WO2021089036A1

WO2021089036A1 - Data transmission method, network device, network system and chip

Info

Publication number: WO2021089036A1
Application number: PCT/CN2020/127446
Authority: WO
Inventors: 游俊
Original assignee: 华为技术有限公司
Priority date: 2019-11-07
Filing date: 2020-11-09
Publication date: 2021-05-14
Also published as: CN112788079A

Abstract

The present application provides a data transmission method, network device, network system and chip. When the network device is configured as a destination network device, it will receive a write request from a source device, and execute the write request to write the data to be written in the write request to the destination device. After the data to be written is written into the destination device, the written data is read from the destination device. Comparing the data to be written in the write request with the read written data; when the data to be written is the same as the read written data, a write success response is returned to the source device; when the data to be written is different from the read written data, a write failure response is returned to the source device. Using the method of the present application, the written data is verified by the destination network device, and there is no need to transmit the data to the source device for verification, thereby reducing the consumption of network transmission resources, and reducing the time delay of data transmission.

Description

Data transmission method, network equipment, network system and chip

Technical field

This application relates to the storage field, and in particular to a data transmission method, network equipment, network system and chip.

Background technique

In order to reduce the occupation of the computing resources of the device during the data transmission process, the data is generally transferred by means of Remote Direct Memory Access (RDMA), for example, when the data from the source device needs to be transferred to the destination device , The data of the source device is first transmitted to the network card of the destination device, and the network card of the destination device is equipped with an RDMA chip. In this way, the data can be directly written into the memory of the destination device through the RDMA chip in the network card, without the need The participation of the processor of the destination device saves the computing resources of the destination device.

In addition, in order to ensure the reliability of the transmitted data, after the data is written to the destination device, the written data needs to be verified. Because the data is transmitted through RDMA, the processor of the destination device is not required to participate , The destination device cannot verify the written data and needs to be verified by the source device. When the source device is verified, the data written in the memory of the destination device needs to be read to the source. Equipment, this not only consumes network resources, but also increases the delay of data transmission.

Summary of the invention

The present invention provides a data transmission method, network equipment, network system, and chip for performing data verification on the network equipment of the destination equipment, which not only saves network transmission resources, but also reduces the time delay of data transmission.

The first aspect of the present invention provides a network device configured as a destination network device (such as a network card of the destination device). The network device receives a write request from a source device (such as a source computing device), and executes the write request to write the data to be written in the write request to a destination device (such as a destination computing device). ); After the data to be written is written to the destination device, read the written data from the destination device; combine the data to be written in the write request with the read written data Data is compared; when the data to be written is the same as the read written data, a write success response is returned to the source device; when the data to be written is different from the read written data , And return a write failure response to the source device.

Using the network equipment provided by the present invention, the data transmitted from the source equipment (such as the source computing equipment) to the destination equipment (such as the destination computing equipment) can pass through the network equipment of the destination equipment (referred to as the destination network for short) After the device) is written to the destination device, the network device of the destination device (the network card of the destination device) verifies the written data without transmitting the data to the source device (such as the source computing device). ) Performs verification, thereby reducing the consumption of network transmission resources in the prior art. After the network device of the destination device verifies the data that has been written, it directly returns a successful data write response to the source device to avoid This reduces the communication overhead of the prior art, thereby reducing the data transmission delay as a whole.

In a possible implementation of the first aspect of the present invention, the network device is configured as a destination network device and is further configured to, after receiving a write request from the source device, cache the data to be written in the write request . In a specific implementation, an area can be allocated in the cache of the destination network device to cache the data to be written in the write request.

In another possible implementation of the first aspect of the present invention, the network device is configured as a destination network device for specific execution: parsing the write request to obtain the data to be written and the data to be written Destination address; and writing the data to be written into the storage space identified by the destination address in the destination device (such as the destination computing device). In a specific implementation, the aforementioned storage space is the memory of the destination device.

