CN116893788A - Metadata processing method, hardware acceleration network card, system and readable storage medium - Google Patents

Abstract

The invention discloses a metadata processing method, a hardware acceleration network card, a system and a readable storage medium in the technical field of storage, wherein the hardware acceleration network card receives a message corresponding to a metadata request and judges whether the message comprises metadata or not; if not, sending the metadata request to the metadata service, and processing the metadata request by using the metadata service; if yes, decoding the network data stream of the metadata into a metadata cache format by using an accelerator, and storing the network data stream of the metadata into a shared cache; and sending the metadata request to a metadata service, and carrying out request processing on the metadata in the shared cache by using the metadata service. The invention has the technical effects that: the link which occupies the CPU processing performance most in the small file processing process, namely the reverse serialization of metadata is unloaded to the hardware acceleration network card for processing, so that the CPU processing workload can be directly reduced, and the bottleneck of small file processing efficiency can be broken through by means of the acceleration performance of hardware.

Description

Metadata processing methods, hardware acceleration network cards, systems and readable storage media

Technical field

The present invention relates to the field of storage technology, and in particular to a metadata processing method, a hardware acceleration network card, a system and a readable storage medium.

Background technique

As user needs continue to change, storage systems save an increasing number of small files (files smaller than a block size).

Storage systems need to manage both metadata and data. Reading each small file requires multiple metadata accesses to complete the data access. Since the small files themselves are very small, more than 50% of the processing is performed in metadata processing, which not only takes up a lot of CPU, but also the current metadata processing is also a key bottleneck in the performance of small file processing.

To sum up, how to effectively solve problems such as small file processing is an urgent technical problem that those skilled in the art currently need to solve.

Contents of the invention

The purpose of the present invention is to provide a metadata processing method, a hardware acceleration network card, a system and a readable storage medium, and offload the link that most consumes CPU processing performance in the small file processing process to the hardware acceleration network card for processing, which can directly reduce the CPU To handle workloads, you can also rely on hardware acceleration performance to break through the bottleneck of small file processing.

In order to solve the above technical problems, the present invention provides the following technical solutions:

A metadata processing method applied to hardware accelerated network cards, including:

Receive the message corresponding to the metadata request and determine whether the message includes metadata;

If not, send the metadata request to the metadata service, and use the metadata service to process the metadata request;

If so, use an accelerator to decode the metadata network data stream into a metadata cache format and then store it in the shared cache;

Send the metadata request to the metadata service, and use the metadata service to perform request processing on the metadata in the shared cache.

Preferably, the accelerator is used to decode the network data stream of the metadata into a metadata cache format and then store it in the shared cache, including:

Utilize the accelerator to apply for an empty cache from the shared cache;

After the network data stream of the metadata is decoded into a metadata cache format, it is stored in the empty cache.

Preferably, it also includes:

Create a shared cache pool with the host;

Divide the shared cache pool into multiple units according to a fixed allocation size;

Each of the units serves as a cache of file metadata.

Preferably, the step of determining whether the message includes metadata includes:

Read the header type of the message;

Determine whether the header type is a Lookup response, Getattr response, Readdir response, Setattr send, Setxattr send or Create send;

If so, determine that the metadata is included in the message;

If not, it is determined that the metadata is not included in the message.

Preferably, the message corresponding to the metadata request is a request message or a response message corresponding to the metadata request.

Preferably, the hardware acceleration network card is a data processor, and the accelerator is an FPGA.

Preferably, it also includes:

After the metadata service or the client generates a sending request, determine whether the sending message corresponding to the sending request includes target metadata;

If so, use the accelerator to pull the target metadata cache that needs to be serialized from the shared cache through the metadata reference of the sending request, and encode the target metadata cache into a network data stream to obtain target message;

Send the target message.

Preferably, determining whether the sending message corresponding to the sending request includes target metadata includes:

Read the header type of the sent message;

Determine whether the header type is a Lookup response, Getattr response, Readdir response, Setattr send, Setxattr send or Create send;

If so, determine that the target message includes the target metadata;

If not, it is determined that the target metadata is not included in the target message.

A hardware-accelerated network card, including:

The steps of the above metadata processing method can be implemented in the hardware acceleration network card;

The hardware acceleration network card processes the serial number and deserialization of metadata through an accelerator.

A metadata processing system including:

Client and service clusters;

Wherein, the client computer includes a client, a shared cache and the above-mentioned hardware acceleration network card;

The service cluster includes a storage node, and the storage node includes a metadata service, a shared cache, and the above-mentioned hardware acceleration network card.

A metadata processing method, applied to the above-mentioned metadata processing system, including:

After the client generates a request, the hardware acceleration network card local to the client sends a request message to the storage node;

After receiving the request message, the storage node uses the local metadata service, the shared cache and the hardware acceleration network card to process the request message and sends it to the client. response message;

The client receives the response message and processes the request result.

Preferably, after the client generates a request, the local hardware acceleration network card of the client sends a request message to the storage node, including:

Determine whether the request needs to carry metadata;

If so, the local hardware acceleration network card of the client sends a request message corresponding to the request type of the request to the storage node;

If not, the client's local hardware acceleration network card uses an accelerator to encode the metadata corresponding to the request into a network data stream, and reports the network data stream and the request corresponding to the request type of the request. The headers are combined to obtain the request message and send the request message to the storage node.

Preferably, determining whether the request needs to carry metadata includes:

Determine whether there is a metadata reference in the request content of the request;

If so, it is determined that the request needs to carry metadata;

If not, it is determined that the request does not need to carry metadata.

Preferably, after receiving the request message, the storage node processes the request message using the local metadata service, the shared cache and the hardware acceleration network card, and sends the request message to the The client sends a response message, including:

After the storage node receives the request message, the local hardware acceleration network card is used to identify whether the request message includes metadata. If so, the network data stream is decoded into a metadata cache format based on the accelerator, And store it in the shared cache, and send the request to the metadata service;

The metadata service responds to the request and generates a response message based on the request and the metadata in the shared cache;

If the request message does not include metadata, the request is directly sent to the metadata service for response and a response message is generated;

If metadata needs to be added to the response message, the hardware acceleration network card can read the corresponding metadata from the shared cache according to the metadata reference, encode the metadata into a network data stream, and convert the metadata into a network data stream. The network data flow is combined with the corresponding response message header to obtain a response message with metadata, and is sent to the client.

Preferably, the client receives the response message and processes the request result, including:

The client's local hardware accelerator network card identifies whether the response message includes metadata. If so, the accelerator is used to decode the metadata in the response message from the network data stream into a metadata cache format and store it in the The local shared cache and the response message are sent to the client;

The client determines the request result and outputs the result according to the response message and the metadata in the local shared cache.

Preferably, the request is a Lookup request, a Getattr request, a Readdir request, a Setattr request, a Setxattr request or a Create request.

