WO2022257685A1

WO2022257685A1 - Storage system, network interface card, processor, and data access method, apparatus, and system

Info

Publication number: WO2022257685A1
Application number: PCT/CN2022/092015
Authority: WO
Inventors: 任仁; 王晨; 叶利杰; 崔文林; 张鹏
Original assignee: 华为技术有限公司
Priority date: 2021-06-07
Filing date: 2022-05-10
Publication date: 2022-12-15

Abstract

A storage system, a network interface card, a processor, and a data access method, apparatus, and system. In the present application, the storage system comprises an I/O stack and a processing unit, and the I/O stack comprises a plurality of storage layers. The processing unit can receive a data read request, the data read request being used for reading target data stored in the storage system; the processing unit queries a global index on the basis of the data read request, wherein the global index can indicate the storage layer in the I/O stack where the target data is located; and the processing unit can read the target data from the storage layer after determining, according to the global index, the storage layer where the target data is located. When processing the data read request, the processing unit in the storage system directly determines, by querying the global index, the storage layer where the target data is located, and there is no need to sequentially traverse a storage medium in the storage system, such that delay caused due to traversing the storage medium is omitted, the processing process of the data read request is more efficient, and the data read efficiency can be effectively improved.

Description

Storage system, network card, processor, data access method, device and system

Cross References to Related Applications

This application claims the priority of the Chinese patent application with the application number 202110634011.5 and the application name "" submitted to the China Patent Office on June 07, 2021, the entire contents of which are incorporated in this application by reference; this application claims that in 2021 The priority of the Chinese patent application filed with the China Patent Office on August 17, the application number is 202110944933.6, and the application title is "storage system, network card, processor, data access method, device and system", the entire content of which is incorporated herein by reference Applying.

technical field

The present application relates to the technical field of storage, and in particular to a storage system, a network card, a processor, a data access method, device and system.

Background technique

In the storage field, when the storage system receives a data read request from a client device, the storage system needs to traverse each storage medium in the storage system in a specific order to determine whether the target data to be read is stored in each storage medium. For example, the storage system first checks whether the target data is stored in the write cache, and if not, continues to check whether the target data is stored in the read cache, and if not, then continues to search for the lower storage medium.

It can be seen that if the target data is stored in a lower-ranked storage medium, the storage system needs to traverse each storage medium that is higher in the storage system in turn until the storage medium where the target data is located can be successfully found. If the sorting position of the storage medium where the target data is located is lower, the time consumption of the data search process will be greater, so that the processing efficiency of the storage system for the data read request is reduced.

Contents of the invention

The present application provides a storage system, a network card, a processor, a data access method, device and system to improve data reading efficiency.

In the first aspect, the embodiment of the present application provides a storage system, the storage system includes an I/O stack and a processing unit, and the I/O stack includes multiple storage layers (the storage layer may be referred to simply as a layer in the embodiment of the application) , the I/O stack is formed by layering the storage media in the storage system. Each storage layer may include one or more storage media, and the data read delay of each storage layer may be different. The processing unit may be located in a device in the storage system (the device may be referred to as a storage device), and the embodiment of the present application does not limit the specific form of the processing unit. The processing unit can process a data reading request generated inside the client device or the storage system, and feed back the requested target data.

The specific process is as follows:

The processing unit receives a data read request, and the data read request is used to request to read the target data stored in the storage system; the processing unit can query the global index based on the data read request, and the global index can indicate where the target data is located in the I/O stack storage layer, the processing unit may determine the storage layer where the target data is located according to the global index. After the storage layer where the target data is located is determined, the target data can be read from the storage layer.

Through the above system, when processing a data read request, the processing unit in the storage system can directly determine the storage layer where the target data is located by querying the global index, without traversing the storage media in the storage system in order, eliminating the need to traverse the storage media The delay, the processing process of the data reading request is more efficient, which can effectively improve the data reading efficiency.

In a possible implementation manner, in addition to processing the data read request, the processing unit may also process the data write request. The specific process is as follows: the processing unit can receive the data write request, and the data write request is used to request to write the target data in the storage system; the processing unit can write the target data into the storage system according to the data write request, and can also according to The target data updates the global index, and the updated global index is used to indicate the storage layer of the target data in the I/O stack. Specifically, the so-called updating the global index refers to storing the target data in a certain storage layer, and this information is recorded in the first index item, so that when the target data needs to be read later, it can be accessed through Query the first index item in the global index to obtain the storage location of the data. The index items are called subgroups in the embodiment, and the first index item is called the first subgroup.

Through the above system, the processing unit can update the global index when the target data is written, and the updated global index can indicate the storage layer where the target data is located, so that when the processing unit needs to read the target data, it can follow the updated The global index accurately determines the storage layer where the target data resides, and reads the target data.