In another possible implementation of the first aspect of the present invention, the network device is configured as a source-end network device (such as a network card of a source-end computing device) for executing: receiving the source-end device (such as: The source computing device) sends striped data containing at least two strips of data, the write address of each striped data contained in the striped data and the write address of at least one check data; according to the at least two strips of data; Generate the at least one check data for each piece of data; for each piece of data, generate a write request according to the piece of data and the write address of the piece of data; for each piece of data, according to the The verification data and the write address of the verification data generate a write request; each write request generated above is sent to a destination device.

In the above possible implementation, the generation of verification data is offloaded from the source device (such as the source computing device) to the source network device (such as the network card of the source computing device). Reduce the load of the source-end device, save the computing resources of the source-end device (such as the source-end computing device), thereby improving the data processing efficiency of the source-end device as a whole.

In another possible implementation of the first aspect of the present invention, the network device is configured as a source-end network device for executing: receiving the data to be written sent by the source device, and storing the data to be written At least one logical address of at least one logical copy of the at least one logical address; lock the at least one logical address; for each logical address, generate a write request based on the data to be written and the logical address; combine the generated at least one write request Each write request in is sent to a target network device.

In the above possible implementation manner, the generation of duplicate data is offloaded from the source device (such as the source computing device) to the source network device (such as the network card of the source computing device), you can Reduce the load of the source-end device, save the computing resources of the source-end device (such as the source-end computing device), thereby improving the data processing efficiency of the source-end device as a whole.

In yet another possible implementation manner of the first aspect of the present invention, the network device is configured as a source-end network device for: parallelly sending each of the at least one generated write request to A target network device.

In the above possible implementation manner, by sending multiple write requests to the target network device in parallel, the concurrency of data transmission of the source network device (such as the network card of the source computing device) is enhanced, thereby improving data efficiency. Transmission efficiency.

The second aspect of the present invention provides a data transmission method, which can be executed by a destination network device. The method includes:

Receiving a write request from the source device, and executing the write request to write the data to be written in the write request to the destination device;

After the data to be written is written into the destination device, read the written data from the destination device;

Comparing the data to be written in the write request with the read data that has been written;

When the data to be written is the same as the read data that has been written, returning a write success response to the source device;

When the to-be-written data is different from the read-written data, a write failure response is returned to the source device.

Using the data transmission method provided by the second aspect of the present invention, the data transmitted from the source device to the destination device can be written into the destination device through the network device of the destination device (referred to as the destination network device), The network device of the destination device verifies the written data without transmitting the data to the source device for verification, thereby reducing the consumption of network resources in the prior art. After the device verifies the written data, it directly returns a data write successful response to the source device, avoiding the communication overhead of the prior art, thereby reducing the data transmission delay as a whole.

In a possible implementation of the second aspect of the present invention, the method is executed by a destination network device, and further includes: after receiving the write request, buffering the data to be written in the write request.

In another possible implementation of the second aspect of the present invention, the write request further includes the destination address of the data to be written, and the execution of the write request is to write the data to be written in the write request to the destination. The end device includes: writing the data to be written in the write request into the storage unit of the destination device according to the destination address in the write request.

In another possible implementation of the second aspect of the present invention, the method is executed by a source network device. The method also includes:

Receiving the striped data sent by the source-end device and the write address of the plurality of striped data constituting the striped data and the write address of at least one check data;

Generating the at least one verification data according to the plurality of striped data;

According to the write address of the plurality of striped data, the write address of the at least one parity data, the plurality of striped data and the at least one parity data generate a plurality of write requests, wherein each partition Each piece of data corresponds to a write request, and each check data corresponds to a write request;

Each of the multiple write requests is sent to a destination device respectively.

In the above possible implementation, the generation of verification data is offloaded from the source device (such as the source computing device) to the source network device (such as the network card of the source computing device). Reduce the load of the source device and save the computing resources of the source device (such as the source computing device), thereby improving the data processing efficiency of the source device.

In another possible implementation of the second aspect of the present invention, the method further includes:

Receiving, by the source network device, the data to be written sent by the source device and the logical address of the data to be written;

Acquiring a copy address corresponding to the logical address storing multiple copies of the data to be written;

Lock the logical address;

Generate multiple write requests according to the data to be written and each copy address;

Each of the multiple write requests is sent to a target network device respectively.