Preferably, after the client generates a request, the local hardware acceleration network card of the client sends a request message to the storage node, including:

The client generates a lookup request; wherein the lookup request is a request to query whether the metadata of the target path exists;

The hardware acceleration network card local to the client encapsulates the lookup request into a lookup request message, and sends the lookup request message to the storage node;

Correspondingly, after receiving the request message, the storage node uses the local metadata service, the shared cache and the hardware acceleration network card to process the request message, and sends the request message to the The client sends a response message, including:

The hardware acceleration network card local to the storage node distributes the lookup request message to the local metadata service for processing, and retrieves the metadata cache through the path;

If the metadata cache is not retrieved, a lookup request response message without metadata is sent to the client;

If the metadata cache is retrieved, the cache reference is associated with the lookup request and the response processing thread is called;

The response processing thread calls the local hardware acceleration network card. After identifying the metadata cache associated with the response to the lookup request, the metadata cache address pointer is submitted to the FPGA for encoding and the network data stream is output;

Combine the network data stream and the response header of the lookup request to obtain a lookup request response message with metadata, and send it to the client;

Correspondingly, the client receives the response message and processes the request results, including:

The hardware acceleration network card local to the client receives the lookup request response message and identifies whether the lookup request response message includes metadata information;

If so, decoding of metadata information is triggered. The local hardware acceleration network card of the client applies for an empty cache from the local shared cache, decodes the network data stream into the metadata cache format, and then sends it to the client. response result.

Preferably, after the client generates a request, the local hardware acceleration network card of the client sends a request message to the storage node, including:

The client generates a getattr request; wherein the getattr request is a request for whether the target id metadata exists;

The hardware acceleration network card local to the client encapsulates the getattr request into a getattr request message, and sends the getattr request message to the storage node;

Correspondingly, after receiving the request message, the storage node uses the local metadata service, the shared cache and the hardware acceleration network card to process the request message, and sends the request message to the The client sends a response message, including:

The hardware acceleration network card local to the storage node distributes the getattr request message to the local metadata service for processing, and retrieves the metadata cache through the id;

If the metadata cache is not retrieved, a getattr request response message without metadata is sent to the client;

If the metadata cache is retrieved, the cache reference is associated with the getattr request and the response processing thread is called;

The response processing thread calls the local hardware acceleration network card, and after identifying the metadata cache associated with the response to the getattr request, submits the metadata cache address pointer to the FPGA for encoding and outputs the network data stream;

Combine the network data stream and the response message header of the getattr request to obtain a getattr request response message with metadata, and send it to the client;

Correspondingly, the client receives the response message and processes the request result, including:

The hardware acceleration network card local to the client receives the getattr request response message and identifies whether the getattr request response message includes metadata information;

If so, decoding of metadata information is triggered. The local hardware acceleration network card of the client applies for an empty cache from the local shared cache, decodes the network data into the metadata cache format and sends the response result to the client. .

Preferably, after the client generates a request, the local hardware acceleration network card of the client sends a request message to the storage node, including:

The client generates a readdir request; wherein the readdir request is a request to query all metadata information of the target directory;

The hardware acceleration network card local to the client encapsulates the readdir request into a readdir request message, and sends the readdir request message to the storage node;

Correspondingly, after receiving the request message, the storage node uses the local metadata service, the shared cache and the hardware acceleration network card to process the request message, and sends the request message to the The client sends a response message, including:

The hardware acceleration network card local to the storage node distributes the readdir request message to the local metadata service for processing, retrieves all metadata caches through the id of the target directory, and selects according to the offset of readdir. The returned metadata cache collection;

If the metadata cache set is not retrieved, a readdir request response message without metadata is sent to the client;

When the metadata cache collection is retrieved, the cache reference is associated with the readdir request and the response processing thread is called;

The response processing thread calls the local hardware acceleration network card, identifies the metadata cache associated with the response to the readdir request, then submits the metadata cache in the metadata cache set to FPGA encoding, and outputs the network data stream;

Combine the network data stream and the response message header of the readdir request to obtain a readdir request response message with metadata, and send it to the client;

Correspondingly, the client receives the response message and processes the request result, including:

The hardware acceleration network card local to the client receives the readdir request response message and identifies whether the readdir request response message includes metadata information;

If so, decoding of metadata information is triggered. The local hardware acceleration network card of the client applies for multiple empty caches from the local shared cache, decodes the network stream into a metadata cache format and sends it to the client. response result.

Preferably, after the client generates a request, the local hardware acceleration network card of the client sends a request message to the storage node, including:

The client generates a create request; wherein the create request is a request to create target metadata;

The hardware acceleration network card local to the client encodes the metadata referenced by the create request into a network data stream, and encapsulates the network data stream and the message header of the create request into a create request message, And send the create request message to the storage node;

Correspondingly, after receiving the request message, the storage node uses the local metadata service, the shared cache and the hardware acceleration network card to process the request message, and sends the request message to the The client sends a response message, including:

The hardware acceleration network card local to the storage node recognizes that the create request message includes metadata information, triggers the FPGA to apply for an empty cache from the shared cache, and decodes the network data stream into a metadata cache format. After writing the empty cache, send the create request to the metadata service;

The metadata server performs metadata creation processing according to the create request and the target metadata in the shared cache, and associates the target metadata to the file metadata cache; after the creation is completed, it accelerates the local hardware to The network card returns a create request response message;

The local hardware acceleration network card of the client receives the create request response message and directly sends the create request response message to the client;

Correspondingly, the client receives the response message and processes the request results, including:

The local hardware acceleration network card of the client receives the create request response message and sends the create request response message to the client.

Preferably, after the client generates a request, the local hardware acceleration network card of the client sends a request message to the storage node, including:

The client generates a setattr request; wherein the setattr request is a request to update the metadata of the target file or target directory;

The hardware acceleration network card local to the client encodes the metadata referenced by the setattr request into a network data stream, encapsulates the setattr request message header and the network data stream into a setattr request message, and sends the setattr request message. Request the message to the storage node;

Correspondingly, after receiving the request message, the storage node uses the local metadata service, the shared cache and the hardware acceleration network card to process the request message, and sends the request message to the The client sends a response message, including:

The hardware acceleration network card local to the storage node recognizes that the setattr request message includes metadata information, triggers the FPGA to apply for an empty cache from the shared cache, and decodes the network data stream into a metadata cache format. After writing the empty cache, send the setattr request to the metadata service;

After the metadata server performs metadata replacement processing based on the setattr request and the target metadata in the shared cache, it returns a setattr request response message to the local hardware acceleration network card;

The local hardware acceleration network card of the storage node receives the setattr request response message and directly sends the setattr request response message to the client;

Correspondingly, the client receives the response message and processes the request results, including:

The local hardware acceleration network card of the client receives the setattr request response message and directly sends the response result to the client.

A readable storage medium. A computer program is stored on the readable storage medium. When the computer program is executed by a processor, the steps of the above metadata processing method are implemented.

Applying the method provided by the embodiment of the present invention to a hardware acceleration network card includes: receiving a message corresponding to a metadata request, determining whether the message includes metadata; if not, sending the metadata request to the metadata service, and Use the metadata service to process the metadata request; if so, use the accelerator to decode the metadata network data stream into the metadata cache format and store it in the shared cache; send the metadata request to the metadata service, use the metadata service Performs request processing on metadata in the shared cache.

In the present invention, after receiving the message corresponding to the metadata request, the hardware acceleration network card first determines whether the message includes metadata. If there is no metadata in the message, the metadata request is directly sent to the metadata service so that the metadata service can process the metadata request. If there is metadata in the message, the accelerator is used to decode the metadata from the network data stream into the metadata cache format, write it to the shared cache, and then send the metadata request to the metadata service. In this way, the metadata service can handle requests directly based on the metadata in the shared cache.