In a possible implementation, multiple character sub-blocks are set in the global index, each character sub-block corresponds to a storage layer in the I/O stack, and points to the logical storage space mapped by the storage medium in the storage system a space in . In other words, the value of the character sub-block can indicate whether the data in the pointed space is located in the corresponding storage layer. In essence, the global index is to establish the corresponding relationship between the logical storage space and the storage medium.

The logical storage space mapped by the storage medium in the storage system is refined, and the logical storage space is divided into large logical blocks. The logic block can be further refined, and a logic block can be divided into multiple logic sub-blocks.

The global index may include multiple character blocks, and one character block points to at least one logical block. A character block includes multiple subgroups, and each subgroup points to a logical subblock in the at least one logical block. Each subgroup includes a plurality of character subblocks, and the character subblocks correspond to a storage layer in the I/O stack.

The data read request includes the logical address of the target data, and the logical address indicates a space in the logical storage space. When the processing unit queries the global index based on the data read request, it can be determined in the global index according to the logical address of the target data A plurality of character sub-blocks pointing to the logical address of the target data; then, determine the storage layer where the target data in the I/O stack is located according to the values of the plurality of character sub-blocks.

Through the above system, the global index can associate a logical storage space (also can be understood as a logical address) with a storage layer (also can be understood as a storage medium) in the I/O stack. The processing unit can use the logical address of the target data to query the character sub-block pointing to the logical address in the global index. According to the value of the character sub-block, the storage layer where the target data is located can be conveniently and quickly determined, ensuring the accuracy of data reading. Efficiency.

In a possible implementation manner, the character sub-blocks in the global index may be represented in various forms, for example, a character sub-block may be a bit, and the value of the bit may be 0 or 1. When the value of this bit is 1, it means that the target data is located in the storage layer corresponding to the character sub-block. When the value of this bit is 0, it means that the target data is located in the storage layer corresponding to the character sub-block. If there are multiple non-zero bits, then the target data is located in the highest layer among the storage layers corresponding to the multiple non-zero bits.

For another example, a character sub-block can be a counter, and the value of the counter can be 0 or a non-zero integer. When a counter has a value of 0, it means that no data is written into the storage layer corresponding to the counter. When a counter is a non-zero integer, the non-zero integer represents the number of times data is written to the storage layer corresponding to the counter. A non-zero counter among the plurality of counters determined to point to the logical address of the target data in the global index may indicate that the target data is located in the storage layer corresponding to the non-zero counter. If there are multiple non-zero counters, then the target data is located in the highest layer among the storage layers corresponding to the multiple non-zero counters.

Through the above system, there are many kinds of presentation forms of character sub-blocks in the global index, and the presentation methods are relatively flexible and can be applied in different scenarios.

In a possible implementation, when the processing unit determines multiple character sub-blocks pointing to the logical address of the target data in the global index according to the logical address of the target data, it may perform a hash operation on the logical address of the target data, such as Query the hash table or act on the hash function, and determine multiple character sub-blocks pointing to the logical address of the target data according to the result of the hash operation.

Through the above-mentioned system, the method of hash operation is relatively simple and fast, and multiple character sub-blocks pointing to the logical address of the target data can be quickly determined to ensure the efficiency of data reading.

In a possible implementation manner, when the processing unit determines multiple character sub-blocks pointing to the logical address of the target data according to the result of the hash operation, it may first determine the logical block of the logical address pointing to the target data in the global index. The character block, and then determine a plurality of character sub-blocks pointing to the logical address of the target data from the character block according to the logical address of the target data.

Through the above system, the processing unit can first locate a character block pointing to a larger logical block, and then locate a character sub-block pointing to a smaller logical sub-block from the character block. The processing unit first locates the character blocks in a large range, and then locates the character sub-blocks in a small range, which can improve the efficiency of locating the character sub-blocks.

In a possible implementation, the processing unit determines a plurality of character sub-blocks pointing to the logical address of the target data from the character block according to the logical address of the target data. The offset between the logical blocks determines the target subgroup among the multiple subgroups in the character block, and the character subblocks in the target subgroup are multiple character subblocks pointing to the logical address of the target data. That is to say, the processing unit can determine the address indicated by the logical address in the global index according to the offset between the start address of the logical block pointed to by the character block and the logical address of the target data, and the data length of the target data. Each subgroup of the logical subblock (ie, the target subgroup). The character sub-blocks in each subgroup are the character sub-blocks pointing to logical addresses in the global index.

Through the above system, the processing unit can more accurately determine the character sub-block pointing to the logical address in the global index through the offset of the logical address in the logical block and the data length.

In a possible implementation manner, the processing unit may be located in a network card of a device in the storage system, or in a processor of a device in the storage system, and the processor may also be a data processor, or may be a A separate hardware component.

Through the above system, the network card or processor of the device in the storage device can have the function of processing data read requests, which effectively expands the application scenarios.

In a possible implementation, when the processing unit is a network card, the global index and the metadata of the target data are located in the memory of the device in the storage system, the client device can obtain the global index from the storage system through unilateral RDMA and metadata for the target data.

