Time sequence database acceleration system and method for heterogeneous time sequence data preprocessing

By using heterogeneous computing systems in the timing database, the timing data processing is offloaded to parallel logic devices (such as GPGPU or FPGA) for acceleration, the problem of large computing power and low efficiency during timing database processing is solved, and more efficient Data processing and performance improvements.

CN119088543APending Publication Date: 2024-12-06山东浪潮数据库技术有限公司
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411068008.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-06
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

The timing database occupies a large computing power and is less efficient when processing timing data, which affects the performance of the database.

Method used

Using heterogeneous computing system, the timing data processing process is offloaded to parallel logic devices (such as GPGPU or FPGA) through the CPU+X architecture for acceleration, and the parallel logic devices are used as coprocessors to complete the preprocessing of timing data, including computing acceleration cards, Storage devices, peripheral electronic devices and input and output interfaces realize data format conversion, cleaning, integration, transformation, feature extraction, normalization and discretization.

Benefits of technology

The data processing computing power of the timing database is liberated, the efficiency and speed of timing data processing is improved, the real-time requirements of the timing database are met, and the overall performance is significantly improved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119088543A_ABST
    Figure CN119088543A_ABST
Patent Text Reader

Abstract

The invention discloses a time sequence database acceleration system and method for heterogeneous time sequence data preprocessing, belongs to the technical field of heterogeneous calculation and time sequence databases, and aims at solving the technical problems that when a time sequence database processes time sequence data, the occupied calculation power is large, and the efficiency is low. According to the technical scheme, the system preprocesses time sequence data collected by a time sequence database on the basis of a heterogeneous computing system, the time sequence data processing process of the time sequence database is unloaded into X for acceleration through a CPU + X framework, and the X serves as a coprocessor to complete preprocessing of the time sequence data collected by all nodes of the time sequence database; wherein X refers to a parallel logic device, and the parallel logic device adopts a GPGPU or an FPGA.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of heterogeneous computing and time series database technology, and specifically to a time series database acceleration system and method for preprocessing heterogeneous time series data. Background Art

[0002] The full name of the time series database is time series database, which is mainly used to process data with time tags (data that changes in the order of time, that is, time serialization). Data with time tags is also called time series data or time series data. Time series data is mainly collected and generated by various types of real-time monitoring, inspection and analysis equipment such as finance, electricity, chemical industry, meteorology, and geographic information. The typical characteristics of these data are: fast generation frequency, heavy dependence on collection time, and large amount of information with many measurement points.

[0003] Compared with relational databases, time series databases focus more on inserting and analyzing data. Because time series data is generated frequently and in large amounts, and the data formats collected by different sensors (such as temperature, humidity, and pressure sensors) are all different, time series databases need to process the collected data before storing it for subsequent queries. The processing of time series data occupies most of the computing power of the time series database, which seriously affects the database performance. Summary of the invention

[0004] The technical task of the present invention is to provide a time series database acceleration system and method for preprocessing heterogeneous time series data to solve the problem that the time series database occupies a large amount of computing power and has low efficiency when processing time series data.

[0005] The technical task of the present invention is achieved in the following way: a time series database acceleration system for preprocessing heterogeneous time series data, which preprocesses the time series data collected by the time series database based on a heterogeneous computing system, unloads the time series data processing process of the time series database to X for acceleration through the CPU+X architecture, and X is used as a coprocessor to complete the preprocessing of the time series data collected by each node of the time series database;

[0006] Here, X refers to a parallel logic device, and the parallel logic device uses a GPGPU or FPGA.

[0007] Preferably, the time series database is a distributed database, consisting of several database nodes, each of which is a heterogeneous computing system composed of a CPU and a computing acceleration card. Each database node is connected to several sensors, and the sensors are used to collect time series data. The computing acceleration card is used to offload the time series data processing process of the time series database to parallel logic devices, thereby releasing the database computing power.

[0008] Preferably, the computing acceleration card includes a parallel logic device, a storage device, a peripheral electronic device and an input and output interface;

[0009] Among them, the parallel logic device is the core device of the computing acceleration card, which is used to receive the time series data transmitted by the database CPU, complete the format conversion according to the time series data type, and then transmit the conversion result back to the database CPU to complete the hardware offloading and acceleration of time series data processing;

[0010] The storage device includes Flash and DDR4 devices. Flash stores firmware logic programs, and DDR4 stores data to be processed and processing result data.

