaining historical I/O data; historical I/O data is the data accessed by the solid state drive within a preset time period; utilization prediction
  • the model learns historical I/O data to obtain prediction results; the prediction results include: the data intensity of the solid state drive in the future window period, which is determined according to the period of the solid state drive being accessed; and the solid state drive is managed according to the predicted result.
  • this method uses the predictive model to learn the historical I/O data, so as to obtain the prediction results;
  • the prediction results include: the data intensity of the solid state drive in the future window period,
  • the data intensity can indicate the data read and write pressure of the solid state drive in the future window period; when the data intensity is greater, it indicates that the data read and write pressure of the solid state drive in the future window period is greater; when the data strength is small, it indicates the solid state drive In the future window period, the data read and write pressure is less.
  • the management of SSDs based on the forecast results includes: when the data read and write pressure of the SSDs in the future window period is high, suspend the low-priority business requests in the SSDs in the future window period; when the SSDs have data in the future window period When the read and write pressure is small, respond to low-priority service requests in the solid-state drive in a timely manner in the future window period, so as to prevent low-priority services from affecting the response of the solid-state drive to user read and write operations.
  • the future window period is determined according to the period in which the solid state drive is accessed, that is, the prediction result output by the prediction model is related to the period in which the solid state drive is accessed.
  • the access period has a certain rule, it provides a guarantee for the accuracy of the prediction result , So as to provide reliable data support for the effective management of solid state drives. Therefore, the present application realizes the effective management of the solid state drive, improves the read and write performance of the solid state drive, and avoids the delay of the user's operation response time, thereby improving the service capability of the solid state drive.
  • a data processing device, equipment, and readable storage medium provided in this application also have the above technical effects.
  • Figure 1 is a flow chart of a data processing method disclosed in this application.
  • Figure 2 is a structural diagram of a prediction model disclosed in this application.
  • Figure 3 is a structural diagram of another prediction model disclosed in this application.
  • FIG. 4 is a structural diagram of an LSTM Cell disclosed in this application.
  • FIG. 5 is a flowchart of a method for determining a future window period disclosed in this application.
  • Fig. 6 is a schematic diagram of a data processing device disclosed in this application.
  • Fig. 7 is a schematic diagram of a data processing device disclosed in this application.
  • the management strategy may not be accurate enough.
  • the management operation of the solid state drive will affect the external services provided by the solid state drive, which will reduce the read and write performance of the solid state drive, extend the response time to user operations, and reduce the service capability of the low solid state drive .
  • the present application provides a data processing solution, which can realize effective management of the solid state drive, so as to improve the read and write performance of the solid state drive.
  • an embodiment of the present application discloses a first data processing method, which is applied to a solid state hard disk, and includes:
  • the method of acquiring data may be: acquiring data through a software interface, or acquiring from a storage medium through a hardware interface, or receiving data sent by a data sending end through a network line, etc.
  • the historical I/O data is: data accessed by the solid state drive within a preset time period, that is, data involved when some objects perform read and/or write operations on the solid state drive during the preset time period.
  • Historical I/O data includes: I/O type (read or write), time stamp, I/O size (data size), etc.
  • the preset time period is the time period during which historical I/O data is generated. Generally, the preset time period is not less than the time length of the future window period. For example, taking the current moment as the limit, the preset time period is the first 10 seconds of the current moment, and the future window period is the last 10 seconds, the last 8 seconds, the last 5 seconds or less of the current moment.
  • the prediction model can predict the data and data intensity involved when certain objects read and/or write to the solid state drive during the future window period based on historical I/O data. Therefore, the prediction result includes the solid state drive in the future window period.
  • the future window period is a certain time period in the future.
  • the future window period is determined according to the period during which the solid state drive is accessed, that is, the prediction result output by the prediction model is associated with the period during which the solid state drive is accessed.
  • data intensity refers to: read I/O intensity, write I/O intensity, or read/write ratio.
  • the process of determining the future window period includes: if the solid state disk is accessed by only one service, determining the period during which the solid state disk is accessed by the current service as the future window period.
  • the forecast results can be set to include: the data and data involved when the current business performs read and/or write operations on the SSD
  • the amount of data, that is, the intensity of the data, the future window period is the period of the current business accessing the solid state drive.
  • the prediction result can include: the current business will read and/or write to the solid state drive in the next 10 seconds The data involved, and the intensity of the data.
  • the prediction interval can be set in the future window period, for example, 10 prediction points are set in the future window period, and each prediction point is predicted once for the read operation.
  • the prediction result Y ⁇ y1, y2,..., y10, y11,..., y20 ⁇ , where ⁇ y1, y2,..., y10 ⁇ is the predicted value for the read operation, ⁇ y11, y12 ,...,Y20 ⁇ are the predicted values for write operations. Since the current business periodically accesses the solid state drive, theoretically, the prediction model can predict the I/O data at the future time from the I/O data at the previous moment.
  • the starting point of the future window period can be the end point of the time period where the input data of the prediction model is located.
  • the time period of the input data of the prediction model is from the 13th to the 14th second
  • the future window period can be from the 14th second to the 15th second, where the starting point of the future window period is the 14th second, that is, the input The end of the time period of the data.
  • the total length of input data time and the number of sampling points can be increased.
  • the input of the original prediction model is: I/O data in the first 10 seconds, of which I/O data is collected every 5 seconds; if the cycle changes, then the input of the prediction model can be adjusted to: I/O data in the first 20 seconds /O data, which collects I/O data every 2 seconds.
  • the total time length of the input data (that is, the preset time period) can be equal to the future window period, so that the I/O data of each sampling point corresponds to the predicted value of the predicted point one-to-one.
  • the process of determining the future window period includes: if the solid state drive is accessed by multiple services, determining the period during which the solid state drive is accessed by each service; calculating the least common multiple of all periods, and determining the future through the least common multiple window period.
  • the period for business A to access the SSD is 10 seconds
  • the period for business B to access the SSD is 20 seconds
  • the period for business C to access the SSD is 30 seconds.
  • the least common multiple of all cycles is 60 seconds
  • the future window period is set to 60 seconds
  • the prediction result includes: the data involved when the solid state drive is read and/or written in the next 60 seconds, and the data intensity .
  • the business that accesses the solid-state hard disk may belong to the upper application layer, and specifically may belong to the application on the host side, such as a database.
  • using a prediction model to learn historical I/O data to obtain a prediction result includes: using an LSTM model to learn historical I/O data to obtain a prediction result. See Figure 2 for the structure of the prediction model. You can also set up a multi-layer structure, so that the predictive ability of the predictive model is stronger. The two-layer structure shown in Figure 3. Each box in Figure 3 is an LSTM Cell set. You can also set up more LSTM Cells accordingly to form a three-layer, four-layer, or even more layer network structure, thereby improving the prediction accuracy of the prediction model.
  • the LSTM model can be: seq2seq.
  • the write operation data of a certain service to the solid state drive in the previous 4 seconds is used as model input to predict the write operation of the service to the solid state drive in the next 4 seconds. Since the operation type and time have been specified here, the input sequence does not include variables such as timestamp and I/O type. 4 seconds is the future window period, and the prediction interval is 1 second.
  • the input-output relationship of the example in Figure 2 is: 75, 100, 125, and 100 in the example of the model input part in Figure 2 are the first 4 seconds, the first 3 seconds, the first 2 seconds, and the first 1 second, respectively.
  • V in the example of the model input part is the intermediate vector obtained by the model input processing.
  • 75, 50, 25, and 50 are the predicted values of the data intensity corresponding to the last 1 second, the last 2 seconds, the last 3 seconds, and the last 4 seconds respectively.
  • the input-output relationship illustrated in FIG. 3 is similar to the input-output relationship illustrated in FIG. 2, and reference may be made to the above description.
  • each box in Fig. 2 and Fig. 3 is a set of LSTM Cells, please refer to Fig. 4 for the structure of each LSTM Cell.
  • x t and h t-1 are the input data of the LSTM Cell
  • x t is the data to be processed by the current LSTM Cell
  • h t-1 is the output result of the previous LSTM Cell.
  • the input data also includes the previous LSTM Cell.
  • the output c t-1 are the four processing gates of the current LSTM Cell, and these four processing gates process x t and h t-1 respectively .