The technical effect of the present invention: offloading the link that consumes the most CPU processing performance in the small file processing process, that is, the deserialization of metadata, to the hardware acceleration network card for processing can directly reduce the CPU processing workload, and can also use the The hardware acceleration performance breaks through the bottleneck of small file processing efficiency.

Correspondingly, embodiments of the present invention also provide a metadata processing method, a hardware acceleration network card, a system, and a readable storage medium corresponding to the above metadata processing method, which have the above technical effects and will not be described again here.

Description of the drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention or related technologies, the drawings needed to be used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the drawings in the following description are only for illustration purposes. For some embodiments of the invention, those of ordinary skill in the art can also obtain other drawings based on these drawings without exerting creative efforts.

Figure 1 is an implementation flow chart of a metadata processing method in an embodiment of the present invention;

Figure 2 is a schematic diagram of metadata deserialization processing in an embodiment of the present invention;

Figure 3 is a flow chart of serialization implementation in a metadata processing method in an embodiment of the present invention;

Figure 4 is a schematic diagram of metadata serialization processing in an embodiment of the present invention;

Figure 5 is a schematic structural diagram of a hardware acceleration network card in an embodiment of the present invention;

Figure 6 is a schematic structural diagram of a metadata processing system in an embodiment of the present invention;

Figure 7 is a specific implementation flow chart of a metadata processing method applied to a metadata processing system in an embodiment of the present invention;

Figure 8 is a schematic interactive flow diagram of a metadata processing method in an embodiment of the present invention.

Detailed ways

In order to enable those skilled in the art to better understand the solution of the present invention, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention.

In order to facilitate those skilled in the art to better understand and implement the technical solutions provided by the embodiments of the present invention, the relevant terms are explained below:

Distributed File System: Distributed File System (DFS) means that the physical storage resources managed by the file system are not necessarily directly connected to the local node, but are connected to the node through the computer network (that is, the node can be regarded as a computer ) are connected; or a complete hierarchical file system is formed by combining several different logical disk partitions or volumes.

File metadata service: Common metadata management can be divided into centralized and distributed metadata management architectures. The centralized metadata management architecture uses a single metadata server, which is simple to implement, but has problems such as single points of failure. The distributed metadata management architecture disperses metadata across multiple nodes.

Metadata: Data describing data, descriptive information about data and information resources.

File metadata: File system metadata is data that describes file data. For example, the inode (mode) of the linux file system. File metadata is the attribute information of the file, including: File type: used to indicate the type of file, such as ordinary files, directories, linked files, etc.

Permissions: What operations a user or group can perform on a file (read, write, execute).

Number of links: The number of hard links to the file.

UID: The ID of the file owner, that is, the user ID.

GID: The ID of all groups of the file, that is, the group ID.

File size: The storage space occupied by the file.

Timestamp: the time when the file was last accessed (Access time); the time when the file was last modified (Modify time, mtime); the time when the file inode was last modified (Change time, ctime).

Access control list: Special user or group permissions, such as the permissions a certain user has on the file.

File data layout: index information of data blocks. The file contains storage location information or address of data blocks.

The distributed storage architecture consists of three parts: client, metadata server and data server. The client is responsible for sending read and write requests, caching file metadata and file data; the metadata server is responsible for managing metadata and processing client requests. , is the core component of the entire system; the data server is responsible for storing file data and ensuring the availability and integrity of the data.

FPGA: It is a semi-customized circuit (Field-Programmable GateArray) among application-specific integrated circuits. It is a programmable logic array.

DPU: Data Processing Unit is a data-centered computing processor that integrates complete data center functions on a single chip.

A metadata processing method in the embodiment of the present invention can be applied to hardware accelerated network cards. Since the serialization and deserialization of metadata can occur in the client or in the storage node in the service cluster, the hardware The acceleration network card can be located in the client or in the storage node of the service cluster, that is, the method can be implemented in the client or the storage node respectively, or the method can be implemented in the client or the storage node at the same time. Regardless of which device it is implemented on, refer to Figures 1 and 2.

The following takes the hardware acceleration network card as the data processor and the accelerator as the FPGA as an example to describe the metadata processing method in detail. For the implementation modes of other hardware acceleration network cards and accelerators, you can refer to this and will not go into details here.

S101. Receive the message corresponding to the metadata request and determine whether the message includes metadata.

After receiving the message corresponding to the metadata request, you can first determine whether the message includes metadata.

Generally, if the received message includes metadata, the metadata that exists in the network data stream format in the report needs to be decoded/deserialized into data in the metadata cache format.

Therefore, in the present invention, after receiving the message corresponding to the metadata request, it is necessary to identify and determine whether the message includes metadata.

The message corresponding to the metadata request is a request message or a response message corresponding to the metadata request. Since the client may receive a message carrying metadata sent by the server, and the server may also receive a message carrying metadata sent by the client, in this embodiment, the metadata request corresponds to The message can be specifically a request message sent by the client, or a response message fed back by the server.

In a specific implementation of the present invention, determining whether a message includes metadata includes:

Read the header type of the message;

Determine whether the header type is Lookup response, Getattr response, Readdir response, Setattr sending, Setxattr sending or Create sending;

If so, it is determined that the message includes metadata;

If not, it is determined that metadata is not included in the message.

For the convenience of description, the above steps are combined for description below.

Generally, for Lookup, metadata is usually carried in the response message fed back by the server;

For Getattr, metadata is usually carried in the response message fed back by the server; for Readdir, metadata is usually carried in the response message fed back by the server; for Setattr, metadata is usually carried in the request message sent by the client; for Readdir, metadata is usually carried in the response message fed back by the server. Setxattr, usually the request message sent by the client will carry metadata; for Create, usually the request message sent by the client will carry metadata; therefore, it can be based on whether the header type in the message header belongs to Lookup response or Getattr response , Readdir response, Setattr sending, Setxattr sending or Create sending, if it belongs, it can be determined that the current message contains metadata, otherwise, it is none.

When it is determined that the judgment result is no, step S102 can be executed; when it is determined that the judgment result is yes, step S103 can be executed.

S102. Send the metadata request to the metadata service, and use the metadata service to process the metadata request.

Specifically, the metadata request message can be sent directly to the metadata service.

After receiving the metadata request message, the metadata service can perform relevant metadata processing.

S103. Use an accelerator to decode the metadata network data stream into a metadata cache format and store it in the shared cache.

In this embodiment, the accelerator can be a conventional accelerator that can implement serialization and deserialization.

The accelerator can be used to decode metadata in network data stream format into metadata cache format and store it in the shared cache.

Decoding metadata from network data stream format to metadata cache format is deserialization of metadata; conversely, encoding metadata from metadata cache format to network data stream format is serialization of metadata.

Among them, after decoding the metadata, the identification (ino), name (name), mode (mode), user ID (uid), user group ID (Gid), creation time (Ctime), and modification time (Mtime) can be obtained , access time (Atime), size (Size) and other information.

In a specific implementation manner of the present invention, in order to facilitate the use of the shared cache, the shared cache can also be created by performing the following steps:

Create a shared cache pool with the host;

Divide the shared cache pool into multiple units according to fixed allocation sizes;

Among them, each unit serves as a cache of file metadata.