In this possible implementation manner, offloading the generation of duplicate data from the source device (such as the source computing device) to the source network device (such as the network card of the source computing device) can reduce The load of the source-end device saves the computing resources of the source-end device (such as the source-end computing device), thereby improving the data processing efficiency of the source-end device as a whole.

In yet another possible implementation of the second aspect of the present invention, each of the multiple write requests is sent to a target network device in parallel.

The third aspect of the present invention provides a network system. The network system includes a source device (such as a source computing device), a destination device (such as a destination computing device), and any one of the first aspect or the first aspect connected to the destination device. The realization of the provided destination network device (such as the network card of the destination computing device).

The fourth aspect of the present invention provides a chip, including an interface and a processor, the interface is used to provide input and output data for the processor; the processor is used to execute the second aspect or any of the second aspect of the present invention A possible implementation of the provided method. In the specific implementation process, the chip can use CPU (Central processing unit, central processing unit), MCU (Micro controller Unit, microcontroller), MPU (Micro processing unit, microprocessor), DSP (Digital signal processing, digital Signal processor), SoC (System on Chip), ASIC (application-specific integrated circuit, application-specific integrated circuit), FPGA (Field Programmable Gate Array) or PLD (Programmable Logic Device), editable Logic device).

A fifth aspect of the present invention provides a network device, the network device includes a processor and a memory, the memory stores program instructions, and the processor executes the program instructions in the memory to execute the second aspect of the present invention or In the second aspect, any possible implementation of the method provided.

The sixth aspect of the present invention provides a computer-readable storage medium, and the computer-readable storage medium stores instructions, which when run on a computer, cause the computer to execute the second aspect or any possible implementation of the second aspect of the present invention The method provided. These computer-readable storages include but are not limited to one or more of the following: ROM (Read-Only Memory), PROM (Programmable ROM), EPROM (Erasable PROM), Flash memory, EEPROM (Electrically EPROM, electric EPROM) and hard drive (Hard drive).

In a seventh aspect, the present application also provides a computer program product containing instructions, which when run on a computer, causes the computer to execute the method provided by the second aspect or any one of the possible implementations of the second aspect of the present invention.

It can be understood that any one of the possible implementation manners of the third aspect and the third aspect of the present invention corresponds to any one of the first aspect and the first aspect respectively, and the fourth, fifth, sixth, and seventh aspects and Any one of their possible implementation manners corresponds to the second aspect and any one of the possible implementation manners of the second aspect respectively, and the technical effects of the implementation are not repeated here.

Description of the drawings

In order to explain the embodiments of the present invention or the technical solutions in the prior art more clearly, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art.

FIG. 1 is an architecture diagram of a data transmission system provided by an embodiment of the present invention.

Fig. 2 is a flowchart of a data transmission method provided by an embodiment of the present invention.

Fig. 3 is a block diagram of a network device provided by an embodiment of the present invention.

Fig. 4 is a hardware structure diagram of a network device provided by an embodiment of the present invention.

Detailed ways

The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments.

As shown in Figure 1, it is an architecture diagram of a network transmission system. The source device 10 and the destination device 20 are connected through a network 30. An applicable scenario of the embodiment of the present invention is: the source device 10 and the destination device 20 may be storage arrays. After the host (not shown) sends a write request to the source device 10, the source device 10 is based on a preset According to the reliability strategy, the source device 10 writes the data to be written in the write request into multiple destination devices 20. For ease of description, the embodiment of the present invention only takes one destination device 20 as an example for description.

The source device 10 includes a processing unit 101, a storage unit 102, and a network device 103. The network device 103 includes a remote direct memory access (RDMA) chip 1031. The destination device 20 includes a processing unit 201, a storage unit 202, and a network device 203. The network device 203 includes an RDMA chip 2031. The storage unit 102 and the storage unit 202 are the memories of the source device 10 and the destination device 20, respectively, which can be dynamic random access memory (Dynamic Random Access Memory, DRAM) or storage class memory (SCM) .