[0011] The input and output interfaces include network port, PCIE interface and JTAG interface; PCIE interface is used for high-speed communication between the computing accelerator card and the database CPU; JTAG interface is used for burning and debugging FPGA logic programs;

[0012] Peripheral electronic components include power chips, clock chips and necessary resistors and capacitors.

[0013] More preferably, the parallel logic device includes a PCIE module, a logic control module, a data loading module, a data processing pipeline module and a result return module;

[0014] Among them, the PCIE module instantiates the PCIE interface to achieve high-speed communication with the database CPU and realizes data transmission between the CPU memory and DDR4 through DMA;

[0015] The logic control module is used to receive the control signal received by the PCIE module, and implement logic control and parameter input for the data loading module, the data processing pipeline module and the result return module;

[0016] The data loading module is used to read data from the corresponding address of DDR4 according to the control parameters (start signal, DDR4 read address, data volume, etc.) sent by the logic control module, and send it to the data processing pipeline module.

[0017] The data processing pipeline module is the core module of the internal logic, which is used to convert and process the input time series data through efficient pipelines. Based on the high parallelism of parallel logic devices, multiple pipelines can be instantiated inside the data processing pipeline module to process the input data in parallel at the same time, thereby improving the data processing speed and efficiency.

[0018] The result return module is used to receive the processing results from the data processing pipeline module and write the results into DDR4 according to the input parameters (DDR4 write address) of the logic control module.

[0019] Preferably, the data processing pipeline module realizes the processing of time series data through the processes of data cleaning, data integration, data transformation, data feature extraction, data normalization, data discretization and data standardization. The specific implementation is determined according to the storage format of the time series database and the collection process of the time series data.

[0020] Preferably, there may be multiple pipelines in the data processing pipeline module working simultaneously, and the results sent to the result return module are sent in multiple parallels. The result return module needs to maintain an internal FIFO to cache the sent results; at the same time, when the result data in the FIFO reaches the maximum bandwidth transmission width, the data is taken out from the FIFO and written to DDR4 to obtain higher bandwidth utilization.

[0021] A time series database acceleration method for heterogeneous time series data preprocessing, which realizes hardware offloading and heterogeneous acceleration of the time series data processing process of the time series database by using the high-speed communication between the computing acceleration card and the database node and the high parallelism and pipeline characteristics of the parallel logic device; the details are as follows:

[0022] ①The database node stores the time series data collected from the sensor or the Internet into the server host memory;

[0023] ②The database node CPU establishes an acceleration task, calls the acceleration API, and sends the acceleration task to the computing acceleration card;

[0024] ③The database node transfers the collected timing data to the DDR4 of the computing acceleration card through DMA, and then starts the acceleration task;

[0025] ④After receiving the acceleration task and start signal, the computing acceleration card starts processing the timing data;

[0026] ⑤ After the computing acceleration card completes data processing acceleration, it sends a completion signal to the database node CPU;

[0027] ⑥ After the database node CPU detects the completion signal, it migrates the processing results from the DDR4 of the computing acceleration card back to the host memory, completing the hardware offloading and heterogeneous acceleration of time series data processing;

[0028] ⑦ All database nodes perform the heterogeneous acceleration process according to steps ① to ⑥.

[0029] As a preferred embodiment, the processing process of time series data is as follows:

[0030] First, read the data to be processed from the corresponding address of DDR4;

[0031] Then the data processing is completed through the data processing pipeline;

[0032] Finally, the processing results are written into DDR4.

[0033] Preferably, the computing acceleration card includes a parallel logic device, a storage device, a peripheral electronic device and an input and output interface;

[0034] Among them, the parallel logic device is the core device of the computing acceleration card, which is used to receive the time series data transmitted by the database CPU, complete the format conversion according to the time series data type, and then transmit the conversion result back to the database CPU to complete the hardware offloading and acceleration of time series data processing;

[0035] The storage device includes Flash and DDR4 devices. Flash stores firmware logic programs, and DDR4 stores data to be processed and processing result data.

[0036] The input and output interfaces include network port, PCIE interface and JTAG interface; PCIE interface is used for high-speed communication between the computing accelerator card and the database CPU; JTAG interface is used for burning and debugging FPGA logic programs;

[0037] Peripheral electronic components include power chips, clock chips and necessary resistors and capacitors.