That is to say, you can create a shared cache pool that can be used by both the DPU and the host. In the shared cache pool, the shared cache pool is divided into n units according to a fixed allocation size, such as mKB, and each unit is used as a Cache of file metadata.

In a specific implementation manner of the present invention, an accelerator is used to decode the network data stream of metadata into a metadata cache format and then store it in the shared cache, including:

Use the accelerator to apply for an empty cache from the shared cache;

After decoding the network data stream of metadata into the metadata cache format, it is stored in the empty cache.

That is, when the accelerator is required to deserialize metadata, the accelerator can apply for an empty cache from the shared cache, and then directly decode the network data stream of metadata into the metadata cache format and store it in the requested Empty cache. It should be noted that when the accelerator applies for an empty cache, it can apply according to actual needs. For example, when it needs to store one file metadata, it applies for an empty cache. When it needs to store multiple file metadata, it applies for multiple files. an empty cache.

S104. Send a metadata request to the metadata service, and use the metadata service to perform request processing on the metadata in the shared cache.

After metadata deserialization is complete, a metadata request can be sent to the metadata service.

The metadata service can then handle requests directly based on the metadata in the shared cache.

Specifically, for how the metadata service handles requests, you can refer to the relevant metadata processing process, which will not be described in detail here.

Applying the method provided by the embodiment of the present invention to a hardware acceleration network card includes: receiving a message corresponding to a metadata request, determining whether the message includes metadata; if not, sending the metadata request to the metadata service, and Use the metadata service to process the metadata request; if so, use the accelerator to decode the metadata network data stream into the metadata cache format and store it in the shared cache; send the metadata request to the metadata service, use the metadata service Performs request processing on metadata in the shared cache.

In the present invention, after receiving the message corresponding to the metadata request, the hardware acceleration network card first determines whether the message includes metadata. If there is no metadata in the message, the metadata request is directly sent to the metadata service so that the metadata service can process the metadata request. If there is metadata in the message, the accelerator is used to decode the metadata from the network data stream into the metadata cache format, write it to the shared cache, and then send the metadata request to the metadata service. In this way, the metadata service can handle requests directly based on the metadata in the shared cache.

The technical effect of the present invention: offloading the link that consumes the most CPU processing performance in the small file processing process, that is, the deserialization of metadata, to the hardware acceleration network card for processing can directly reduce the CPU processing workload, and can also use the The acceleration performance of hardware breaks through the bottleneck of small file processing efficiency.

It should be noted that, based on the above embodiments, the embodiments of the present invention also provide corresponding improvement solutions. In the preferred/improved embodiments, the same steps or corresponding steps as in the above-mentioned embodiments may be referred to each other, and the corresponding beneficial effects may also be referred to each other. They will not be described one by one in the preferred/improved embodiments herein.

Please refer to Figure 3 and Figure 4 for the specific processing process, including:

S201. After the metadata service or the client generates a sending request, determine whether the sending message corresponding to the sending request includes the target metadata.

When sending a request or a response to a request, it often involves carrying metadata, which requires converting the metadata into a network data stream form to form a message for transmission.

Since the message carrying metadata may be a request sent by the client or a feedback response from the server, after the metadata service or the client generates a send request, it is first clear whether the send message corresponding to the send request includes the target. metadata.

In a specific implementation manner of the present invention, determining whether the sending message corresponding to the sending request includes target metadata includes:

Read the header type of the sent message;

Determine whether the header type is Lookup response, Getattr response, Readdir response, Setattr sending, Setxattr sending or Create sending;

If yes, it is determined that the target message includes target metadata;

If not, it is determined that the target metadata is not included in the target message.

That is to say, when the header type of the sent message is specifically Lookup response, Getattr response, Readdir response, Setattr sending, Setxattr sending or Create sending, it can be determined that the message needs to carry target metadata, so the operation of step S202 can be performed; Otherwise, it means that the sent message does not need to carry target metadata, and the message corresponding to the send request can be sent directly.

The target metadata here can be the metadata of one file, the metadata of multiple files, or the metadata of a directory, etc.

S202. If so, use the accelerator to pull the target metadata cache that needs to be serialized from the shared cache by sending the requested metadata reference, and encode the target metadata cache into a network data stream to obtain the target message.

When it is clear that the target metadata needs to be sent, the target metadata cache that requires the sequence number can be pulled from the shared cache based on the metadata reference in the send request, and then the target metadata cache can be encoded into a network data stream. Then, the message header corresponding to the sending request and the network data flow are assembled to obtain the target message.

S203. Send the target message.

If the send request is a request sent by the client, the target message is sent to the server; if the send request is a corresponding send request sent by the server, the target message is sent to the client.

Corresponding to the above method embodiments, embodiments of the present invention also provide a hardware acceleration network card. The hardware acceleration network card described below and the metadata processing method described above can be mutually referenced.

As shown in Figure 5, the hardware acceleration network card includes:

The steps of the above metadata processing method can be implemented in the hardware acceleration network card 300;

The hardware acceleration network card processes the serial number and deserialization of metadata through the accelerator 301.

Since the steps of the above metadata processing method can be implemented in the hardware acceleration network card, the hardware acceleration network card also has the technical effects of the above metadata processing method, which will not be described again here.

Corresponding to the above method embodiments, embodiments of the present invention also provide a metadata processing system. The metadata processing system described below, the metadata processing method and the hardware acceleration network card described above can correspond to each other.

Referring to Figure 6, the metadata processing system includes:

Client 400 and service cluster 500;

Among them, the client includes client 401, shared cache 402 and hardware acceleration network card 300;

The service cluster includes a storage node 510, and the storage node includes a metadata service 511, a shared cache 512, and a hardware acceleration network card 300.

In Figure 4, both the hardware accelerator network card of the client and the hardware acceleration network card of the storage node can be the hardware accelerator network card shown in Figure 3.

Since the metadata processing system has the hardware acceleration network card provided by the embodiment of the present invention, and because the metadata processing method provided by the above embodiment can be implemented on the hardware acceleration network card, the metadata processing system also has the metadata processing method. The technical effects of the method will not be repeated here.

Please refer to Figure 7, which is a specific implementation flow chart of a metadata processing method applied to a metadata processing system in an embodiment of the present invention. The method includes:

S701. After the client generates a request, the client's local hardware acceleration network card sends a request message to the storage node.

Among them, the request is a Lookup request, Getattr request, Readdir request, Setattr request, Setxattr request or Create request. In other words, the metadata processing method provided by the embodiment of the present invention can cope with the metadata serial number and deserialization involved in current metadata processing types.

In a specific implementation manner of the present invention, after the client generates a request, the client's local hardware acceleration network card sends a request message to the storage node, including:

Determine whether the request needs to carry metadata;

If so, the client's local hardware acceleration network card sends a request message corresponding to the requested request type to the storage node;

If not, the client's local hardware acceleration network card uses the accelerator to encode the metadata corresponding to the request into a network data stream, and combines the network data stream with the request header corresponding to the requested request type to obtain the corresponding request report. message and send the request message to the storage node.

In other words, when the client side generates a request message, if it needs to carry metadata, the metadata can be serialized by the hardware acceleration network card.