The processing unit 101 and the processing unit 201 can have a variety of specific implementation forms, such as a central processing unit (CPU) or graphics processing unit (GPU), and can be a single-core processor or a multi-core processor. If it is implemented as a multi-core processor, the number of processors is not limited.

The storage unit 102 and the storage unit 202 are the memories of the source device 10 and the destination device 20, respectively, which can be dynamic random access memory (Dynamic Random Access Memory, DRAM) or storage class memory (SCM) . The storage unit 102 and the storage unit 202 are used to store program codes and data, so that the processing unit can call and execute to implement related functions.

In the embodiment of the present invention, the network device 103 and the network device 203 are network cards belonging to the source device 10 and the destination device 20, respectively, but in other embodiments, the network device 103 and the network device 203 are respectively connected to the source The switch of the end device 10 and the destination device 20.

In the embodiment of the present invention, when data is transmitted from the source device 10 to the destination device 20, the RDMA method is adopted. Specifically, the processing unit 101 of the source device 10 first sends the write request to the network of the source device 10 Device 103, the RDMA chip 1031 of the network device 103 forwards the write request to the network device 203 of the destination device 20, and the RDMA 2031 of the network device 203 writes the data in the write request to the storage unit 202 of the destination device 20. When data is written from the source device 10 to the destination device 20 through RDMA, the processing unit 201 of the destination device 20 does not need to participate, so the computing resources of the destination device 20 are saved.

Generally, in order to prevent errors when data is written into the storage unit, the written data must be verified. However, when RDMA is used for data transmission, because the processing unit 201 of the destination device 20 does not participate in the data transmission, it cannot pass the destination. The end device performs data verification. At present, the commonly used method is to use the processing unit 101 of the source device 10 to verify the data written in the storage unit 202 of the destination device 20. The verification process includes:

(1) After the data is written into the storage unit 202 of the destination device 20, the destination device 20 sends response information to the source device 10.

(2) The processing unit 101 of the source device 10 initiates a read request again to read the data written in the storage unit 202 of the destination device 20 to the source device 10, and the processing unit 101 of the source device 10 compares the write request Whether the data in is consistent with the data read from the storage unit 202 of the destination device 20.

(3) If the comparison result shows that the two are consistent, the data verification is successful, and the write completion response is sent to the application that generated the write request.

In the above-mentioned verification method in the prior art, the written data needs to be read from the destination device 20 to the source device 10, which increases the consumption of network bandwidth. Because the source device is required to send and read the destination device’s data. The instruction to write data increases the communication overhead of the data transmission process, thereby increasing the time delay of the data writing operation as a whole.

In order to solve the above-mentioned problems, the present invention provides a data transmission method. After the data transmitted from the source device 10 to the destination device 20 is written into the storage unit 202 of the destination device through the network device 203 of the destination device 20, The network device 203 of the destination device 20 verifies the written data without transmitting the data to the source device 10 for verification, thereby reducing the consumption of network resources. The network device of the destination device verifies the data that has been written. After the written data is verified, the response of data writing success is directly returned to the source end device, which avoids the communication overhead in the prior art, thereby reducing the data transmission delay as a whole. In the following, the data transmission method provided by the embodiment of the present invention will be described as an example in which the data generated by the source device 10 is written into the destination device 20 with reference to Fig. 2.

As shown in FIG. 2, it is a flowchart of a method for writing data from a source device 10 to a destination device 20 in an embodiment of the present invention.

In step S201, the source device 10 receives a write request sent by the host, and the destination device of the write request carries the data to be written and the logical address of the data to be written.

When performing this step, when the source device 10 is used as a storage device, after receiving a write request from the host, it will store the data to be written in the write request. In specific implementation, the write request of the host may be issued by an application running on the host or an application running in a virtual machine of the host.

In step S202, the processing unit 101 of the source device 10 determines the reliability policy of the source device 10. If the source device 10 supports erasure code (EC) or redundant array of disks (Redundant Arrays of Independent) Drives, RAID) strategy, step S203 is executed, and if the source device 10 supports a multi-copy reliability strategy, step S205 is executed.