The processing unit may feed back the global index and the metadata of the target data to the client device under a first instruction of the client device, where the first instruction is based on RDMA transmission.

Wherein, the global index fed back by the processing unit to the client device may be the entire global index or a partial global index, for example, only all character sub-blocks or part of the character sub-blocks in the global index that only need the logical address of the target data are fed back.

Wherein, the storage system may notify the client device of the address (that is, the memory address) of the global index in the memory of the storage system in advance.

Through the above system, the client device can read the global index and the metadata of the target data through unilateral RDMA, without the participation of the processor in the storage system, which can improve the efficiency of data interaction.

In a possible implementation, when the processing unit is a network card, the global index is located in the memory of the device in the storage system, and the metadata of the target data is located in the persistent memory of the device in the storage system, the client device can use unilateral RDMA The global index is obtained from the storage system by means of direct access, and the metadata of the target data is obtained from the storage system by means of direct access.

The processing unit may feed back the global index to the client device under the second instruction of the client device, and the second instruction is based on RDMA transmission; the processing unit may also obtain the target data from the persistent storage under the third instruction of the client device The metadata of the target data is fed back to the client device.

Through the above system, the client device can read the global index through unilateral RDMA, and obtain the metadata of the target data through direct access, without the participation of the processor in the storage system, which can improve the efficiency of data interaction.

In a possible implementation, when the metadata of the target data indicates that the target data is located in the memory of the device in the storage system (wherein, the memory of the device in the storage system may belong to a layer in the I/O stack of the storage system or multiple layers). The client device can determine whether the metadata of the target data is valid through the obtained global index, that is, determine whether the storage layer indicated by the global index is consistent with the location indicated by the metadata of the target data. If they are consistent, it indicates that the metadata of the target data The data is valid, and the client device can obtain the target data from the storage system through unilateral RDMA. The processing unit may feed back the target data to the client device under a fourth instruction of the client device, where the fourth instruction is initiated according to metadata of the target data and based on RDMA transmission.

Through the above system, the storage system allows the client device to obtain the global index through unilateral RDMA, and allows the client device to obtain the target data through unilateral RDMA, which effectively simplifies the interaction process between the storage system and the client device. The processor in the storage system does not need to participate, and the occupation of the processor in the storage system can also be reduced.

In a possible implementation, when the target data is located in the persistent memory of the device in the storage system (the persistent memory of the device in the storage system may belong to one or more layers of the I/O stack of the storage system middle). The client device can determine whether the metadata of the target data is valid through the obtained global index, that is, determine whether the storage layer indicated by the global index is consistent with the location indicated by the metadata of the target data. If they are consistent, it indicates that the metadata of the target data The data is valid, and the client device can obtain target data from the storage system through direct access. The processing unit may obtain the target data from the persistent storage under the fifth instruction of the client device, and feed back the target data to the client device, where the fifth instruction is initiated according to the metadata of the target data.

Through the above system, the storage system allows the client device to obtain the global index through unilateral RDMA, and allows the client device to obtain the target data through the direct access method, so that the client device can directly access the index without going through the processor in the storage system. Get target data.

In a possible implementation, the processing unit can also control data flow in the I/O stack (data flows out of one storage layer and then flows into another storage layer) and data elimination (data in one storage layer is Delete), and global indexes can also be updated according to data flow in the I/O stack and data elimination. The processing unit may be a network card or a processor.

Through the above system, the processing unit can update the global index for data flow and data elimination in the I/O stack, so that the global index can accurately indicate the storage layer where each data is located, and ensure the accuracy and effectiveness of the data reading process.

In the second aspect, the embodiment of the present application provides a data access method, which can be executed by a processing unit in the storage system. For the beneficial effects, please refer to the first aspect and the related descriptions in any possible implementation of the first aspect. I won't repeat them here. The storage system also includes an I/O stack. For the description of the I/O stack, reference may be made to the foregoing content, and details will not be repeated here. In this method, the processing unit may receive a data read request, where the data read request is used to request to read target data stored in the storage system. After receiving the data read request, the processing unit can query the global index based on the data read request, and the global index is used to indicate the storage layer where the target data in the I/O stack is located; after that, the processing unit can according to the storage layer indicated by the global index , to read the target data.

In a possible implementation manner, the processing unit may also process the data writing request. The processing unit may receive a data write request, and the data write request is used for requesting to write target data in the storage system. Afterwards, the processing unit can write the target data into the storage system according to the data write request, and can also update the global index according to the target data, and the updated global index is used to indicate the storage layer where the target data in the I/O stack is located.

In a possible implementation, the data read request includes the logical address of the target data, and when the processing unit queries the global index based on the data read request, it can determine the logical address pointing to the target data in the global index according to the logical address of the target data. A plurality of character sub-blocks of the address; after that, determine the storage layer where the target data in the I/O stack is located according to the values of the plurality of character sub-blocks.