Among them, determine whether the request needs to carry metadata, including:

Determine whether there is a metadata reference in the request content of the request;

If so, it is determined that the request needs to carry metadata;

If not, it is determined that the request does not need to carry metadata.

That is to say, when there is a metadata reference in the request content, it means that the request message needs to carry the corresponding metadata. When encoding metadata, the corresponding metadata can be found and encoded based on the metadata reference.

S702. After receiving the request message, the storage node uses the local metadata service, shared cache and hardware acceleration network card to process the request message, and sends a response message to the client.

The creation of the shared cache may refer to the creation process described in the above embodiment.

The storage node can be any storage node in the storage cluster.

The specific processing and generation of response messages can refer to the relevant metadata processing flow, and will not be described in detail here.

In a specific implementation manner of the present invention, after receiving the request message, the storage node uses the local metadata service, shared cache and hardware acceleration network card to process the request message, and sends a response message to the client. Articles, including:

After the storage node receives the request message, it uses the local hardware acceleration network card to identify whether the request message contains metadata. If so, the network data stream is decoded into the metadata cache format based on the accelerator and stored in the shared cache. and sends the request to the metadata service;

The metadata service responds to the request and generates a response message based on the request and the metadata in the shared cache;

If the request message does not include metadata, the request is sent directly to the metadata service for response and a response message is generated;

If metadata needs to be added to the response message, the hardware acceleration network card can read the corresponding metadata from the shared cache based on the metadata reference, encode the metadata into a network data stream, and combine the network data stream with the corresponding response message header. Combine them to obtain a response message with metadata and send it to the client.

That is to say, when the response message needs to carry metadata, the hardware-accelerated network card reads the corresponding metadata from the shared cache based on the metadata reference, and then encodes the metadata into a network data stream. By combining the network data flow and the response message header, a response message with metadata can be obtained. It can be seen that the process of serializing metadata to generate response messages and the process of serializing metadata to generate request messages can be referenced from each other.

S703. The client receives the response message and processes the request result.

The hardware acceleration network card in the client receives the response message, and then identifies the response message. If it contains metadata information, it converts the metadata from the network data stream format into a metadata cache structure and stores it concurrently. into the shared cache so that the client can read the metadata in the shared cache. After the client reads the metadata, it can feed back the processing results to the visual interface for users to view.

In a specific implementation manner of the present invention, the client receives the response message and processes the request result, including:

The client's local hardware accelerator network card identifies whether the response message includes metadata. If so, the accelerator is used to decode the metadata in the response message from the network data stream into a metadata cache format and store it in the local shared cache. The response message is sent to the client;

The client determines the request result and displays the output based on the response message and the metadata in the local shared cache.

That is to say, if there is metadata in the response message, the accelerator is used to decode the metadata; if there is no metadata in the response message, the accelerator may need to process it, and then the response message can be directly forwarded to the client.

Applying the method provided by the embodiment of the present invention includes: after the client generates a request, the client's local hardware acceleration network card sends a request message to the storage node; the storage node, after receiving the request message, uses the local metadata service , shared cache and hardware acceleration network card process the request message and send a response message to the client. The client receives the response message and processes the request results.

In the present invention, after the client generates a request, the client's local hardware acceleration network card stores or sends the request message. After the storage node receives the request message, the local metadata service, shared cache and hardware acceleration network card process the request message and send a response message to the client. The client receives the response message and processes the request results.

The technical effect of the present invention: the link that consumes the most CPU processing performance in the small file processing process, that is, the deserialization of metadata, can be offloaded to the hardware acceleration network card for processing, which can directly reduce the CPU processing workload, and can also use the help of Based on the acceleration performance of hardware, it breaks through the bottleneck of small file processing efficiency.

In order to facilitate those skilled in the art to better understand the metadata processing method provided by the embodiment of the present invention, the following is an example of the processing process of different metadata requests:

In the specific implementation of the present invention, for a lookup request, the processing process is as follows:

After the client generates a request, the client's local hardware acceleration network card sends a request message to the storage node, including:

The client generates a lookup request; where the lookup request is a request to query whether the metadata of the target path exists;

The client's local hardware acceleration network card encapsulates the lookup request into a lookup request message, and sends the lookup request message to the storage node;

Accordingly, after receiving the request message, the storage node uses the local metadata service, shared cache and hardware acceleration network card to process the request message, and sends a response message to the client, including:

The local hardware acceleration network card of the storage node distributes the lookup request message to the local metadata service for processing, and retrieves the metadata cache through the path;

If the metadata cache is not retrieved, a lookup request response message without metadata is sent to the client;

If the metadata cache is retrieved, the cache reference is associated with the lookup request and the response processing thread is called;

The response processing thread calls the local hardware acceleration network card. After identifying the metadata cache associated with the response to the lookup request, it submits the metadata cache address pointer to the FPGA for encoding and outputs the network data stream;

Combine the network data flow and the response header of the lookup request to obtain the lookup request response message with metadata and send it to the client;

Accordingly, the client receives the response message and processes the request results, including:

The client's local hardware acceleration network card receives the lookup request response message and identifies whether the lookup request response message includes metadata information;

If so, the decoding of the metadata information is triggered. The client's local hardware acceleration network card applies for an empty cache from the local shared cache, decodes the network data stream into the metadata cache format, and sends the response result to the client.

That is to say, the client sends a lookup request with the purpose of querying whether the metadata of a certain path exists; the client encapsulates the request into a request and sends it to the metadata service; the node where the metadata service is located, the DPU receives the client's request , identify whether metadata information is included; Lookup does not contain metadata information, and is directly sent to the metadata service for processing, and the metadata cache is retrieved through the path; the metadata cache is not retrieved, and the client message is directly responded to; the metadata cache is retrieved , associate the cache reference to the request, and call the response processing; the response processing calls the DPU, recognizes that the response is associated with the metadata cache, submits the metadata cache address pointer to the FPGA encoding, and outputs the data stream; transfer the data stream of g) and the response message header, together form a response message and send it to the client; the client DPU receives the response message and identifies whether it contains metadata information; if the metadata cache is not retrieved, no metadata information is included; if it is The metadata cache is retrieved and needs to be decoded; the decoding of the metadata information is triggered, and the empty metadata cache applied by the DPU converts the network stream into a metadata cache data structure; the response result is sent.

In the specific implementation of the present invention, for the getattr request, the processing process is as follows:

After the client generates a request, the client's local hardware acceleration network card sends a request message to the storage node, including:

The client generates a getattr request; where the getattr request is a request for whether the target id metadata exists;

The client's local hardware acceleration network card encapsulates the getattr request into a getattr request message, and sends the getattr request message to the storage node;

Accordingly, after receiving the request message, the storage node uses the local metadata service, shared cache and hardware acceleration network card to process the request message, and sends a response message to the client, including:

The local hardware acceleration network card of the storage node distributes the getattr request message to the local metadata service for processing, and retrieves the metadata cache through the id;

If the metadata cache is not retrieved, a getattr request response message without metadata is sent to the client;

If the metadata cache is retrieved, the cache reference is associated with the getattr request and the response processing thread is called;

The response processing thread calls the local hardware acceleration network card. After identifying the metadata cache associated with the response to the getattr request, it submits the metadata cache address pointer to the FPGA for encoding and outputs the network data stream;

Combine the network data stream and the response header of the getattr request to obtain the getattr request response message with metadata and send it to the client;

Accordingly, the client receives the response message and processes the request results, including:

The client's local hardware acceleration network card receives the getattr request response message and identifies whether the getattr request response message includes metadata information;

If so, the decoding of the metadata information is triggered. The client's local hardware acceleration network card applies for an empty cache from the local shared cache, decodes the network data into the metadata cache format and sends the response result to the client.