Data loss may occur during data transmission. In addition, due to storage device failure, data loss may also occur. Therefore, in order to ensure the reliability of data, it is necessary to set up some reliability strategies and implement these reliability strategies (such as: Redundant data storage), so that when data is lost, these redundant data can be used for data recovery. At present, there are mainly two reliability strategies, one is erasure code (EC) or redundant array of disks (Redundant Arrays of Independent Drives, RAID) strategies, both of which are to write to the write request The data is combined into one strip of data, and then one strip of data is divided into multiple strips of data, the verification data of multiple strips is calculated, and the multiple strips of data and the verification data are stored separately. In this way, in any one After the striped data is lost, the lost striped data can be recovered through other striped data and parity data. For example, for RAID 5, a striped data can be divided into three striped data and three striped data Calculate a parity data. For RAID6, a stripe data can be divided into three stripe data, and two parity data are calculated for the three stripe data; the other is a multiple copy strategy, which is writing When data is being stored, multiple backup data are generated for the data to be written, and different backup data are written to different destination devices, so that when the data is damaged, the data can be restored by reading the backup data. A general application will set a reliability strategy in advance. After the processing unit 101 generates a write request, it will check the reliability strategy preset by the application and perform different processing according to different reliability strategies.

Step S203, if it is determined in step S202 that the reliability strategy supported by the application is an EC or RAID strategy, when the processor 101 of the source device 10 determines that the data to be written constitutes striped data, it will constitute the striped data to be written The data and related address information are sent to the network device (also referred to as the source network device) 103 of the source device.

When the source device 10 adopts the EC or Raid strategy to ensure the reliability of the data, the data to be written needs to be combined into striped data. Each striped data is composed of multiple striped data. The source device 10 will set each data in advance. The size of each striped data and the number of striped data that make up the striped data. When the received data to be written is enough for the size of the set striped data, the made-up striped data will be sent to the destination Equipment network equipment 103. For example, when the source device 10 adopts RAID 5, the data to be written needs to be filled with striped data composed of three striped data and then sent to the network device 130. In addition, according to the RAID algorithm, different striped data will be written to different locations. In addition, check data of the striped data needs to be generated. Therefore, the source device 10 will also write each striped data and check data. The address of the written location and the address of the verification data are sent to the network device 103 of the source device. In the embodiment of the present invention, although the verification data for the strip data is not generated in the source device 10, the address where the verification data is stored is sent to the network device 103.

Step S204: After receiving the striped data, the network device (ie, the source-end network device) 103 calculates check data for the striped data.

In order to ensure the reliability of the data, in the RAID algorithm, check data is generated for the striped data, so that after one of the strips is damaged, the check data can be used to recover. At present, the general practice is that the processing unit of the source device generates the check data after forming the striped data, which will occupy the computing resources of the processing unit of the source device, thereby affecting the IO processing. In the embodiment of the present invention, the check data of the stripe data can be calculated in the network device 103, so that the computing resources of the source device 10 can be saved. The specific implementation can be to add a dedicated chip to the network device 103 to generate verification data, and the other is to install a program instruction in the network device 103, which is called by the microprocessor in the network device 103 To generate the verification data. In addition, after the network device 103 receives the striped data, it will first cache the striped data in the cache of the network device, and then calculate the check data.

Step S205: The network device (ie, the source network device) 103 generates multiple sub-write requests, and sends the sub-write requests to different destination devices in parallel according to the addresses of the sub-write requests.

After the verification data is generated, the processor of the source device 10 further generates a plurality of sub-write requests according to the strip data transmitted by the source device processor 101, that is, the address of each strip and the address of the verification data, one of which is Striping corresponds to one sub-write request, and one check data corresponds to one sub-write request. In the embodiment of the present invention, the network device 103 can send multiple sub-write requests to different destination devices in parallel, which improves the efficiency of data transmission compared with the serial transmission by the source device.