In a possible implementation manner, the character sub-blocks in the global index have multiple representation forms. For example, a character sub-block may be a bit, and the value of the bit may be 0 or 1. When the value of this bit is 1, it means that the target data is located in the storage layer corresponding to the character sub-block. When the value of this bit is 0, it means that the target data is located in the storage layer corresponding to the character sub-block.

For another example, a character sub-block can be a counter, and the value of the counter can be 0 or a non-zero integer. When a counter has a value of 0, it means that no data is written into the storage layer corresponding to the counter. When a counter is a non-zero integer, the non-zero integer indicates that the target data is located in the corresponding storage layer, and may also indicate the number of times data (the data includes the target data) has been written into the storage layer corresponding to the counter.

In a possible implementation manner, the global index is located in the memory of the device in the method, the metadata of the target data is located in the persistent storage of the device in the method, and the processing unit may feed back the The global index, the second indication is based on RDMA transmission.

The processing unit may also acquire the metadata of the target data from the persistent storage under the third instruction of the client device, and feed back the metadata of the target data to the client device.

The processing unit may feed back the global index to the client device under the second instruction of the client device, and the second instruction is based on RDMA transmission; and may also obtain the metadata of the target data from the persistent storage under the third instruction of the client device. Data, which feeds back metadata of the target data to the client device.

In a possible implementation manner, the processing unit may also control data flow and data elimination in the I/O stack; and update the global index according to the data flow and data elimination in the I/O stack.

In the third aspect, the embodiment of the present application also provides a network card. The network card may be a network card on a device in a storage system, and the network card has the method examples in the above-mentioned second aspect and each possible implementation manner of the second aspect. For the function and beneficial effect of the behavior, please refer to the description of the first aspect and will not go into details here.

In a fourth aspect, the embodiment of the present application also provides a processor, which may be a processor on a device in a storage system, and the processor has the following functions for realizing the above second aspect and each possible implementation manner of the second aspect The function of the behavior in the method example, the beneficial effect can refer to the description of the first aspect and will not be repeated here.

In the fifth aspect, the embodiment of the present application also provides a data access device, the data access device has the function of implementing the behavior in the method example of the second aspect above, and the beneficial effects can be referred to the description of the first aspect, which will not be repeated here. The functions may be implemented by hardware, or may be implemented by executing corresponding software through hardware. Hardware or software includes one or more modules corresponding to the above-mentioned functions. In a possible design, the structure of the device includes a transmission module, a reading module, and optionally a writing module, and a control module. These modules can perform the corresponding functions in the method example of the second aspect above, specifically Refer to the detailed description in the method example, and do not repeat them here.

In the sixth aspect, the embodiment of the present application provides a data access system. The data access system includes a storage system and a client device. The storage system includes an I/O stack and a processing unit. For the description of the I/O stack, please refer to the foregoing content. I won't repeat them here.

The client device may send a data read request to the storage system, where the data read request is used to request to read target data stored in the storage system. After receiving the data read request, the processing unit in the storage system can query the global index based on the data read request, and the global index is used to indicate the storage layer where the target data in the I/O stack is located. After that, read the target data according to the storage layer indicated by the global index, and feed back the target data to the client device.

In a possible implementation manner, the client device may also send a data write request to the storage system, where the data write request is used to request to write target data in the storage system. After receiving the data write request, the processing unit can write the target data according to the data write request, and can update the global index according to the target data, and the updated global index is used to indicate the storage layer where the target data is located in the I/O stack.

In a possible implementation, the character sub-block is a bit, and the value of the bit includes 0 or 1. 1 indicates that the target data is located in the storage layer corresponding to the character sub-block, and 0 indicates that the target data is located in the storage layer corresponding to the character sub-block. layer.

In a possible implementation manner, the character sub-block is a counter, and the counter is 0 or a non-zero integer, and the non-zero integer indicates the number of times data is written to the storage layer corresponding to the character sub-block.

In a possible implementation manner, the processing unit is a processing chip with computing power, such as a data processor, which may be located in the network card of the storage system, may also be located in the central processing unit, or may be a independent hardware components.

In a possible implementation, the processing unit determines a plurality of character sub-blocks pointing to the logical address of the target data from the character block according to the logical address of the target data. The offset between the logical blocks determines the target subgroup among the multiple subgroups in the character block, and the character subblocks in the target subgroup are multiple character subblocks pointing to the logical address of the target data.

In a possible implementation manner, when the global index and the metadata of the target data are located in the memory of the device in the system.

The client device may initiate a first indication to the storage system based on RDMA, where the first indication is used to request the global index and metadata of the target data. The processing unit may feed back the global index and the metadata of the target data to the client device under the first instruction of the client device.