That is to say, the client sends a getattr request to query whether a certain id metadata exists; the client encapsulates the request into a request and sends it to the metadata service; on the node where the metadata service is located, the DPU receives the client's request and identifies whether it contains Metadata information; getattr does not contain metadata information, and is directly sent to the metadata service for processing, and the metadata cache is retrieved through the id; the metadata cache is retrieved, the cache reference is associated with the request, and response processing is called; response processing is called DPU recognizes that the response is associated with the metadata cache, submits the metadata cache address pointer to the FPGA for encoding, and outputs the data stream; the converted network data stream and the response message header are combined into a response message and sent to the client; The client DPU receives the response message and identifies whether it contains metadata information; if the metadata cache is not retrieved, the metadata information is not included; if the metadata cache is retrieved, decoding is required; triggering the decoding of the metadata information, The empty metadata cache requested by the DPU converts the network stream into a metadata cache data structure and sends the response result.

In the specific implementation of the present invention, for readdir requests, the processing process is as follows:

After the client generates a request, the client's local hardware acceleration network card sends a request message to the storage node, including:

The client generates a readdir request; where the readdir request is a request to query all metadata information of the target directory;

The client's local hardware acceleration network card encapsulates the readdir request into a readdir request message, and sends the readdir request message to the storage node;

Accordingly, after receiving the request message, the storage node uses the local metadata service, shared cache and hardware acceleration network card to process the request message, and sends a response message to the client, including:

The local hardware acceleration network card of the storage node distributes the readdir request message to the local metadata service for processing, retrieves all metadata caches through the ID of the target directory, and selects the returned metadata cache set based on the offset of readdir;

If the metadata cache set is not retrieved, a readdir request response message without metadata is sent to the client;

When the metadata cache collection is retrieved, the cache reference is associated with the readdir request and the response processing thread is called;

The response processing thread calls the local hardware acceleration network card, identifies the metadata cache associated with the response to the readdir request, and then submits the metadata cache in the metadata cache set to FPGA encoding and outputs the network data stream;

Combine the network data stream and the response header of the readdir request to obtain the readdir request response message with metadata and send it to the client;

Accordingly, the client receives the response message and processes the request results, including:

The client's local hardware acceleration network card receives the readdir request response message and identifies whether the readdir request response message includes metadata information;

If so, the decoding of the metadata information is triggered. The client's local hardware acceleration network card applies for multiple empty caches from the local shared cache, decodes the network stream into the metadata cache format and sends the response result to the client.

That is to say, the client sends a readdir request to query all metadata information of a certain directory; the client encapsulates the request into a request and sends it to the metadata service; on the node where the metadata service is located, the DPU receives the client's request and identifies whether Contains metadata information; readdir does not contain metadata information, and is directly sent to the metadata service for processing. All included are retrieved through the directory id, and the metadata cache set is returned based on the offset of readdir; the metadata cache set is retrieved , associate the cache set reference to the request, and call the response processing; the response processing calls the DPU, recognizes that the response is associated with the metadata cache, and submits the metadata cache in the metadata cache set one by one to the FPGA for encoding and outputs Network data flow; combine the network data flow and the response message header to form a response message and send it to the client; the client DPU receives the response message and identifies whether it contains metadata information; if the metadata cache is retrieved, it needs to be Decoding; triggering the decoding of metadata information, multiple empty metadata caches requested by the DPU convert the network flow of readdir response messages into metadata cache data structures one by one; sending the response results.

In the specific implementation of the present invention, for the create request, the processing process is as follows: after the client generates the request, the client's local hardware acceleration network card sends a request message to the storage node, including:

The client generates a create request; where the create request is a request to create target metadata;

The client's local hardware acceleration network card encodes the metadata referenced by the create request into a network data stream, encapsulates the network data stream and the header of the create request into a create request message, and sends the create request message to the storage node;

Accordingly, after receiving the request message, the storage node uses the local metadata service, shared cache and hardware acceleration network card to process the request message, and sends a response message to the client, including:

The local hardware acceleration network card of the storage node recognizes that the create request message includes metadata information, triggers the FPGA to apply for an empty cache from the shared cache, decodes the network data stream into the metadata cache format and writes it into the empty cache, and then sends the create request. to metadata services;

The metadata server performs metadata creation processing based on the create request and the target metadata in the shared cache, and associates the target metadata to the file metadata cache; after the creation is completed, it returns a create request response message to the local hardware acceleration network card;

The local hardware acceleration network card receives the create request response message and directly sends the create request response message to the client;

Accordingly, the client receives the response message and processes the request results, including:

The client's local hardware acceleration network card receives the create request response message and sends the create request response message to the client.

That is to say, the client sends a create request to create a certain name metadata; the client encapsulates the request into a request, and the DPU recognizes that the request refers to the metadata cache; the FPGA encodes the referenced metadata into a network data stream, and the DPU The message header and the metadata data stream are sent to the metadata service together; at the node where the metadata service is located, the DPU receives the client's request and identifies whether it contains metadata information; create contains metadata information, triggers FPGA decoding, and FPGA applies for an empty Metadata caching decodes the network data stream into a metadata cache; the DPU sends and distributes create requests to the metadata service; the metadata service performs creation and processing based on the request and the metadata cache, and associates the decoded metadata cache to the file metadata. Data cache; after completing the processing, the metadata service returns a response message; the client DPU receives the response message, recognizes that the message does not have metadata cache information, directly sends the response message, and returns that the creation is successful.

In the specific implementation of the present invention, for the setattr request, the processing process is as follows:

After the client generates a request, the client's local hardware acceleration network card sends a request message to the storage node, including:

The client generates a setattr request; where the setattr request is a request to update the metadata of the target file or target directory;

The client's local hardware acceleration network card encodes the metadata referenced by the setattr request into a network data stream, encapsulates the setattr request message header and network data stream into a setattr request message, and sends the setattr request message to the storage node;

Accordingly, after receiving the request message, the storage node uses the local metadata service, shared cache and hardware acceleration network card to process the request message, and sends a response message to the client, including:

The local hardware acceleration network card of the storage node recognizes that the setattr request message includes metadata information, triggers the FPGA to apply for an empty cache from the shared cache, decodes the network data stream into the metadata cache format and writes it into the empty cache, and then sends the setattr request. to metadata services;

After the metadata server performs metadata replacement processing based on the setattr request and the target metadata in the shared cache, it returns a setattr request response message to the local hardware acceleration network card;

The local hardware acceleration network card receives the setattr request response message and directly sends the setattr request response message to the client;

Accordingly, the client receives the response message and processes the request results, including:

The client's local hardware acceleration network card receives the setattr request response message and directly sends the response result to the client.