Step S206: If it is determined in step S202 that the reliability policy supported by the application is a multiple copy policy, the processing unit 101 of the source device 10 forwards the write request to the network device 103 of the source device 10, and the write The request carries the data to be written, that is, the logical address of the data to be written.

Step S207: The network device 103 of the source device locks the logical address of the data to be written carried in the write request.

Generally, a persistence log (Plog) is generated for each application in the source device 10, and the log records the logical space allocated by the source device 10 for each application. The number of backup data and the backup address set by the application data set the physical address of the storage unit in the destination device 20 storing the backup data for the logical space of the application. After the source device 10 receives the write request, it locks the logical address carried by the write request recorded in the plog to prevent the logical address from being assigned to other write requests. When the logical address is locked, the logical address corresponding to the The physical address of the storage unit 202 in the other destination device 20 will also be locked, thus ensuring the reliability of the data. However, the Plog space allocation performed by the source device 10 will increase the load of the source device 10, so in the embodiment of the present invention, this function is offloaded to the network device 103. That is, the Plog is transferred to the source device 10 for storage. After the network device 103 receives the write request sent by the source device 10, the logical address of the write request in the Plog is locked to prevent the logical address from being allocated to other write requests. In addition, after the network device 103 receives the write request, it will first cache the data to be written in the write request in the cache of the network device, and then perform subsequent processing.

Step S208: The network device 103 of the source device 10 obtains the write addresses of the multiple copies of the data to be written carried in the write request, and generates multiple sub-files based on the write addresses of the multiple copies and the data to be written. Write request.

Since the mapping relationship between the logical address space and the physical addresses of the storage units 202 of the multiple destination devices 20 is recorded in the plog, the write addresses of multiple copies of data can be determined according to the logical addresses of the data to be written. In this way, multiple sub-write requests can be generated according to the data to be written and the write addresses of the multiple copies. When generating the sub-write request, the data to be written and the write address of each copy are carried in each sub-write request.

Step S209: The network device 103 of the source device sends the multiple sub-write requests to different destination devices 20 according to the address of each sub-write request.

In the embodiment of the present invention, in order to achieve better reliability, generally each stripe or different copy data in the striped data is stored in different destination devices, but the striped striped data is stored in different destination devices. Or different copies of data stored in different addresses of the same destination device are also within the protection scope of the present invention.

In step S210, after receiving the sub-write request, the network device 203 of the destination device 20 writes data into the storage unit 202 of the destination device 20 through RDMA. After the data is successfully written into the storage unit 202, the storage unit 202 Generate a successful write response.

In step S211, after receiving the successful write response, the network device 203 reads data from the address indicated by the sub-write request in the storage unit 202 through RDMA, and compares the read data with the sub-write request. Compare the data to be written in.

When data is written into the storage unit 202, errors such as address write offset may occur. Therefore, in order to detect such errors in time, after the data is written into the storage unit 202, it needs to be verified. The embodiment of the present invention adopts The verification method is that after the network device 203 writes the data to be written into the storage unit 202, the written address is read from the written address, so that if an address offset occurs, the data read from the written address The data will be different from the written data, so you can find the address writing error.

In step S212, if the data is consistent, the network device 203 of the destination device 20 returns a write success response to the source device 10.

In step S213, if the data is inconsistent, the network device 203 of the destination device 20 returns a write failure response to the source device 10.

Compared with the traditional data verification method, in the embodiment of the present invention, the data written into the storage unit 202 of the destination device 20 is verified by the destination device 20 and the network device 203, without consuming network bandwidth, and reducing The latency of data writing.

As shown in FIG. 3, it is an architecture diagram of a network device in an embodiment of the present invention. The network device may be the network device 103 of the source device 10 in FIG. 1 or the network device 203 of the destination device 20.

The network device includes a verification data generation module 301, an address lock module 302, a concurrency module 303, an RDMA module 304, and a post-write verification module 305. Among them, the verification data generation module 301, the address locking module 302, the concurrency module 303, and the RDMA module 304 are used when the network device is used as the network device of the source device 10, and the RDMA module 304 and the post-write verification module 305 are used in The network device is used as a destination device.