[0038] More preferably, the parallel logic device includes a PCIE module, a logic control module, a data loading module, a data processing pipeline module and a result return module;

[0039] Among them, the PCIE module instantiates the PCIE interface to achieve high-speed communication with the database CPU and realizes data transmission between the CPU memory and DDR4 through DMA;

[0040] The logic control module is used to receive the control signal received by the PCIE module, and implement logic control and parameter input for the data loading module, the data processing pipeline module and the result return module;

[0041] The data loading module is used to read data from the corresponding address of DDR4 according to the control parameters (start signal, DDR4 read address, data volume, etc.) sent by the logic control module, and send it to the data processing pipeline module.

[0042] The data processing pipeline module is the core module of the internal logic, which is used to convert and process the format of the input time series data through an efficient pipeline. Based on the high parallelism of parallel logic devices, multiple pipelines can be instantiated inside the data processing pipeline module to process the input data in parallel at the same time, thereby improving the data processing speed and efficiency. The data processing pipeline module realizes the processing of time series data through the process of data cleaning, data integration, data transformation, data feature extraction, data normalization, data discretization and data standardization. The specific implementation is determined according to the storage format of the time series database and the collection process of time series data.

[0043] The result return module is used to receive processing results from the data processing pipeline module, and write the results into DDR4 according to the input parameters (DDR4 write address) of the logic control module; there may be multiple pipelines working simultaneously in the data processing pipeline module, and the results sent to the result return module are multi-parallel input. The result return module needs to maintain an internal FIFO to cache the input results; at the same time, when the result data in the FIFO reaches the maximum bandwidth transmission width, the data is taken out from the FIFO and written into DDR4 to obtain higher bandwidth utilization.

[0044] The time series database acceleration system and method for preprocessing heterogeneous time series data of the present invention have the following advantages:

[0045] (1) The present invention uses a CPU+X (GPGPU, FPGA, etc.) heterogeneous architecture to offload the time series data processing process of the time series database to X for acceleration. X is used as a coprocessor to complete the preprocessing of the time series data collected by the sensors or networks of each node of the time series database. This not only frees up the computing power of the time series database for time series data processing, but also X, as a high-parallelism device, can complete data processing more efficiently and faster than the database CPU, better meet the real-time requirements of the time series database, and improve the overall performance of the time series database;

[0046] (ii) The present invention uses parallel logic devices. Parallel logic devices can realize almost real-time data processing due to their high parallelism and pipeline advantages. By unloading the processing of time series data to parallel logic devices, it not only liberates the data processing computing power of the time series database, but also greatly improves the processing efficiency and speed of time series data, thereby improving the overall performance of the database. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The present invention is further described below in conjunction with the accompanying drawings.

[0048] Attached Figure 1 A schematic diagram of the structure of a time series database acceleration system for preprocessing heterogeneous time series data;

[0049] Attached Figure 2 It is a schematic diagram of the internal logic structure of a parallel logic device;

[0050] Attached Figure 3 A flowchart of a time series database acceleration method for preprocessing heterogeneous time series data. DETAILED DESCRIPTION

[0051] The time series database acceleration system and method for preprocessing heterogeneous time series data of the present invention are described in detail below with reference to the accompanying drawings and specific embodiments of the specification.

[0052] Embodiment 1:

[0053] As attached Figure 1 As shown, this embodiment provides a time series database acceleration system for preprocessing heterogeneous time series data. The system preprocesses the time series data collected by the time series database based on the heterogeneous computing system, and unloads the time series data processing process of the time series database to X for acceleration through the CPU+X architecture. X is used as a coprocessor to complete the preprocessing of the time series data collected by each node of the time series database;

[0054] Here, X refers to a parallel logic device, and the parallel logic device uses a GPGPU or FPGA.

[0055] The time series database in this embodiment is a distributed database, which is composed of several database nodes. Each database node is composed of a CPU and a computing acceleration card to form a heterogeneous computing system. Each database node is connected to several sensors (temperature sensors, humidity sensors, pressure sensors, etc.), and the sensors are used to collect time series data (temperature, humidity, pressure, etc.); in addition, time series data can also be obtained from the Internet, such as commodity sales data, stock data, etc.; the computing acceleration card is used to offload the time series data processing process of the time series database to parallel logic devices, thereby releasing the database computing power.

[0056] The computing acceleration card in this embodiment includes a parallel logic device (FPGA, GPGPU, etc.), a storage device, peripheral electronic devices, and an input and output interface;

[0057] Among them, the parallel logic device is the core device of the computing acceleration card, which is used to receive the time series data transmitted by the database CPU, complete the format conversion according to the time series data type, and then transmit the conversion result back to the database CPU to complete the hardware offloading and acceleration of time series data processing;

[0058] The storage device includes Flash and DDR4 devices. Flash stores firmware logic programs, and DDR4 stores data to be processed and processing result data.