  • FIG. 4 Represents the vector matrix multiplication operation, Means addition, Represents Hadamard Product (Hadamard Product), Represents the activation function.
  • W x is the weight value of x t in the current processing gate
  • W h is the weight value of h t-1 in the current processing gate.
  • W xi is the weight value of x t in the “Input Gate”
  • W hi is the weight value of h t-1 “Input Gate”
  • the training process of the LSTM model includes: acquiring historical I/O data, using the historical I/O data as the training data of the LSTM model, and the future window period of the LSTM model is determined by the period of business access to the solid state drive.
  • the LSTM model can input the data involved when the business performs read and/or write operations on the solid state drive, and output the corresponding prediction results, and the prediction results include: The data and data intensity involved in the read and/or write operation of the solid state drive.
  • managing the solid state drive according to the prediction result includes: performing cache management and/or garbage collection on the solid state drive according to the prediction result.
  • the embodiment of the present application uses the predictive model to learn the historical I/O data, so as to obtain the predictive result;
  • the predictive result includes: the data accessed by the solid state drive in the future window period Strength, the data strength can indicate the data read and write pressure of the solid state drive in the future window period; when the data strength is greater, it indicates that the data read and write pressure of the solid state drive in the future window period is greater; when the data strength is small, it indicates The data read and write pressure of SSDs in the future window period is relatively small.
  • the management of SSDs based on the forecast results includes: when the data read and write pressure of the SSDs in the future window period is high, suspend the low-priority business requests in the SSDs in the future window period; when the SSDs have data in the future window period When the read and write pressure is small, respond to low-priority service requests in the solid-state drive in a timely manner in the future window period, so as to prevent low-priority services from affecting the response of the solid-state drive to user read and write operations.
  • the future window period is determined according to the period in which the solid state drive is accessed, that is, the prediction result output by the prediction model is related to the period in which the solid state drive is accessed.
  • the access period has a certain rule, it provides a guarantee for the accuracy of the prediction result , So as to provide reliable data support for the effective management of solid state drives. Therefore, the present application realizes the effective management of the solid state drive, improves the read and write performance of the solid state drive, and avoids the delay of the user's operation response time, thereby improving the service capability of the solid state drive.
  • Figure 5 discloses a method for determining the future window period, including:
  • S501 If the solid state drive is accessed by multiple services, determine the period during which the solid state drive is accessed by each service;
  • S502 Determine the maximum period of all periods as the lower limit of the value of the future window period
  • S504. Determine any target value that is not less than the lower limit of the value and not greater than the upper limit of the value as the future window period.
  • multiple services that access the SSD may belong to the same system or different systems. If these multiple services belong to the target system, then the period of different services in the target system accessing the solid state drive is different. Therefore, the least common multiple of each period may be relatively large. If the least common multiple is directly determined as the future window period, it will be Cause the future window to be too long. To prevent the future window period from being too long, this embodiment determines any target value that is not less than the lower limit of the value and not greater than the upper limit of the value as the future window period.
  • the period of business A accessing the SSD is 10 seconds
  • the period of business B accessing the SSD is 20 seconds
  • the period of business C accessing the SSD is 30 seconds
  • the least common multiple of all cycles is 60 seconds.
  • it further includes: calculating the greatest common divisor of all periods, and determining the greatest common divisor as the prediction interval; determining multiple prediction points in the future window period according to the prediction interval; wherein the prediction point is the same as the prediction value.
  • the predicted result includes the predicted value.
  • the future window period is determined to be 60 seconds based on services of A, B, and C, then the greatest common divisor of 10 seconds, 20 seconds, and 30 seconds is determined as the prediction interval.
  • 6 prediction points can be determined, and these 6 prediction points correspond to the 10th, 20th, 30th, 40th, 50th, and 60th second respectively, then each prediction point can correspond to one Predicted value, and the predicted result can include 6 predicted values. According to these six prediction values, it can be determined whether the corresponding prediction point is suitable for the management of the solid state drive.
  • Each predicted value is the data intensity at the time of the predicted point.
  • a suitable future window period can be determined according to the period of the target system accessing the solid state drive. If the future window period is too long, the pressure on the forecasting model will be greater, and the forecast results will not have reference value; the future window period is too short, and the forecast results will be meaningless due to the inability to operate in time.
  • a suitable future window period can be determined according to the actual access situation of the target system to the solid state drive, so that the accuracy of the prediction result can be improved, thereby effectively managing the current solid state drive and improving the readability of the solid state drive. Write performance provides a reliable basis.
  • the following describes a data processing device provided by an embodiment of the present application.
  • the data processing device described below and the data processing method described above can be cross-referenced.
  • an embodiment of the present application discloses a data processing device applied to a solid state hard disk, including:
  • the acquiring module 601 is used to acquire historical I/O data; the historical I/O data is data accessed by the solid state hard disk within a preset time period;
  • the prediction module 602 is used for learning historical I/O data using the prediction model to obtain prediction results; the prediction results include: the data intensity of the solid state drive being accessed in the future window period, and the future window period is determined according to the period during which the solid state drive is accessed ;
  • the management module 603 is used to manage the solid state disk according to the prediction result.
  • determining the future window period further includes: a determining module for determining the future window period, and the determining module is specifically configured to:
  • the period during which the solid state drive is accessed by the current business is determined as the future window period.
  • the determining module includes:
  • the first determining unit is configured to determine the period during which the solid state drive is accessed by each service if the solid state drive is accessed by multiple services;
  • the first calculation unit is used to calculate the least common multiple of all periods, and determine the future window period through the least common multiple.
  • the calculation unit includes:
  • the first determining subunit is used to determine the maximum period of all periods as the lower limit of the value of the future window period
  • the second determining subunit is used to determine the least common multiple as the upper limit of the value of the future window period
  • the third determining subunit is used to determine any target value that is not less than the lower limit of the value and not greater than the upper limit of the value as the future window period.
  • the determining module further includes:
  • the second calculation unit is used to calculate the greatest common divisor of all periods, and determine the greatest common divisor as the prediction interval;
  • the second determining unit is used to determine multiple prediction points in the future window period according to the prediction interval.
  • the prediction module is specifically used to:
  • the management module is specifically used for:
  • this embodiment provides a data processing device that realizes effective management of solid state drives, improves the read and write performance of solid state drives, avoids delays in user operation response time, and thereby improves the service capability of solid state drives.
  • the following describes a data processing device provided by an embodiment of the present application.
  • the data processing device described below and the data processing method and device described above can be cross-referenced.
  • an embodiment of the present application discloses a data processing device, including:
  • the memory 701 is used to store computer programs
  • the processor 702 is configured to execute the computer program to implement the method disclosed in any of the foregoing embodiments.
  • the following describes a readable storage medium provided by an embodiment of the present application.
  • the readable storage medium described below and the data processing method, device, and device described above can be cross-referenced.
  • a readable storage medium used to store a computer program, where the computer program implements the data processing method disclosed in the foregoing embodiment when the computer program is executed by a processor.
  • the computer program implements the data processing method disclosed in the foregoing embodiment when the computer program is executed by a processor.
  • the steps of the method or algorithm described in combination with the embodiments disclosed in this document can be directly implemented by hardware, a software module executed by a processor, or a combination of the two.
  • the software module can be placed in random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disks, removable disks, CD-ROMs, or all areas in the technical field. Any other form of well-known readable storage medium.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Human Computer Interaction (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Debugging And Monitoring (AREA)