It should be noted that the client device can request to obtain the entire global index, or only obtain a part of the global index. For example, the client device can determine the global index pointing to the The memory address of part or all of the character sub-blocks of the logical address, according to the memory address of the part or all of the character sub-blocks, initiates a first instruction to the storage system to request the part or all of the character sub-blocks.

In a possible implementation manner, the processing unit may notify the client device of the memory address of the global index in the storage system.

In a possible implementation manner, when the global index is located in the memory of the device in the system, the metadata of the target data is located in the persistent storage of the device in the system.

The client device may initiate a second indication to the storage system based on RDMA, where the second indication is used to request a global index; and may also initiate a third indication to the storage system, where the third indication is used to request metadata of the target data.

The processing unit may feed back the global index to the client device under the second instruction of the client device; obtain the metadata of the target data from the persistent storage under the third instruction of the client device, and feed back the metadata of the target data to the client device. data.

It should be noted that the client device can request to obtain the entire global index, or only obtain a part of the global index. For example, the client device can determine the global index pointing to the The memory address of part or all of the character sub-blocks of the logical address, according to the memory address of the part or all of the character sub-blocks, initiates a second instruction to the storage system to request the part or all of the character sub-blocks.

In a possible implementation manner, when the target data is located in the memory of the device in the system; the client device can verify the validity of the metadata of the target data according to the global index; when it is determined that the metadata of the target data is valid, A fourth indication is initiated to the storage system according to the metadata of the target data, where the fourth indication is based on RDMA transmission. Afterwards, the processing unit may feed back the target data to the client device under the fourth instruction of the client device.

In a possible implementation manner, when the target data is located in the persistent memory of the device in the system.

The client device may check the validity of the metadata of the target data according to the global index; and initiate a fifth indication to the storage system according to the metadata of the target data when it is determined that the metadata of the target data is valid.

The processing unit may obtain the target data from the persistent storage under the fifth instruction of the client device, and feed back the target data to the client device.

In the seventh aspect, the present application also provides a computer-readable storage medium, where instructions are stored in the computer-readable storage medium, and when it is run on a computer, the computer executes the above-mentioned second aspect and various possible implementations of the second aspect methods in methods.

In an eighth aspect, the present application further provides a computer program product including instructions, which, when run on a computer, cause the computer to execute the above second aspect and the method in each possible implementation manner of the second aspect.

In the ninth aspect, the present application also provides a computer chip, the chip is connected to the memory, and the chip is used to read and execute the software program stored in the memory, and execute the method in the above-mentioned second aspect and each possible implementation manner of the second aspect .

Description of drawings

FIG. 1 is a schematic structural diagram of a system provided by the present application;

Fig. 2 is a schematic structural diagram of an I/O stack provided by the present application;

3A to 3B are schematic diagrams of a global index provided by the present application;

4 to 5 are schematic diagrams of a data access method provided by the present application;

6 to 7 are schematic diagrams of another data access method provided by this application;

FIG. 8 is a schematic structural diagram of a data processing device provided by the present application.

Detailed ways

Before explaining the data processing method provided by this application, first explain the concepts involved in this application:

1. Metadata.

Also known as intermediary data and relay data. Metadata is data describing data (data about data). Metadata can indicate the attributes of data. For example, metadata can record the physical address of data, modification information of data, etc.

2. Remote direct memory access (RDMA).

RDMA is a technology that bypasses the operating system kernel of a remote device (such as a storage device) to access data in its memory. Because it does not go through the operating system, it not only saves a lot of processor resources, but also improves system throughput and reduces system traffic. Network communication delay, especially suitable for wide application in large-scale parallel computer clusters.

RDMA has several major characteristics, (1) data is transmitted between the network and the remote device; (2) without the participation of the operating system kernel, all content related to sending and transmitting is offloaded to the smart network card; (3) virtualized in the user space Direct data transmission between the memory and the iNIC does not involve the operating system kernel, and there is no additional data movement and copying.

3. Unilateral RDMA and bilateral RDMA.

Here, the two ends that need to exchange information are respectively referred to as a client device (client for short) and a server (in this embodiment of the application, the server can be understood as a storage device). The client is deployed on the user side, and the user can initiate a request to the server through the client. The server can be deployed at the remote end. The server generally refers to the storage system, and can be specifically understood as a device in the storage system.

Unilateral RDMA can be divided into RDMA read (READ) and RDMA write (WRITE).

Taking RDMA READ in unilateral RDMA as an example, the client can directly determine the location of the target data in the memory of the server. The message is sent to the server. On the server side, the network card on the server side reads the data on the location information. In the above process, the processor on the server side is not aware of a series of operations on the client side. In other words, the processor on the server side does not know that the client has performed a read operation, thus reducing the consumption of the processor participating in the data transmission process and improving the performance of the system for processing business, with high bandwidth, low latency and low CPU usage. rate features.

In the embodiment of this application, the client device can read the subgroups in the global index from the server through unilateral RDMA, and can also read the data of a certain layer of the target data in the I/O stack from the server through unilateral RDMA metadata.