[0059] The input and output interfaces include network port, PCIE interface and JTAG interface; PCIE interface is used for high-speed communication between the computing accelerator card and the database CPU; JTAG interface is used for burning and debugging FPGA logic programs;

[0060] Peripheral electronic components include power chips, clock chips and necessary resistors and capacitors.

[0061] As attached Figure 2 As shown, the internal logic of the parallel logic device in this embodiment includes the engineering module structure and specific data processing methods required for data processing, including a PCIE module, a logic control module, a data loading module, a data processing pipeline module and a result return module;

[0062] Among them, the PCIE module instantiates the PCIE interface to achieve high-speed communication with the database CPU and realizes data transmission between the CPU memory and DDR4 through DMA;

[0063] The logic control module is used to receive the control signal received by the PCIE module, and implement logic control and parameter input for the data loading module, the data processing pipeline module and the result return module;

[0064] The data loading module is used to read data from the corresponding address of DDR4 according to the control parameters (start signal, DDR4 read address, data volume, etc.) sent by the logic control module, and send it to the data processing pipeline module.

[0065] The data processing pipeline module is the core module of the internal logic, which is used to convert and process the input time series data through efficient pipelines. Based on the high parallelism of parallel logic devices, multiple pipelines can be instantiated inside the data processing pipeline module to process the input data in parallel at the same time, thereby improving the data processing speed and efficiency.

[0066] The result return module is used to receive the processing results from the data processing pipeline module and write the results into DDR4 according to the input parameters (DDR4 write address) of the logic control module.

[0067] The data processing pipeline module in this embodiment realizes the processing of time series data through the processes of data cleaning, data integration, data transformation, data feature extraction, data normalization, data discretization and data standardization. The specific implementation is determined according to the storage format of the time series database and the collection process of the time series data.

[0068] The data processing pipeline module in this embodiment may have multiple pipelines working simultaneously, and the results sent to the result return module are sent in multiple parallels. The result return module needs to maintain an internal FIFO to cache the sent results; at the same time, when the result data in the FIFO reaches the maximum bandwidth transmission width, the data is taken out from the FIFO and written to DDR4 to obtain higher bandwidth utilization.

[0069] Embodiment 2:

[0070] As attached Figure 3 In the embodiment, a method for accelerating heterogeneous time series data preprocessing in a time series database is described. The method realizes hardware unloading and heterogeneous acceleration of the time series data processing process of the time series database by computing high-speed communication between the acceleration card and the database node and the high parallelism and pipeline characteristics of the parallel logic device. The details are as follows:

[0071] ①The database node stores the time series data collected from the sensor or the Internet into the server host memory;

[0072] ②The database node CPU establishes an acceleration task, calls the acceleration API, and sends the acceleration task to the computing acceleration card;

[0073] ③The database node transfers the collected timing data to the DDR4 of the computing acceleration card through DMA, and then starts the acceleration task;

[0074] ④After receiving the acceleration task and start signal, the computing acceleration card starts processing the timing data;

[0075] ⑤ After the computing acceleration card completes data processing acceleration, it sends a completion signal to the database node CPU;

[0076] ⑥ After the database node CPU detects the completion signal, it migrates the processing results from the DDR4 of the computing acceleration card back to the host memory, completing the hardware offloading and heterogeneous acceleration of time series data processing;

[0077] ⑦ All database nodes perform the heterogeneous acceleration process according to steps ① to ⑥.

[0078] The processing process of the time series data in this embodiment is as follows:

[0079] First, read the data to be processed from the corresponding address of DDR4;

[0080] Then the data processing is completed through the data processing pipeline;

[0081] Finally, the processing results are written into DDR4.

[0082] The computing acceleration card in this embodiment includes a parallel logic device, a storage device, peripheral electronic devices and an input and output interface;

[0083] Among them, the parallel logic device is the core device of the computing acceleration card, which is used to receive the time series data transmitted by the database CPU, complete the format conversion according to the time series data type, and then transmit the conversion result back to the database CPU to complete the hardware offloading and acceleration of time series data processing;

[0084] The storage device includes Flash and DDR4 devices. Flash stores firmware logic programs, and DDR4 stores data to be processed and processing result data.