Abstract

一种数据处理方法、装置、设备及可读存储介质。方法包括：获取历史I/O数据（S101）；历史I/O数据为固态硬盘在预设时间段内被访问的数据；利用预测模型对历史I/O数据进行学习，获得预测结果(S102)；预测结果包括：固态硬盘在未来窗口期内被访问的数据强度，未来窗口期根据固态硬盘被访问的周期确定；根据预测结果管理固态硬盘(S103)。实现了固态硬盘的有效管理，提高了固态硬盘的读写性能，避免了用户操作响应时间的延时，从而提高了固态硬盘的服务能力。

Description

一种数据处理方法、装置、设备及可读存储介质
本申请要求于2019年11月06日提交至中国专利局、申请号为201911077491.9、发明名称为“一种数据处理方法、装置、设备及可读存储介质”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及计算机技术领域，特别涉及一种数据处理方法、装置、设备及可读存储介质。
背景技术
在现有技术中，对固态硬盘的管理大多是技术人员根据自身经验确定，因此固态硬盘的管理依赖于技术人员的工作经验和专业知识，对技术人员要求较高。同时，由于技术人员按照经验总结的管理策略带有更多主观意见，导致管理策略可能不够准确。当管理策略存在偏差时，固态硬盘的管理操作会与固态硬盘对外提供的服务相互影响，如此会降低固态硬盘的读写性能，延长对用户操作的响应时间，从而降低了低固态硬盘的服务能力。
因此，如何实现固态硬盘的有效管理，以提高固态硬盘的读写性能，是本领域技术人员需要解决的问题。
发明内容
有鉴于此，本申请的目的在于提供一种数据处理方法、装置、设备及可读存储介质，以实现固态硬盘的有效管理，以提高固态硬盘的读写性能。其具体方案如下：
第一方面，本申请提供了一种数据处理方法，应用于固态硬盘，包括：
获取历史I/O数据；历史I/O数据为固态硬盘在预设时间段内被访问的数据；
利用预测模型对历史I/O数据进行学习，获得预测结果；预测结果包括：固态硬盘在未来窗口期内被访问的数据强度，未来窗口期根据固态硬盘被访问的周期确定；
根据预测结果管理固态硬盘。
优选地，未来窗口期的确定过程包括：
若固态硬盘仅被一个业务访问，则将固态硬盘被当前业务访问的周期确定为未来窗口期。
优选地，未来窗口期的确定过程包括：
若固态硬盘被多个业务访问，则确定固态硬盘被每个业务访问的周期；
计算所有周期的最小公倍数，并通过最小公倍数确定未来窗口期。
优选地，通过最小公倍数确定未来窗口期，包括：
将所有周期中的最大周期确定为未来窗口期的取值下限；
将最小公倍数确定为未来窗口期的取值上限；
将任一个不小于取值下限且不大于取值上限的目标值确定为未来窗口期。
优选地，还包括：
计算所有周期的最大公约数，并将最大公约数确定为预测间隔；
按照预测间隔在未来窗口期中确定多个预测点。
优选地，利用预测模型对历史I/O数据进行学习，获得预测结果，包括：
利用LSTM模型对历史I/O数据进行学习，获得预测结果。
优选地，根据预测结果管理固态硬盘，包括：
根据预测结果对固态硬盘进行缓存管理和/或垃圾回收。
第二方面，本申请提供了一种数据处理装置，应用于固态硬盘，包括：
获取模块，用于获取历史I/O数据；历史I/O数据为固态硬盘在预设时间段内被访问的数据；
预测模块，用于利用预测模型对历史I/O数据进行学习，获得预测结果；预测结果包括：固态硬盘在未来窗口期内被访问的数据强度，未来窗口期根据固态硬盘被访问的周期确定；
管理模块，用于根据预测结果管理固态硬盘。
第三方面，本申请提供了一种数据处理设备，包括：
存储器，用于存储计算机程序；
处理器，用于执行计算机程序，以实现前述公开的数据处理方法。
第四方面，本申请提供了一种可读存储介质，用于保存计算机程序，其中，计算机程序被处理器执行时实现前述公开的数据处理方法。
通过以上方案可知，本申请提供了一种数据处理方法，应用于固态硬盘，包括：获取历史I/O数据；历史I/O数据为固态硬盘在预设时间段内被访问的数据；利用预测模型对历史I/O数据进行学习，获得预测结果；预测结果包括：固态硬盘在未来窗口期内被访问的数据强度，未来窗口期根据固态硬盘被访问的周期确定；根据预测结果管理固态硬盘。
可见，该方法在获取到历史I/O数据后，利用预测模型对历史I/O数据进行学习，从而可获得预测结果；预测结果中包括：固态硬盘在未来窗口期内被访问的数据强度，该数据强度可表明固态硬盘在未来窗口期的数据读写压力；当该数据强度较大时，表明固态硬盘在未来窗口期的数据读写压力较大；当该数据强度较小时，表明固态硬盘在未来窗口期的数据读写压力较小。因此根据预测结果管理固态硬盘包括：当固态硬盘在未来窗口期的数据读写压力较大时，在未来窗口期内暂缓固态硬盘内的低优先级业务请求；当固态硬盘在未来窗口期的数据读写压力较小时，在未来窗口期内及时响应固态硬盘内的低优先级业务请求，从而可避免低优先级业务影响固态硬盘对用户读写操作的响应。其中，未来窗口期根据固态硬盘被访问的周期确定，也就是预测模型输出的预测结果与固态硬盘被访问的周期相关联，由于访问周期具有一定的规律，因此为预测结果的准确性提供了保障，从而为固态硬盘的有效管理提供了可靠的数据支持。因此本申请实现了固态硬盘的有效管理，提高了固态硬盘的读写性能，避免了用户操作响应时间的延时，从而提高了固态硬盘的服务能力。
相应地，本申请提供的一种数据处理装置、设备及可读存储介质，也同样具有上述技术效果。
附图说明
为了更清楚地说明本申请实施例或现有技术中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本申请的实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据提供的附图获得其他的附图。