Bilateral RDMA can be divided into RDMA transmission (SEND) and RDMA reception (RECEIVE).

Taking RDMA RECEIVE in bilateral RDMA as an example, the client does not know where the metadata of the target data is stored in the memory of the server, so the message initiated by the client to request to read data does not carry metadata location information. After the server receives the message, the processor on the server side queries the location information of the metadata and returns it to the client, and the client sends a message to the server again to request to read the data. The text contains the location information of the metadata (that is, the address of the metadata). The network card of the server obtains the metadata according to the location information of the metadata, further obtains the target data, and sends the target data to the client. In this process, the processor on the server side is required to participate, that is to say, bilateral RDMA requires the processor on the server side to process messages from the client, so unilateral RDMA takes less time to read data than bilateral RDMA , lower processor usage and better user experience. Therefore, the application of unilateral RDMA is getting wider and wider.

4. Direct access.

Pass-through access is a way to read and write data from the server-side persistent storage (such as hard disk) without going through the server-side processor. In the direct access method, the client can determine the location of the target data in the server's persistent storage, and the client can communicate with the controller in the server's hard disk through the server's network card, and then read from the server's hard disk. data or write data.

In the embodiment of this application, when the metadata (or index) of the data is stored on the hard disk of the server, the client can read the data of a certain layer of the data in the I/O stack from the server through direct access. metadata. When the data of a certain layer in the I/O stack is stored on the persistent memory of the server, the client can also read the data of the target data in this layer of the I/O stack from the server through direct access.

As shown in FIG. 1, it is a schematic structural diagram of a data access system provided by an embodiment of the present application. The system includes a client device 200 and a storage system 100. The storage system 100 includes multiple storage devices. In FIG. 1 only A storage device 110 of the storage system is exemplarily shown.

Users access data through applications. The computers running these applications are referred to as "client devices 200". The client device 200 may be a physical machine or a virtual machine. Physical client devices 200 include, but are not limited to, desktop computers, servers, laptop computers, and mobile devices.

A user may initiate a data access request, such as a data write request or a data read request, to the storage device 110 in the storage system 100 through the client device 200 . The storage device 110 receives the data access request and processes the data access request.

Here, the storage device 110 processes the data access request and executes the data access method provided in the embodiment of the present application as an example for description. For example, the data access request is a data write request, which is used to request to write target data in the storage system 100 . The data writing request includes the target data and the logical address of the target data. After receiving the data write request, the storage device 110 may first write the target data to the position indicated by the logical address of the target data according to the data write request, and the position may be located at a position in the I/O stack. layer; and update the global index, and the updated global index can indicate the layer where the target data is located in the I/O stack. The I/O stack is a layered structure formed by dividing the storage media in the storage system into layers. Each layer in the I/O stack includes one or more storage media that can be used to store data. The global index can associate the logical address of the data with the layer where the data resides, so as to indicate the layer where the data resides in the I/O stack. For the description of the I/O stack and the global index, please refer to the relevant description below.

For another example, the data access request is a data read request, which is used to request to read target data from the storage system 100, and the data read request carries a logical address of the target data. After receiving the data read request, the storage device 110 can determine the layer where the target data is located in the I/O stack according to the global index and the logical address of the target data, and then read the target data from the layer where the target data is located. data.

Specifically, referring to FIG. 1 for the structure of the storage device 110 , the storage device 110 includes a bus 111 , a processor 112 , a memory 113 , a network card 114 and a hard disk 115 . Memory 113 may be located in processor 112 .

It should be noted that, in the embodiment of the present application, a hard disk is used as an example of a persistent memory of the storage device for illustration, but the embodiment of the present application is also applicable to mechanical hard disks or other types of hard disks.

Processor 112 can be central processing unit (central processing unit, CPU), and this processor 112 can also be other general processors, digital signal processor (digital signal processor, DSP), application specific integrated circuit (application specific integrated circuit, ASIC) ), field programmable gate array (field programmable gate array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, artificial intelligence chips, chip-on-chip, etc.

The memory 113 may include a volatile memory (volatile memory), such as a random access memory (random access memory, RAM), a dynamic random access memory (dynamic random access memory, DRAM), and the like. It can also be a non-volatile memory (non-volatile memory), such as a storage-class memory (storage-class memory, SCM), etc., or a combination of a volatile memory and a non-volatile memory. The memory 113 can be divided from a functional point of view, and the memory 113 can be divided into a write cache, a read cache, and the like. The write cache refers to a cache that can provide high-efficiency write capabilities, and the read cache refers to a cache that can store data with a high read frequency.

The storage device 110 may also include one or more hard disks 115 . The hard disk 115 can be used to permanently store data. Specifically, inside the hard disk 115, the hard disk 115 may also include a hard disk cache and a persistent storage medium.