If the reliability strategy adopted in the source device 10 is an EC/RAID strategy, the source device 10 will send the striped data to the source network device, and the check data generation module 301 will be the stripe The data generates the verification data. For details, please refer to the description of step S204 in FIG. 2. After the verification data is generated, the concurrency module will generate each stripe data and verification data according to the number of strips and verification data in the striped data, and the address of each stripe data and verification data. Multiple sub-write requests, and the generated multiple sub-write requests are concurrently sent to different destination devices 20 according to addresses. For details, please refer to the description of step S205 in FIG. 2.

If the reliability strategy adopted by the source device 10 is a multiple copy strategy, the source device will send the write request to the network device of the source device and at the same time send the addresses written by multiple copies to the network device of the source device. Device, the address locking module 302 will lock the address in the storage unit in the destination device according to the address of each copy sent by the source device to ensure the consistency of multiple copies. For details, please refer to Figure 2 Description of step S207. After the address is locked, the concurrent module generates a sub-write request for each copy according to the data to be written in the write request, that is, the address of each copy, and sends multiple sub-write requests to different addresses in parallel according to the address in the sub-write request. The destination device 20.

When the network device is the network device of the destination device, when receiving a sub-write request sent by the network device of the source device, the RDMA module first writes the data to be written in the sub-write request In the storage unit of the destination device, please refer to the description of step S210 in FIG. 2 for details. After the data is written, the post-write verification module 305 verifies whether the written data is correct. For a detailed description of verification, please refer to steps S211-S213 in FIG. 2.

Each module in the network device can be implemented by software, hardware or a combination of both.

When any of the above modules or units are implemented by software, the software exists in the form of computer program instructions and is stored in the memory, and the processor can be used to execute the program instructions to implement the above method flow. The processor may include but is not limited to at least one of the following: a central processing unit (CPU), a microprocessor, a digital signal processor (DSP), a microcontroller (microcontroller unit, MCU), or artificial intelligence Various computing devices such as processors that run software. Each computing device may include one or more cores for executing software instructions to perform operations or processing. The processor can be a single semiconductor chip, or it can be integrated with other circuits to form a semiconductor chip. For example, it can form an SoC (on-chip) with other circuits (such as codec circuits, hardware acceleration circuits, or various bus and interface circuits). System), or it can be integrated into the ASIC as a built-in processor of an ASIC, and the ASIC integrated with the processor can be packaged separately or together with other circuits. In addition to the core used to execute software instructions for calculation or processing, the processor may further include necessary hardware accelerators, such as field programmable gate array (FPGA) and PLD (programmable logic device) , Or a logic circuit that implements dedicated logic operations.

When the above modules are implemented by hardware circuits, the hardware circuits may be general-purpose CPU (Central processing unit, central processing unit), MCU (Micro controller Unit, microcontroller), MPU (Micro processing unit, microprocessor), DSP (Digital signal processing, digital signal processor), SoC (System on Chip, system-on-chip) to achieve, of course, can also use application-specific integrated circuit (application-specific integrated circuit, ASIC) to achieve, or programmable logic device (programmable logic) device, PLD), the above-mentioned PLD can be a complex programmable logical device (CPLD), a field-programmable gate array (FPGA), a generic array logic (generic array logic, GAL) or its In any combination, it can run necessary software or does not rely on software to execute the above method flow.

As shown in FIG. 4, it is a hardware architecture diagram of the network device in an embodiment of the present invention. The network device includes a processor 410, a memory 420, an interface 430, and an RDMA chip. Program instructions are stored in the memory. When the verification data generation module 301, address locking module 302, concurrency module 303, and post-write verification module 305 in FIG. 3 are implemented in software, the programs corresponding to the above modules are stored In the memory 420, these programs are called and executed by the processor 410 to implement the functions performed by the verification data generation module 301, the address locking module 302, the concurrency module 303, and the post-write verification module 305. The interface 430 is used to receive data from the source device or the destination device or send data to the source device or the destination device. The RDMA chip is used to directly write the data received from the source device into the storage unit of the destination device. The processor may include but is not limited to at least one of the following: central processing unit (central processing unit, processing unit), microprocessor, digital signal processor (DSP), microcontroller (microcontroller unit, MCU), or manual Various computing devices such as smart processors that run software. Each computing device may include one or more cores for executing software instructions to perform operations or processing.