图1为本申请公开的一种数据处理方法流程图；
图2为本申请公开的一种预测模型结构图；
图3为本申请公开的另一种预测模型结构图；
图4为本申请公开的一种LSTM Cell结构图；
图5为本申请公开的一种未来窗口期的确定方法流程图；
图6为本申请公开的一种数据处理装置示意图；
图7为本申请公开的一种数据处理设备示意图。
具体实施方式
下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本申请一部分实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本申请保护的范围。
目前，由于技术人员按照经验总结的管理策略带有更多主观意见，导致管理策略可能不够准确。当管理策略存在偏差时，固态硬盘的管理操作会与固态硬盘对外提供的服务相互影响，如此会降低固态硬盘的读写性能，延长对用户操作的响应时间，从而降低了低固态硬盘的服务能力。为此，本申请提供了一种数据处理方案，能够实现固态硬盘的有效管理，以提高固态硬盘的读写性能。
参见图1所示，本申请实施例公开了第一种数据处理方法，应用于固态硬盘，包括：
S101、获取历史I/O数据；
需要说明的是，获取数据的方式可以为：通过软件接口获取数据，或通过硬件接口从存储介质中获取，或接收数据发送端通过网络线路发送的数据等。其中，历史I/O数据为：固态硬盘在预设时间段内被访问的数据，也就是在预设时间段内某些对象对固态硬盘进行读和/或写操时涉及的数据。历史I/O数据包括：I/O类型(读或写)、时间戳、I/O大小(数据量大小)等。其中，预设时间段即产生历史I/O数据的时间段，一般情况下，预设时间段不小于未来窗口期的时间长度。如：以当前时刻为界限，预设 时间段为当前时刻的前10秒，未来窗口期为当前时刻的后10秒、后8秒、后5秒或更短时间。
S102、利用预测模型对历史I/O数据进行学习，获得预测结果；
具体的，预测模型可根据历史I/O数据预测未来窗口期内某些对象对固态硬盘进行读和/或写操时涉及的数据和数据强度，因此预测结果中包括固态硬盘在未来窗口期内被访问的数据强度。未来窗口期即未来某一个时间段。未来窗口期根据固态硬盘被访问的周期确定，也就是预测模型输出的预测结果与固态硬盘被访问的周期相关联。其中，数据强度指：读I/O强度、写I/O强度或读写比例。
在一种具体实施方式中，未来窗口期的确定过程包括：若固态硬盘仅被一个业务访问，则将固态硬盘被当前业务访问的周期确定为未来窗口期。具体的，在预测模型的训练过程中，将固态硬盘被当前业务访问的周期确定为未来窗口期，那么可设置预测结果包括：当前业务对固态硬盘进行读和/或写操时涉及的数据和数据量，也就是数据强度，其中的未来窗口期为当前业务访问固态硬盘的周期。例如：当前业务访问固态硬盘的周期为每10秒访问一次，那么未来窗口期即设置为10秒，那么预测结果可以包括：当前业务在未来10秒内对固态硬盘进行读和/或写操时涉及的数据，以及数据强度。
具体的，在将当前业务访问固态硬盘的周期确定为未来窗口期后，在未来窗口期中可设置预测间隔，如：在未来窗口期中设置10个预测点，每个预测点针对读操作预测一次，针对写操作预测一次，则预测结果Y＝{y1，y2，…，y10，y11，…，y20}，其中，{y1，y2，…，y10}为针对读操作的预测值，{y11，y12，…，y20}为针对写操作的预测值。由于当前业务是周期性地访问固态硬盘，因此理论上，预测模型可以由前一时刻的I/O数据预测未来时刻的I/O数据。实际应用过程中，预测模型的输入可选取更多I/O数据，如：一次样本输入为X＝{x1，x2，…，x8}，X为当前业务访问固态硬盘的历史I/O数据，xi表示每一I/O数据的三元变量(时间戳、I/O类型、I/O大小)。预设时间段设置的越长，预测模型的输入也就越多。
其中，未来窗口期的起点可以为预测模型的输入数据所处时间段的终 点。如：预测模型的输入数据所处时间段为第13秒到第14秒，那么未来窗口期可以为第14秒到第15秒，其中，未来窗口期的起点为第14秒开始，也就是输入数据的所处时间段的终点。
若当前业务访问固态硬盘的周期变化时，可以增加输入数据的时间总长度以及采样点数量。例如：原来预测模型的输入为：前10秒内的I/O数据，其中每5秒采集一次I/O数据；若周期发生变化，那么预测模型的输入可调整为：前20秒内的I/O数据，其中每2秒采集一次I/O数据。其中，输入数据的时间总长度(即预设时间段)可以与未来窗口期相等，这样每个采样点的I/O数据与预测点的预测值一一对应。
在一种具体实施方式中，未来窗口期的确定过程包括：若固态硬盘被多个业务访问，则确定固态硬盘被每个业务访问的周期；计算所有周期的最小公倍数，并通过最小公倍数确定未来窗口期。具体的，假设有3个业务按照各自的周期访问固态硬盘，其中，A业务访问固态硬盘的周期为10秒，B业务访问固态硬盘的周期为20秒，C业务访问固态硬盘的周期为30秒，那么所有周期的最小公倍数为60秒，因此将未来窗口期即设置为60秒，那么预测结果包括：在未来60秒内固态硬盘被进行读和/或写操时涉及的数据，以及数据强度。
需要说明的是，访问固态硬盘的业务可以属于上层应用层，具体可以属于主机端的应用，如数据库等。
在本实施例中，利用预测模型对历史I/O数据进行学习，获得预测结果，包括：利用LSTM模型对历史I/O数据进行学习，获得预测结果。预测模型的结构请参见图2。还可以设置多层结构，这样预测模型的预测能力更强。如图3所示的两层结构。图3中的每个框即为一个LSTM Cell集合。还可以据此设置更多LSTM Cell，以形成三层、四层、甚至更多层的网络结构，从而提高预测模型的预测精度。LSTM模型可以为：seq2seq。