Inside the storage device 110, the data access method provided by the embodiment of the present application may be executed by the processor 112, that is, the processor 112 may execute the data access method provided by the embodiment of the present application by invoking a computer to execute instructions. The data access method provided in the embodiment of the present application can also be executed by the network card 114 , for example, the network card 114 can execute the data processing method provided in the embodiment of the present application by calling the computer-executed instructions stored in the memory 113 . For another example, the network card 114 may also invoke computer-executed instructions stored inside the network card 114 to execute the data access method provided in the embodiment of the present application. For another example, in some possible scenarios, computer storage instructions may also be programmed on the network card 114, and the network card 114 may execute the data access method provided in the embodiment of the present application.

The embodiment of the present application does not limit the type of the storage system. In practice, the storage system in FIG. 1 can be either a centralized storage system or a distributed storage system.

For each storage medium in the storage device in the storage system (such as read cache, write cache, hard disk, etc.), the concept of storage medium layering is introduced. Divide the storage medium into multiple layers from top to bottom (from high to low). The embodiment of the present application does not limit the division standard used when the storage medium is layered. For example, the storage medium can be divided only according to the type of the storage medium, and the storage medium of the same type is divided into one layer to form a top-down ( from high to low) in multiple layers. For another example, the storage medium may only be divided according to the function of the storage medium to form multiple layers from top to bottom (from high to low). For another example, the storage medium may be divided into multiple layers from top to bottom (from high to low) by comprehensively considering the function and type of the storage medium.

As shown in FIG. 2 , in the embodiment of the present application, by layering storage media in the storage system, an input/output (input/out, I/O) stack of the storage system can be formed. In the I/O stack, it is divided into a performance layer and a capacity layer from top to bottom in a coarse-grained manner. The read and write performance of the performance tier is better than that of the capacity tier, but the capacity of the performance tier is relatively small.

Specific to the inside of the performance layer, continue to refine, from top to bottom can be divided into write cache (write cache, Wcache), read cache (read cache, Rcache), hard disk cache (smart cache). Write cache (write cache, Wcache), read cache (read cache, Rcache), and hard disk cache are respectively a layer in the I/O stack.

It should be noted that the inclusion of various caches in the performance layer is only an example. The standards for storage medium tiering are different, and the granular tiering in the performance tier may also be different.

Specific to the inside of the capacity layer, continue to refine, from top to bottom, it can be divided into high-performance layer and ordinary performance layer. Among them, the high-performance layer can include high-performance hard disks, such as solid-state drives (SSD). The general performance tier may include a hard disk with general performance, such as a mechanical hard disk (HDD).

It should be noted that the inclusion of each layer in the capacity layer is only an example. The types of hard disks included in different storage systems 100 may also be different. For example, some storage systems 100 may only include one type of hard disk, and in this case, the capacity layer may not be further refined.

The arrangement of the layers in the I/O stack describes the flow of data in the storage system 100 , and in the storage system 100 data usually flows from top to bottom in the layers of the I/O stack. Taking the I/O stack shown in FIG. 2 as an example, when data A is written into the storage system 100, it will be written into the performance layer first. Inside the performance layer, it will be written to the write cache first. When the data in the write cache exceeds the threshold W, the data in the write cache will be migrated to the read cache. In this way, free storage space will be formed in the write cache to store the latest write The data. When the data in the read cache exceeds a certain threshold R, the data in the read cache will continue to migrate downwards to the hard disk cache. If the data in the hard disk cache exceeds the threshold S, the data in the hard disk cache will flow to the capacity layer. The data in the hard disk cache will be migrated to the high-performance tier in the capacity tier first. When the data in the high-performance tier reaches the threshold H, the data in the high-performance tier will be migrated to the normal performance tier.

Specific to each layer of the I/O stack, each layer of the I/O stack may include one or more storage media, which may be used to store data. In order to quickly find data from a layer of the I/O stack, the data stored in the layer of the I/O stack may be indexed. That is, when data is written into this layer, an index can be created for the data. The index can indicate the correspondence between the logical address of the data and the metadata of the data. According to the logical address of the data, the metadata of the data can be determined, and then the physical address of the data can be determined.

The embodiment of the present application does not limit the method of indexing data in each layer. For example, the data can be indexed through a hash table, that is, the index of the data is stored in the layer in the form of a hash table. The hash table The corresponding relationship between the logical address of the data and the metadata is recorded in . For another example, data can also be indexed in the form of a B+ tree or a linked list.

Based on the I/O stack shown in Figure 2, when the storage system 100 is processing the data read request from the client device 200, the storage media in each layer are sequentially arranged in the order of the I/O stack from top to bottom Find the data requested by the data read request until the data requested by the data read request is found. After the requested data is found, the data is fed back to the client device 200 .

This method of searching for data in each layer in the top-to-bottom order of the I/O stack is a relatively common method. When the data to be searched is located in the lower layer, the time-consuming to find the data is greater. This will reduce the processing efficiency of data read requests.