That is to say, the client sends a setattr request to update the metadata information of a file or directory; the client encapsulates the request into a request, and the DPU recognizes that the request refers to the metadata cache; the FPGA encodes the referenced metadata into network data stream, DPU sends the message header and metadata data stream together to the metadata service; at the node where the metadata service is located, the DPU receives the client's request and identifies whether it contains metadata information; setattr contains metadata information and triggers FPGA decoding. The empty metadata cache requested by the DPU decodes the network data stream into a metadata cache. The DPU sends and distributes setattr requests to the metadata service; the metadata service replaces the metadata based on the request and the metadata cache. After the replacement is completed, the metadata service returns a response message. The client DPU receives the response message, recognizes that the message does not have metadata cache information, and directly sends the response message and returns setattr successfully.

Corresponding to the above method embodiments, embodiments of the present invention also provide a readable storage medium. The readable storage medium described below and the metadata processing method described above may be mutually referenced.

A readable storage medium. A computer program is stored on the readable storage medium. When the computer program is executed by a processor, the steps of the metadata processing method of the above method embodiment are implemented.

The readable storage medium can specifically be a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk that can store program codes. readable storage media.

Each embodiment in this specification is described in a progressive manner. Each embodiment focuses on its differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other. As for the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple. For relevant details, please refer to the description in the method section.

Those skilled in the art may further realize that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented with electronic hardware, computer software, or a combination of both. In order to clearly illustrate the hardware and software Interchangeability, in the above description, the composition and steps of each example have been generally described according to functions. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art may implement the described functionality using different methods for each specific application, but such implementations should not be considered to be beyond the scope of the present invention.

The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein may be implemented directly in hardware, in software modules executed by a processor, or in a combination of both. Software modules may be located in random access memory (RAM), memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disks, removable disks, CD-ROMs, or anywhere in the field of technology. any other known form of storage media.

Finally, it should be noted that in this article, relationships such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that these entities or any such actual relationship or sequence between operations. Furthermore, the terms include, include, or any variation thereof are intended to cover a non-exclusive inclusion such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not expressly listed, or It also includes elements inherent to the process, method, article or equipment.

This article uses specific examples to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method and the core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the present invention There will be changes in the specific implementation and scope of application of the ideas. In summary, the content of this description should not be understood as limiting the present invention.

Claims (22)