在图2中，以前4秒某业务对固态硬盘写操作数据为模型输入，以预测该业务对固态硬盘后4秒的写操作情况。由于这里已经指定了操作类型和时间，因此输入序列不包括时间戳、I/O类型等变量。4秒即为未来窗口期，其中的预测间隔为1秒。
具体的，以当前时刻为界限，图2示例的输入输出关系为：图2中模 型输入部分示例的75、100、125、100分别为前4秒、前3秒、前2秒、前1秒对应的数据强度，模型输入部分示例的V为模型输入处理得到的中间向量。图2中的模型输出部分示例的75、50、25、50分别为后1秒、后2秒、后3秒、后4秒分别对应的数据强度的预测值。图3中示例的输入输出关系与图2示例的输入输出关系类似，可参照上述说明。
其中，图2和图3中的每个框为一个LSTM Cell集合，每个LSTM Cell的结构请参见图4。在图4中，x t和h t-1为LSTM Cell的输入数据，x t为当前LSTM Cell待处理数据，h t-1是前一个LSTM Cell输出的结果，输入数据还包括前一个LSTM Cell输出的c t-1。图4中“Input Gate”、“Forget Gate”、“Cell Gate”和“Output Gate”为当前LSTM Cell的四个处理门，这四个处理门分别处理x t和h t-1。“Input Gate”的处理结果i t与“Cell Gate”的处理结果c t进行内积运算，获得第一个内积结果，“Forget Gate”的处理结果f t与前一个LSTM Cell输出的c t-1进行内积运算，获得第二个内积结果，进而第一个内积结果与第二个内积结果相加，获得一个新的c t，这个新的c t经过激活函数后，与“Output Gate”的处理结果o t进行内积，从而获得当前LSTM Cell输出的h t。其中，当前LSTM Cell输出的h t和c t会同时输入下一个LSTM Cell。
图4中的
Figure PCTCN2020102020-appb-000001
表示向量矩阵乘法运算，
Figure PCTCN2020102020-appb-000002
表示相加，
Figure PCTCN2020102020-appb-000003
表示哈达玛积(Hadamard Product)，
Figure PCTCN2020102020-appb-000004
表示激活函数。W x为x t在当前处理门中的权重值，W h为h t-1当前处理门中的权重值。例如，W xi为x t在“Input Gate”中的权重值，W hi为h t-1“Input Gate”中的权重值，其他以此类推。LSTM中的每个LSTM Cell中有四个处理门。
LSTM模型的训练过程包括：获取历史I/O数据，将历史I/O数据作为LSTM模型的训练数据，LSTM模型的未来窗口期由业务访问固态硬盘的周期确定。训练完成后，该LSTM模型便可以将该业务对固态硬盘进行读和/或写操时涉及的数据为输入，并输出相应的预测结果，且该预测结果包括：在未来窗口期内该业务对固态硬盘进行读和/或写操时涉及的数据和数据强度。
S103、根据预测结果管理固态硬盘。
在本实施例中，根据预测结果管理固态硬盘，包括：根据预测结果对 固态硬盘进行缓存管理和/或垃圾回收。
可见，本申请实施例在获取到历史I/O数据后，利用预测模型对历史I/O数据进行学习，从而可获得预测结果；预测结果中包括：固态硬盘在未来窗口期内被访问的数据强度，该数据强度可表明固态硬盘在未来窗口期的数据读写压力；当该数据强度较大时，表明固态硬盘在未来窗口期的数据读写压力较大；当该数据强度较小时，表明固态硬盘在未来窗口期的数据读写压力较小。因此根据预测结果管理固态硬盘包括：当固态硬盘在未来窗口期的数据读写压力较大时，在未来窗口期内暂缓固态硬盘内的低优先级业务请求；当固态硬盘在未来窗口期的数据读写压力较小时，在未来窗口期内及时响应固态硬盘内的低优先级业务请求，从而可避免低优先级业务影响固态硬盘对用户读写操作的响应。其中，未来窗口期根据固态硬盘被访问的周期确定，也就是预测模型输出的预测结果与固态硬盘被访问的周期相关联，由于访问周期具有一定的规律，因此为预测结果的准确性提供了保障，从而为固态硬盘的有效管理提供了可靠的数据支持。因此本申请实现了固态硬盘的有效管理，提高了固态硬盘的读写性能，避免了用户操作响应时间的延时，从而提高了固态硬盘的服务能力。
请参见图5，图5公开了一种未来窗口期的确定方法，包括：
S501、若固态硬盘被多个业务访问，则确定固态硬盘被每个业务访问的周期；
S502、将所有周期中的最大周期确定为未来窗口期的取值下限；
S503、计算所有周期的最小公倍数，并将最小公倍数确定为未来窗口期的取值上限；
S504、将任一个不小于取值下限且不大于取值上限的目标值确定为未来窗口期。
需要说明的是，访问固态硬盘的多个业务可能属于同一系统，也可能属于不同系统。若这多个业务均属于目标系统，那么目标系统中的不同业务访问固态硬盘的周期长短不一，因此各个周期的最小公倍数可能会比较大，若直接将最小公倍数确定为未来窗口期，将会导致未来窗口期过长。为避免未来窗口期过长，本实施例将任一个不小于取值下限且不大于取值 上限的目标值确定为未来窗口期。
例如：假设目标系统中有3个业务按照各自的周期访问固态硬盘，其中，A业务访问固态硬盘的周期为10秒，B业务访问固态硬盘的周期为20秒，C业务访问固态硬盘的周期为30秒，那么所有周期的最小公倍数为60秒，此时可在30秒～60秒之间(包括端点值)任取一个值(如30秒、40秒、45秒等)，进而将此取值确定为未来窗口期。