[0085] The input and output interfaces include network port, PCIE interface and JTAG interface; PCIE interface is used for high-speed communication between the computing accelerator card and the database CPU; JTAG interface is used for burning and debugging FPGA logic programs;

[0086] Peripheral electronic components include power chips, clock chips and necessary resistors and capacitors.

[0087] The parallel logic device in this embodiment includes a PCIE module, a logic control module, a data loading module, a data processing pipeline module and a result return module;

[0088] Among them, the PCIE module instantiates the PCIE interface to achieve high-speed communication with the database CPU and realizes data transmission between the CPU memory and DDR4 through DMA;

[0089] The logic control module is used to receive the control signal received by the PCIE module, and implement logic control and parameter input for the data loading module, the data processing pipeline module and the result return module;

[0090] The data loading module is used to read data from the corresponding address of DDR4 according to the control parameters (start signal, DDR4 read address, data volume, etc.) sent by the logic control module, and send it to the data processing pipeline module.

[0091] The data processing pipeline module is the core module of the internal logic, which is used to convert and process the format of the input time series data through an efficient pipeline. Based on the high parallelism of parallel logic devices, multiple pipelines can be instantiated inside the data processing pipeline module to process the input data in parallel at the same time, thereby improving the data processing speed and efficiency. The data processing pipeline module realizes the processing of time series data through the process of data cleaning, data integration, data transformation, data feature extraction, data normalization, data discretization and data standardization. The specific implementation is determined according to the storage format of the time series database and the collection process of time series data.

[0092] The result return module is used to receive processing results from the data processing pipeline module, and write the results into DDR4 according to the input parameters (DDR4 write address) of the logic control module; there may be multiple pipelines working simultaneously in the data processing pipeline module, and the results sent to the result return module are multi-parallel input. The result return module needs to maintain an internal FIFO to cache the input results; at the same time, when the result data in the FIFO reaches the maximum bandwidth transmission width, the data is taken out from the FIFO and written into DDR4 to obtain higher bandwidth utilization.

[0093] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A time series database acceleration system for preprocessing heterogeneous time series data, characterized in that: The system preprocesses the time series data collected by the time series database based on a heterogeneous computing system. Through the CPU+X architecture, the time series data processing of the time series database is unloaded to X for acceleration. X acts as a coprocessor to complete the preprocessing of the time series data collected by each node of the time series database. Here, X refers to a parallel logic device, and the parallel logic device uses a GPGPU or FPGA.

2. The time series database acceleration system for heterogeneous time series data preprocessing according to claim 1, characterized in that: The time series database is a distributed database consisting of several database nodes. Each database node is a heterogeneous computing system composed of a CPU and a computing acceleration card. Each database node is connected to several sensors, which are used to collect time series data. The computing acceleration card is used to offload the time series data processing process of the time series database to parallel logic devices, thereby releasing the database computing power.

3. The time series database acceleration system for heterogeneous time series data preprocessing according to claim 1, characterized in that: The computing accelerator card includes parallel logic devices, storage devices, peripheral electronic devices and input and output interfaces; The parallel logic device is used to receive the time series data transmitted by the database CPU, complete the format conversion according to the time series data type, and then transmit the conversion result back to the database CPU to complete the hardware offloading and acceleration of the time series data processing; The storage device includes Flash and DDR4 devices. Flash stores firmware logic programs, and DDR4 stores data to be processed and processing result data. The input and output interfaces include network port, PCIE interface and JTAG interface; PCIE interface is used for high-speed communication between the computing accelerator card and the database CPU; JTAG interface is used for burning and debugging FPGA logic programs; Peripheral electronic components include power chips, clock chips and necessary resistors and capacitors.

4. The time series database acceleration system for preprocessing heterogeneous time series data according to any one of claims 1 to 3, characterized in that: The parallel logic device includes a PCIE module, a logic control module, a data loading module, a data processing pipeline module and a result return module; Among them, the PCIE module instantiates the PCIE interface to achieve high-speed communication with the database CPU and realizes data transmission between the CPU memory and DDR4 through DMA; The logic control module is used to receive the control signal received by the PCIE module, and implement logic control and parameter input for the data loading module, the data processing pipeline module and the result return module; The data loading module is used to read data from the corresponding address of DDR4 according to the control parameters sent by the logic control module, and send it to the data processing pipeline module. The data processing pipeline module is used to convert and process the format of input time series data through an efficient pipeline. Based on the high parallelism of parallel logic devices, multiple pipelines can be instantiated inside the data processing pipeline module to process the input data in parallel at the same time, thereby improving the data processing speed and efficiency. The result return module is used to receive the processing results from the data processing pipeline module and write the results into DDR4 according to the input parameters of the logic control module.