The verification data generation module 301, address lock module 302, concurrency module 303, and post-write verification module 305 in FIG. 3 can also be implemented by hardware. When implemented by hardware, these modules serve as hardware modules in the chip of the network device. Integrated in the chip of the network device. When implemented by hardware, the processor of the network device is an application specific integrated circuit (ASIC) or an independent semiconductor chip. The processor's internal processing is used to execute software instructions to perform calculations or processing, and may further include necessary hardware accelerators, such as field programmable gate array (FPGA), PLD (programmable logic device) , Or a logic circuit that implements dedicated logic operations.

In another embodiment of the present invention, when data is migrated between the application and the source device and the destination device, only steps S209 to S213 need to be performed. In step S209, the data carried in the sent write request is data that needs to be migrated from the source device 10 to the destination device 20. The other steps are the same as S210 and S213, and will not be repeated here.

The above are only specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention. It should be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

A network device, characterized in that the network device is configured as a destination network device for executing:

Receiving a write request from a source device, and executing the write request to write the data to be written in the write request to the destination device;

After the data to be written is written into the destination device, read the written data from the destination device;

Comparing the data to be written in the write request with the read written data;

When the data to be written is the same as the read data that has been written, returning a write success response to the source device;

When the to-be-written data is different from the read-written data, a write failure response is returned to the source device.
The network device according to claim 1, wherein the network device is configured as a destination network device and is further configured to perform:

After receiving the write request of the source-end device, buffer the data to be written in the write request.
The network device according to claim 1 or 2, wherein the network device is configured as a destination network device for specific execution:

Parsing the write request to obtain the data to be written and the destination address of the data to be written;

The data to be written is written into the storage space identified by the destination address in the destination device.
The network device according to any one of claims 1-3, wherein the network device is configured as a source-end network device for executing:

Receiving striped data containing at least two strips of data sent by the source device, writing addresses of at least two strips of data and writing addresses of at least one verification data included in the striped data;

Generating the at least one verification data according to the at least two pieces of data;

For each striped data, generate a write request according to the striped data and the write address of the striped data;

For each check data, generate a write request according to the check data and the write address of the check data;

Each of the generated write requests is sent to a destination device respectively.
The network device according to any one of claims 1-3, wherein the network device is configured as a source-end network device for executing:

Receiving the data to be written sent by the source device, and at least one logical address storing at least one logical copy of the data to be written;

Lock the at least one logical address;

For each logical address, generate a write request according to the data to be written and the logical address;

Each of the generated at least one write request is sent to a target network device respectively.
The network device according to claim 4 or 5, wherein the network device is configured as a source-end network device for:

Each of the generated at least one write request is sent to a target network device in parallel.
A method of data transmission, characterized in that the method includes:

Receiving a write request from a source device, and executing the write request to write the data to be written in the write request to the destination device;

After the data to be written is written into the destination device, read the written data from the destination device;

Comparing the data to be written in the write request with the read data that has been written;

When the data to be written is the same as the read data that has been written, returning a write success response to the source device;

When the to-be-written data is different from the read-written data, a write failure response is returned to the source device.
8. The data transmission method according to claim 7, wherein the method further comprises:

After the write request is received, the data to be written in the write request is cached.
The data transmission method according to claim 7 or 8, wherein the write request further includes the destination address of the data to be written, and the execution of the write request is to transfer the data to be written in the write request. The writing destination device includes:

The data to be written in the write request is written into the storage unit of the destination device according to the destination address in the write request.
A network system, comprising: a source device, a destination device, and the destination network device according to any one of claims 1-3 connected to the destination device.
A chip includes an interface and a processor, and the interface is used to provide input and output data for the processor;

The processor is configured to execute the data transmission method according to any one of claims 7-9.