在一种具体实施方式中，还包括：计算所有周期的最大公约数，并将最大公约数确定为预测间隔；按照预测间隔在未来窗口期中确定多个预测点；其中，预测点与预测值一一对应，预测结果包括预测值。
在上述示例的基础上，若根据A、B、C业务确定未来窗口期为60秒，则将访问周期10秒、20秒和30秒的最大公约数10秒确定为预测间隔，那么在60秒(0-60秒)的未来窗口期内，可确定6个预测点，这6个预测点与第10、20、30、40、50、60秒分别对应，那么每个预测点即可对应一个预测值，而预测结果中就可以包括6个预测值。根据这6个预测值可分别确定在相应的预测点是否适合进行固态硬盘的管理。每个预测值为预测点所处时刻的数据强度。
由上可见，本实施例根据目标系统访问固态硬盘的周期，可确定合适的未来窗口期。未来窗口期过长，预测模型的压力会比较大，预测结果也不具有可参考价值；未来窗口期过短，由于无法及时操作，导致预测结果没有太大意义。按照本实施例提供的方法，可根据目标系统对固态硬盘的实际访问情况，确定合适的未来窗口期，从而可提高预测结果的精确度，从而为现固态硬盘的有效管理，提高固态硬盘的读写性能提供可靠依据。
下面对本申请实施例提供的一种数据处理装置进行介绍，下文描述的一种数据处理装置与上文描述的一种数据处理方法可以相互参照。
参见图6所示，本申请实施例公开了一种数据处理装置，应用于固态硬盘，包括：
获取模块601，用于获取历史I/O数据；历史I/O数据为固态硬盘在预设时间段内被访问的数据；
预测模块602，用于利用预测模型对历史I/O数据进行学习，获得预 测结果；预测结果包括：固态硬盘在未来窗口期内被访问的数据强度，未来窗口期根据固态硬盘被访问的周期确定；
管理模块603，用于根据预测结果管理固态硬盘。
在一种具体实施方式中，还包括：用于确定未来窗口期的确定模块，确定模块具体用于：
若固态硬盘仅被一个业务访问，则将固态硬盘被当前业务访问的周期确定为未来窗口期。
在一种具体实施方式中，确定模块包括：
第一确定单元，用于若固态硬盘被多个业务访问，则确定固态硬盘被每个业务访问的周期；
第一计算单元，用于计算所有周期的最小公倍数，并通过最小公倍数确定未来窗口期。
在一种具体实施方式中，计算单元包括：
第一确定子单元，用于将所有周期中的最大周期确定为未来窗口期的取值下限；
第二确定子单元，用于将最小公倍数确定为未来窗口期的取值上限；
第三确定子单元，用于将任一个不小于取值下限且不大于取值上限的目标值确定为未来窗口期。
在一种具体实施方式中，确定模块还包括：
第二计算单元，用于计算所有周期的最大公约数，并将最大公约数确定为预测间隔；
第二确定单元，用于按照预测间隔在未来窗口期中确定多个预测点。
在一种具体实施方式中，预测模块具体用于：
利用LSTM模型对历史I/O数据进行学习，获得预测结果。
在一种具体实施方式中，管理模块具体用于：
根据预测结果对固态硬盘进行缓存管理和/或垃圾回收。
其中，关于本实施例中各个模块、单元更加具体的工作过程可以参考前述实施例中公开的相应内容，在此不再进行赘述。
可见，本实施例提供了一种数据处理装置，该装置实现了固态硬盘的有效管理，提高了固态硬盘的读写性能，避免了用户操作响应时间的延时， 从而提高了固态硬盘的服务能力。
下面对本申请实施例提供的一种数据处理设备进行介绍，下文描述的一种数据处理设备与上文描述的一种数据处理方法及装置可以相互参照。
参见图7所示，本申请实施例公开了一种数据处理设备，包括：
存储器701，用于保存计算机程序；
处理器702，用于执行所述计算机程序，以实现上述任意实施例公开的方法。
下面对本申请实施例提供的一种可读存储介质进行介绍，下文描述的一种可读存储介质与上文描述的一种数据处理方法、装置及设备可以相互参照。
一种可读存储介质，用于保存计算机程序，其中，所述计算机程序被处理器执行时实现前述实施例公开的数据处理方法。关于该方法的具体步骤可以参考前述实施例中公开的相应内容，在此不再进行赘述。
本申请涉及的“第一”、“第二”、“第三”、“第四”等(如果存在)是用于区别类似的对象，而不必用于描述特定的顺序或先后次序。应该理解这样使用的数据在适当情况下可以互换，以便这里描述的实施例能够以除了在这里图示或描述的内容以外的顺序实施。此外，术语“包括”和“具有”以及他们的任何变形，意图在于覆盖不排他的包含，例如，包含了一系列步骤或单元的过程、方法或设备不必限于清楚地列出的那些步骤或单元，而是可包括没有清楚地列出的或对于这些过程、方法或设备固有的其它步骤或单元。
需要说明的是，在本申请中涉及“第一”、“第二”等的描述仅用于描述目的，而不能理解为指示或暗示其相对重要性或者隐含指明所指示的技术特征的数量。由此，限定有“第一”、“第二”的特征可以明示或者隐含地包括至少一个该特征。另外，各个实施例之间的技术方案可以相互结合，但是必须是以本领域普通技术人员能够实现为基础，当技术方案的结合出现相互矛盾或无法实现时应当认为这种技术方案的结合不存在，也不在本 申请要求的保护范围之内。
本说明书中各个实施例采用递进的方式描述，每个实施例重点说明的都是与其它实施例的不同之处，各个实施例之间相同或相似部分互相参见即可。
结合本文中所公开的实施例描述的方法或算法的步骤可以直接用硬件、处理器执行的软件模块，或者二者的结合来实施。软件模块可以置于随机存储器(RAM)、内存、只读存储器(ROM)、电可编程ROM、电可擦除可编程ROM、寄存器、硬盘、可移动磁盘、CD-ROM、或技术领域内所公知的任意其它形式的可读存储介质中。
本文中应用了具体个例对本申请的原理及实施方式进行了阐述，以上实施例的说明只是用于帮助理解本申请的方法及其核心思想；同时，对于本领域的一般技术人员，依据本申请的思想，在具体实施方式及应用范围上均会有改变之处，综上所述，本说明书内容不应理解为对本申请的限制。