5. The time series database acceleration system for heterogeneous time series data preprocessing according to claim 4, characterized in that: The data processing pipeline module realizes the processing of time series data through the processes of data cleaning, data integration, data transformation, data feature extraction, data normalization, data discretization and data standardization.

6. The time series database acceleration system for preprocessing heterogeneous time series data according to claim 4, characterized in that: The data processing pipeline module may have multiple pipelines working simultaneously, and the results sent to the result return module are sent in parallel. The result return module needs to maintain an internal FIFO to cache the results sent in. At the same time, when the result data in the FIFO reaches the maximum bandwidth transmission width, the data is taken out of the FIFO and written to DDR4.

7. A time series database acceleration method for preprocessing heterogeneous time series data, characterized in that: This method realizes hardware offloading and heterogeneous acceleration of the time series data processing process of the time series database by using the high-speed communication between the computing accelerator card and the database node and the high parallelism and pipeline characteristics of the parallel logic device; The details are as follows: ①The database node stores the time series data collected from the sensor or the Internet into the server host memory; ②The database node CPU establishes an acceleration task, calls the acceleration API, and sends the acceleration task to the computing acceleration card; ③The database node transfers the collected timing data to the DDR4 of the computing acceleration card through DMA, and then starts the acceleration task; ④After receiving the acceleration task and start signal, the computing acceleration card starts processing the timing data; ⑤ After the computing acceleration card completes data processing acceleration, it sends a completion signal to the database node CPU; ⑥ After the database node CPU detects the completion signal, it migrates the processing results from the DDR4 of the computing acceleration card back to the host memory, completing the hardware offloading and heterogeneous acceleration of time series data processing; ⑦ All database nodes perform the heterogeneous acceleration process according to steps ① to ⑥.

8. The time series database acceleration method for preprocessing heterogeneous time series data according to claim 7, characterized in that: The specific process of processing time series data is as follows: First, read the data to be processed from the corresponding address of DDR4; Then the data processing is completed through the data processing pipeline; Finally, the processing results are written into DDR4.

9. The method for accelerating time series database by preprocessing heterogeneous time series data according to claim 7, characterized in that: The computing accelerator card includes parallel logic devices, storage devices, peripheral electronic devices and input and output interfaces; The parallel logic device is used to receive the time series data transmitted by the database CPU, complete the format conversion according to the time series data type, and then transmit the conversion result back to the database CPU to complete the hardware offloading and acceleration of the time series data processing; The storage device includes Flash and DDR4 devices. Flash stores firmware logic programs, and DDR4 stores data to be processed and processing result data. The input and output interfaces include network port, PCIE interface and JTAG interface; PCIE interface is used for high-speed communication between the computing accelerator card and the database CPU; JTAG interface is used for burning and debugging FPGA logic programs; Peripheral electronic components include power chips, clock chips and necessary resistors and capacitors.

10. The time series database acceleration method for preprocessing heterogeneous time series data according to any one of claims 7 to 9, characterized in that: The parallel logic device includes a PCIE module, a logic control module, a data loading module, a data processing pipeline module and a result return module; Among them, the PCIE module instantiates the PCIE interface to achieve high-speed communication with the database CPU and realizes data transmission between the CPU memory and DDR4 through DMA; The logic control module is used to receive the control signal received by the PCIE module, and implement logic control and parameter input for the data loading module, the data processing pipeline module and the result return module; The data loading module is used to read data from the corresponding address of DDR4 according to the control parameters sent by the logic control module, and send it to the data processing pipeline module. The data processing pipeline module is used to convert and process the format of input time series data through an efficient pipeline. Based on the high parallelism of parallel logic devices, multiple pipelines can be instantiated inside the data processing pipeline module to process the input data in parallel at the same time, thereby improving the data processing speed and efficiency. The data processing pipeline module realizes the processing of time series data through the process of data cleaning, data integration, data transformation, data feature extraction, data normalization, data discretization and data standardization. The result return module is used to receive processing results from the data processing pipeline module and write the results into DDR4 according to the input parameters of the logic control module; there may be multiple pipelines working simultaneously in the data processing pipeline module, and the results sent to the result return module are multi-parallel input. The result return module needs to maintain an internal FIFO to cache the input results; at the same time, when the result data in the FIFO reaches the maximum bandwidth transmission width, the data is taken out from the FIFO and written into DDR4.