1. A metadata processing method, characterized in that it is applied to a hardware acceleration network card, including: Receive the message corresponding to the metadata request and determine whether the message includes metadata; If not, send the metadata request to the metadata service, and use the metadata service to process the metadata request; If so, use an accelerator to decode the metadata network data stream into a metadata cache format and then store it in the shared cache; Send the metadata request to the metadata service, and use the metadata service to perform request processing on the metadata in the shared cache. 2. The metadata processing method according to claim 1, wherein the accelerator is used to decode the network data stream of the metadata into a metadata cache format and then store it in a shared cache, including: Utilize the accelerator to apply for an empty cache from the shared cache; After the network data stream of the metadata is decoded into a metadata cache format, it is stored in the empty cache. 3. The metadata processing method according to claim 2, further comprising: Create a shared cache pool with the host; Divide the shared cache pool into multiple units according to a fixed allocation size; Each of the units serves as a cache of file metadata. 4. The metadata processing method according to claim 1, wherein the step of determining whether the message includes metadata includes: Read the header type of the message; Determine whether the header type is a Lookup response, Getattr response, Readdir response, Setattr send, Setxattr send or Create send; If so, determine that the metadata is included in the message; If not, it is determined that the metadata is not included in the message. 5. The metadata processing method according to claim 1, wherein the message corresponding to the metadata request is a request message or a response message corresponding to the metadata request. 6. The metadata processing method according to claim 1, wherein the hardware acceleration network card is a data processor, and the accelerator is an FPGA. 7. The metadata processing method according to any one of claims 1 to 6, further comprising: After the metadata service or the client generates a sending request, determine whether the sending message corresponding to the sending request includes target metadata; If so, use the accelerator to pull the target metadata cache that needs to be serialized from the shared cache through the metadata reference of the sending request, and encode the target metadata cache into a network data stream to obtain target message; Send the target message. 8. The metadata processing method according to claim 7, wherein determining whether the sending message corresponding to the sending request includes target metadata includes: Read the header type of the sent message; Determine whether the header type is a Lookup response, Getattr response, Readdir response, Setattr send, Setxattr send or Create send; If so, determine that the target message includes the target metadata; If not, it is determined that the target metadata is not included in the target message. 9. A hardware acceleration network card, characterized by including: The steps of the metadata processing method according to any one of claims 1 to 8 can be implemented in the hardware acceleration network card; The hardware acceleration network card processes the serial number and deserialization of metadata through an accelerator. 10. A metadata processing system, characterized by including: Client and service clusters; Wherein, the client computer includes a client, a shared cache and a hardware acceleration network card as claimed in claim 9; The service cluster includes a storage node, and the storage node includes a metadata service, a shared cache, and the hardware acceleration network card according to claim 9. 11. A metadata processing method, characterized in that it is applied to the metadata processing system as claimed in claim 10, comprising: After the client generates a request, the hardware acceleration network card local to the client sends a request message to the storage node; After receiving the request message, the storage node uses the local metadata service, the shared cache and the hardware acceleration network card to process the request message and sends it to the client. response message; The client receives the response message and processes the request result. 12. The metadata processing method according to claim 11, characterized in that after the client generates a request, the local hardware acceleration network card of the client sends a request message to the storage node, including: Determine whether the request needs to carry metadata; If so, the local hardware acceleration network card of the client sends a request message corresponding to the request type of the request to the storage node; If not, the client's local hardware acceleration network card uses an accelerator to encode the metadata corresponding to the request into a network data stream, and reports the network data stream and the request corresponding to the request type of the request. The headers are combined to obtain the request message and send the request message to the storage node. 13. The metadata processing method according to claim 12, characterized in that determining whether the request needs to carry metadata includes: Determine whether there is a metadata reference in the request content of the request; If so, it is determined that the request needs to carry metadata; If not, it is determined that the request does not need to carry metadata. 14. The metadata processing method according to claim 11, characterized in that, after receiving the request message, the storage node uses the local metadata service, the shared cache and the hardware The acceleration network card processes the request message and sends a response message to the client, including: After the storage node receives the request message, the local hardware acceleration network card is used to identify whether the request message includes metadata. If so, the network data stream is decoded into a metadata cache format based on the accelerator, and store it in the shared cache, and send the request to the metadata service; The metadata service responds to the request and generates a response message based on the request and the metadata in the shared cache; If the request message does not include metadata, the request is directly sent to the metadata service for response and a response message is generated; If metadata needs to be added to the response message, the hardware acceleration network card can read the corresponding metadata from the shared cache according to the metadata reference, encode the metadata into a network data stream, and convert the metadata into a network data stream. The network data flow is combined with the corresponding response message header to obtain a response message with metadata, and is sent to the client. 15. The metadata processing method according to claim 11, characterized in that the client receives the response message and processes the request result, including: The client's local hardware accelerator network card identifies whether the response message includes metadata. If so, the accelerator is used to decode the metadata in the response message from the network data stream into a metadata cache format and store it in the The local shared cache and the response message are sent to the client; The client determines the request result and outputs the result according to the response message and the metadata in the local shared cache. 16. The metadata processing method according to any one of claims 11 to 15, characterized in that the request is a Lookup request, a Getattr request, a Readdir request, a Setattr request, a Setxattr request or a Create request. 17. The metadata processing method according to claim 11, characterized in that after the client generates a request, the local hardware acceleration network card of the client sends a request message to the storage node, including: The client generates a lookup request; wherein the lookup request is a request to query whether the metadata of the target path exists; The hardware acceleration network card local to the client encapsulates the lookup request into a lookup request message, and sends the lookup request message to the storage node; Correspondingly, after receiving the request message, the storage node uses the local metadata service, the shared cache and the hardware acceleration network card to process the request message, and sends the request message to the The client sends a response message, including: The hardware acceleration network card local to the storage node distributes the lookup request message to the local metadata service for processing, and retrieves the metadata cache through the path; If the metadata cache is not retrieved, a lookup request response message without metadata is sent to the client; If the metadata cache is retrieved, the cache reference is associated with the lookup request and the response processing thread is called; The response processing thread calls the local hardware acceleration network card. After identifying the metadata cache associated with the response to the lookup request, the metadata cache address pointer is submitted to the FPGA for encoding and the network data stream is output; Combine the network data stream and the response header of the lookup request to obtain a lookup request response message with metadata, and send it to the client; Correspondingly, the client receives the response message and processes the request results, including: The hardware acceleration network card local to the client receives the lookup request response message and identifies whether the lookup request response message includes metadata information; If so, decoding of metadata information is triggered. The local hardware acceleration network card of the client applies for an empty cache from the local shared cache, decodes the network data stream into the metadata cache format, and then sends it to the client. response result. 18. The metadata processing method according to claim 11, characterized in that after the client generates a request, the local hardware acceleration network card of the client sends a request message to the storage node, including: The client generates a getattr request; wherein the getattr request is a request for whether the target id metadata exists; The hardware acceleration network card local to the client encapsulates the getattr request into a getattr request message, and sends the getattr request message to the storage node; Correspondingly, after receiving the request message, the storage node uses the local metadata service, the shared cache and the hardware acceleration network card to process the request message, and sends the request message to the The client sends a response message, including: The hardware acceleration network card local to the storage node distributes the getattr request message to the local metadata service for processing, and retrieves the metadata cache through the id; If the metadata cache is not retrieved, a getattr request response message without metadata is sent to the client; If the metadata cache is retrieved, the cache reference is associated with the getattr request and the response processing thread is called; The response processing thread calls the local hardware acceleration network card, and after identifying the metadata cache associated with the response to the getattr request, submits the metadata cache address pointer to the FPGA for encoding and outputs the network data stream; Combine the network data stream and the response message header of the getattr request to obtain a getattr request response message with metadata, and send it to the client; Correspondingly, the client receives the response message and processes the request result, including: The hardware acceleration network card local to the client receives the getattr request response message and identifies whether the getattr request response message includes metadata information; If so, decoding of metadata information is triggered. The local hardware acceleration network card of the client applies for an empty cache from the local shared cache, decodes the network data into the metadata cache format and sends the response result to the client. . 19. The metadata processing method according to claim 11, characterized in that after the client generates a request, the local hardware acceleration network card of the client sends a request message to the storage node, including: The client generates a readdir request; wherein the readdir request is a request to query all metadata information of the target directory; The hardware acceleration network card local to the client encapsulates the readdir request into a readdir request message, and sends the readdir request message to the storage node; Correspondingly, after receiving the request message, the storage node uses the local metadata service, the shared cache and the hardware acceleration network card to process the request message, and sends the request message to the The client sends a response message, including: The hardware acceleration network card local to the storage node distributes the readdir request message to the local metadata service for processing, retrieves all metadata caches through the id of the target directory, and selects according to the offset of readdir. The returned metadata cache collection; If the metadata cache set is not retrieved, a readdir request response message without metadata is sent to the client; When the metadata cache collection is retrieved, the cache reference is associated with the readdir request and the response processing thread is called; The response processing thread calls the local hardware acceleration network card, identifies the metadata cache associated with the response to the readdir request, then submits the metadata cache in the metadata cache set to FPGA encoding, and outputs the network data stream; Combine the network data stream and the response message header of the readdir request to obtain a readdir request response message with metadata, and send it to the client; Correspondingly, the client receives the response message and processes the request results, including: The hardware acceleration network card local to the client receives the readdir request response message and identifies whether the readdir request response message includes metadata information; If so, decoding of metadata information is triggered. The local hardware acceleration network card of the client applies for multiple empty caches from the local shared cache, decodes the network stream into a metadata cache format and sends it to the client. response result. 20. The metadata processing method according to claim 11, characterized in that after the client generates a request, the local hardware acceleration network card of the client sends a request message to the storage node, including: The client generates a create request; wherein the create request is a request to create target metadata; The hardware acceleration network card local to the client encodes the metadata referenced by the create request into a network data stream, and encapsulates the network data stream and the message header of the create request into a create request message, And send the create request message to the storage node; Correspondingly, after receiving the request message, the storage node uses the local metadata service, the shared cache and the hardware acceleration network card to process the request message, and sends the request message to the The client sends a response message, including: The hardware acceleration network card local to the storage node recognizes that the create request message includes metadata information, triggers the FPGA to apply for an empty cache from the shared cache, and decodes the network data stream into a metadata cache format. After writing the empty cache, send the create request to the metadata service; The metadata server performs metadata creation processing according to the create request and the target metadata in the shared cache, and associates the target metadata to the file metadata cache; after the creation is completed, it accelerates the local hardware to The network card returns a create request response message; The hardware acceleration network card local to the storage node receives the create request response message and directly sends the create request response message to the client; Correspondingly, the client receives the response message and processes the request results, including: The local hardware acceleration network card of the client receives the create request response message and sends the create request response message to the client. 21. The metadata processing method according to claim 11, characterized in that after the client generates a request, the local hardware acceleration network card of the client sends a request message to the storage node, including: The client generates a setattr request; wherein the setattr request is a request to update the metadata of the target file or target directory; The hardware acceleration network card local to the client encodes the metadata referenced by the setattr request into a network data stream, encapsulates the setattr request message header and the network data stream into a setattr request message, and sends the setattr request message. Request the message to the storage node; Correspondingly, after receiving the request message, the storage node uses the local metadata service, the shared cache and the hardware acceleration network card to process the request message, and sends the request message to the The client sends a response message, including: The hardware acceleration network card local to the storage node recognizes that the setattr request message includes metadata information, triggers the FPGA to apply for an empty cache from the shared cache, and decodes the network data stream into a metadata cache format. After writing the empty cache, send the setattr request to the metadata service; After the metadata server performs metadata replacement processing based on the setattr request and the target metadata in the shared cache, it returns a setattr request response message to the local hardware acceleration network card; The hardware acceleration network card local to the storage node receives the setattr request response message and directly sends the setattr request response message to the client; Correspondingly, the client receives the response message and processes the request result, including: The local hardware acceleration network card of the client receives the setattr request response message and directly sends the response result to the client. 22. A readable storage medium, characterized in that a computer program is stored on the readable storage medium, and when the computer program is executed by a processor, the computer program implements any one of claims 1 to 8 and 11 to 21. Describe the steps of the metadata processing method.