Claims (10)

  1. 一种数据处理方法，其特征在于，应用于固态硬盘，包括：
    获取历史I/O数据；所述历史I/O数据为所述固态硬盘在预设时间段内被访问的数据；
    利用预测模型对所述历史I/O数据进行学习，获得预测结果；所述预测结果包括：所述固态硬盘在未来窗口期内被访问的数据强度，所述未来窗口期根据所述固态硬盘被访问的周期确定；
    根据所述预测结果管理所述固态硬盘。
  2. 根据权利要求1所述的数据处理方法，其特征在于，所述未来窗口期的确定过程包括：
    若所述固态硬盘仅被一个业务访问，则将所述固态硬盘被当前业务访问的周期确定为所述未来窗口期。
  3. 根据权利要求1所述的数据处理方法，其特征在于，所述未来窗口期的确定过程包括：
    若所述固态硬盘被多个业务访问，则确定所述固态硬盘被每个业务访问的周期；
    计算所有周期的最小公倍数，并通过所述最小公倍数确定所述未来窗口期。
  4. 根据权利要求3所述的数据处理方法，其特征在于，所述通过所述最小公倍数确定所述未来窗口期，包括：
    将所有周期中的最大周期确定为所述未来窗口期的取值下限；
    将所述最小公倍数确定为所述未来窗口期的取值上限；
    将任一个不小于所述取值下限且不大于所述取值上限的目标值确定为所述未来窗口期。
  5. 根据权利要求3或4所述的数据处理方法，其特征在于，还包括：
    计算所有周期的最大公约数，并将所述最大公约数确定为预测间隔；
    按照所述预测间隔在所述未来窗口期中确定多个预测点。
  6. 根据权利要求1所述的数据处理方法，其特征在于，所述利用预测模型对所述历史I/O数据进行学习，获得预测结果，包括：
    利用LSTM模型对所述历史I/O数据进行学习，获得所述预测结果。
  7. 根据权利要求1所述的数据处理方法，其特征在于，所述根据所述预测结果管理所述固态硬盘，包括：
    根据所述预测结果对所述固态硬盘进行缓存管理和/或垃圾回收。
  8. 一种数据处理装置，其特征在于，应用于固态硬盘，包括：
    获取模块，用于获取历史I/O数据；所述历史I/O数据为所述固态硬盘在预设时间段内被访问的数据；
    预测模块，用于利用预测模型对所述历史I/O数据进行学习，获得预测结果；所述预测结果包括：所述固态硬盘在未来窗口期内被访问的数据强度，所述未来窗口期根据所述固态硬盘被访问的周期确定；
    管理模块，用于根据所述预测结果管理所述固态硬盘。
  9. 一种数据处理设备，其特征在于，包括：
    存储器，用于存储计算机程序；
    处理器，用于执行所述计算机程序，以实现如权利要求1至7任一项所述的数据处理方法。
  10. 一种可读存储介质，其特征在于，用于保存计算机程序，其中，所述计算机程序被处理器执行时实现如权利要求1至7任一项所述的数据处理方法。
PCT/CN2020/102020 2019-11-06 2020-07-15 一种数据处理方法、装置、设备及可读存储介质 WO2021088404A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US17/733,225 US20220253214A1 (en) 2019-11-06 2022-04-29 Data processing method, apparatus, device, and readable storage medium

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201911077491.9A CN110764714B (zh) 2019-11-06 2019-11-06 一种数据处理方法、装置、设备及可读存储介质
CN201911077491.9 2019-11-06

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US17/733,225 Continuation US20220253214A1 (en) 2019-11-06 2022-04-29 Data processing method, apparatus, device, and readable storage medium

Publications (1)

Publication Number Publication Date
WO2021088404A1 true WO2021088404A1 (zh) 2021-05-14

Family

ID=69336574

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/102020 WO2021088404A1 (zh) 2019-11-06 2020-07-15 一种数据处理方法、装置、设备及可读存储介质

Country Status (3)

Country Link
US (1) US20220253214A1 (zh)
CN (1) CN110764714B (zh)
WO (1) WO2021088404A1 (zh)

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110764714B (zh) * 2019-11-06 2021-07-27 深圳大普微电子科技有限公司 一种数据处理方法、装置、设备及可读存储介质
CN113971137A (zh) * 2020-07-22 2022-01-25 华为技术有限公司 一种垃圾回收方法及装置
CN112860593A (zh) * 2021-02-09 2021-05-28 山东英信计算机技术有限公司 一种存储系统的gc性能预测方法、系统、介质及设备
CN116700634B (zh) * 2023-08-08 2023-11-03 苏州浪潮智能科技有限公司 分布式存储系统垃圾回收方法、装置及分布式存储系统
CN117806837A (zh) * 2024-02-29 2024-04-02 山东云海国创云计算装备产业创新中心有限公司 一种硬盘任务管理方法、装置、存储介质及系统

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2018184204A1 (en) * 2017-04-07 2018-10-11 Intel Corporation Methods and systems for budgeted and simplified training of deep neural networks
CN109543832A (zh) * 2018-11-27 2019-03-29 北京中科寒武纪科技有限公司 一种计算装置及板卡
CN109670581A (zh) * 2018-12-21 2019-04-23 北京中科寒武纪科技有限公司 一种计算装置及板卡
CN110389909A (zh) * 2018-04-16 2019-10-29 三星电子株式会社 使用深度神经网络优化固态驱动器的性能的系统和方法
CN110764714A (zh) * 2019-11-06 2020-02-07 深圳大普微电子科技有限公司 一种数据处理方法、装置、设备及可读存储介质

Family Cites Families (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7502358B1 (en) * 2004-10-07 2009-03-10 Marvell International Ltd. Diffusion bucket scheduler for wireless network devices
US7996642B1 (en) * 2007-04-25 2011-08-09 Marvell International Ltd. Digital locked loop on channel tagged memory requests for memory optimization
JP2012231445A (ja) * 2011-04-11 2012-11-22 Toshiba Corp パケット配信装置およびパケット配信方法
US9158468B2 (en) * 2013-01-02 2015-10-13 International Business Machines Corporation High read block clustering at deduplication layer
US9940337B2 (en) * 2015-05-31 2018-04-10 Vmware, Inc. Predictive probabilistic deduplication of storage
WO2017058045A1 (en) * 2015-09-29 2017-04-06 Emc Corporation Dynamic storage tiering based on predicted workloads
US10353628B2 (en) * 2017-04-13 2019-07-16 Samsung Electronics Co., Ltd. Opportunity window hints for background operations in SSD
CN109976905B (zh) * 2019-03-01 2021-10-22 联想(北京)有限公司 内存管理方法、装置和电子设备
CN109992210B (zh) * 2019-03-29 2020-10-23 重庆紫光华山智安科技有限公司 数据存储方法、装置及电子设备
CN111913649B (zh) * 2019-05-09 2022-05-06 深圳大普微电子科技有限公司 一种固态硬盘的数据处理方法及装置

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2018184204A1 (en) * 2017-04-07 2018-10-11 Intel Corporation Methods and systems for budgeted and simplified training of deep neural networks
CN110389909A (zh) * 2018-04-16 2019-10-29 三星电子株式会社 使用深度神经网络优化固态驱动器的性能的系统和方法
CN109543832A (zh) * 2018-11-27 2019-03-29 北京中科寒武纪科技有限公司 一种计算装置及板卡
CN109670581A (zh) * 2018-12-21 2019-04-23 北京中科寒武纪科技有限公司 一种计算装置及板卡
CN110764714A (zh) * 2019-11-06 2020-02-07 深圳大普微电子科技有限公司 一种数据处理方法、装置、设备及可读存储介质

Also Published As

Publication number Publication date
CN110764714B (zh) 2021-07-27
CN110764714A (zh) 2020-02-07
US20220253214A1 (en) 2022-08-11

Similar Documents

Publication Publication Date Title
WO2021088404A1 (zh) 一种数据处理方法、装置、设备及可读存储介质
US9652374B2 (en) Sparsity-driven matrix representation to optimize operational and storage efficiency
WO2021120789A1 (zh) 数据写入方法、装置及存储服务器和计算机可读存储介质
WO2021197364A1 (zh) 一种用于服务的扩缩容的方法及相关设备
WO2020119051A1 (zh) 云平台资源使用预测方法及终端设备
EP2494436A1 (en) Allocating storage memory based on future use estimates
TW200939052A (en) Dynamic formulas for spreadsheet cells
TW200807266A (en) System and method of dynamically changing file representations
US10560537B2 (en) Function based dynamic traffic management for network services
CN108418858B (zh) 一种面向Geo-distributed云存储的数据副本放置方法
CN103593452A (zh) 一种基于MapReduce机制的数据密集型成本优化方法
US20120221373A1 (en) Estimating Business Service Responsiveness
WO2023245965A1 (zh) 一种脉冲神经网络加速计算系统、方法、设备及非易失性可读存储介质
CN110471944A (zh) 指标统计方法、系统、设备及存储介质
US6427152B1 (en) System and method for providing property histories of objects and collections for determining device capacity based thereon
TW202138999A (zh) 用於卷積運算的資料劃分方法及處理器
CN109471971B (zh) 一种面向教育领域资源云存储的语义预取方法及系统
Trivedi et al. A decision model for closed queuing networks
WO2022110861A1 (zh) 一种网络训练的数据集缓存方法、装置、设备及存储介质
Löpker et al. The idle period of the finite G/M/1 queue with an interpretation in risk theory
CN114816750A (zh) 大数据数据治理任务运行方法
JPWO2013114911A1 (ja) リスク評価システム、リスク評価方法、及びプログラム
Han et al. Virtual Machine Allocation Strategy Based on Statistical Machine Learning
CN106909522A (zh) Gpu写请求数据的延迟控制方法、装置以及云计算系统
Yao et al. Uniform scheduling of interruptible garbage collection and request IO to improve performance and wear-leveling of SSDs

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20886047

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20886047

Country of ref document: EP

Kind code of ref document: A1