In order to effectively improve the processing efficiency of data read requests, in the embodiment of the present application, a global index (global mask) is set in the storage system 100, and the global index can indicate where the data on the logical address is located in the I/O stack. layer. After the storage device 110 receives the data read request, it can determine the layer where the target data in the I/O stack is located according to the logical address of the requested read target data carried in the data read request and the global index. Afterwards, the target data is read from the layer according to the logical address. In this way, in the process of processing the entire data read request, it is not necessary to search for data in each layer according to the top-down order of the I/O stack, and the global index can be used to directly determine the location of the requested data. layer, which can greatly reduce the delay of data search and improve the processing efficiency of data read requests.

Obviously, the global index is the key to simplifying the data reading process. The composition of the global index will be described below. In the embodiment of the present application, the global index can exist in any one or both of the following forms.

Before introducing the global index, let's briefly explain the division of logical storage space to better understand the existence form of the global index. In the embodiment of the present application, the physical storage space composed of various storage media in the storage system 100 can be mapped to a logical storage space, the logical storage space is divided according to a set size, and the logical storage space is divided into multiple logical blocks. Each logical block can be the same size. For example, each logical block may be equal to 256 kilobytes (Kibibyte, KB), and for another example, each logical block may be equal to 1 megabyte (megabyte, MB). Each logical block can be divided into multiple logical sub-blocks according to the minimum reading and writing unit of the I/O stack. The size of each logical sub-block may be equal to the minimum read/write unit of the I/O stack.

The minimum reading and writing unit of the I/O stack refers to the minimum amount of data written to the I/O stack at one time during the data writing process, or the minimum amount of data read from the I/O stack at one time during the data reading process . Usually the minimum amount of data written at one time is the same as the minimum amount of data read. For example, when writing a 256KB data to the storage system 100, if the minimum read/write unit of the I/O stack is 8KB, then when writing the 256KB data, it will be divided into multiple times (32 times). Write 8KB of data until all the 256KB of data is written. Similarly, when reading a 256K data from the storage system 100, if the minimum reading and writing unit of the I/O stack is 8KB, then when reading the 256KB data, it will be divided into multiple times (32 times). Take the data of 8KB size, and read all the data up to 256KB.

A character block for pointing to a logical block is set in the global index, and the character block includes a character sub-block pointing to a logical sub-block. For any logical block, the specific value of the character block pointing to the logical block can indicate whether data is stored in the logical block, and which layer of the I/O stack the stored data is stored in. Similarly, for any logical sub-block, the character sub-block pointing to the logical sub-block can indicate whether data is stored in the logical sub-block, and which layer of the I/O stack the stored data is stored in.

The global index includes multiple character blocks, each character block is used to point to at least one logical block, each character block includes multiple subgroups, and each subgroup points to a logical subblock in the at least one logical block. Each subgroup includes multiple character sub-blocks, and one character sub-block corresponds to one layer in the I/O stack. The specific value of the character sub-block can indicate whether data is stored in the logical sub-block, and whether the stored data is stored in the layer corresponding to the character sub-block in the I/O stack.

Different existence forms of the global index have different sizes of character blocks and character sub-blocks. For the bitmap, the smallest character sub-block is one bit, each sub-group includes multiple bits corresponding to different layers in the I/O stack, and the character block includes multiple sub-groups pointing to different logical sub-blocks. A character block is a group of bits. For the counter group, the smallest character sub-block is a counter, and usually the counter needs to be represented by multiple bits. Each subgroup includes multiple counters corresponding to different layers in the I/O stack, and the character block includes multiple subgroups pointing to different logical subblocks. A character block is a set of counters.

Form 1, bitmap (bitmap).

As shown in Figure 3A, the schematic diagram of the bitmap includes multiple groups of bits in the bitmap, and each group of bits can point to at least one logical block, and each group of bits includes multiple subgroups (grain), each subgroup The group includes multiple bits, and each subgroup points to a logical subblock in the logical block. Different subgroups in a group of bits point to different logical subblocks. The number of bits in each subgroup can be equal to the layer in the I/O stack can also be equal to the total number of layers in the I/O stack minus one.

A bit in a subgroup corresponds to a layer in the I/O stack, and different bits correspond to different layers in the I/O stack. For example, for a bit in the subgroup, when the value of the bit is 1, it may indicate that the data in the logical sub-block is located at this layer, and when the value of the bit is 0, it may indicate that the logical sub-block The data in is not in this layer. For example, multiple bits in a subgroup correspond to a layer in the I/O stack in the top-down order of the I/O stack, that is, the first bit corresponds to the first layer in the I/O stack (such as corresponding to the write cache), the second bit corresponds to the second layer in the I/O stack (such as corresponding to the read cache), and the third bit corresponds to the third layer in the I/O stack (such as corresponding to the hard disk cache). The fourth bit corresponds to the fourth layer in the I/O stack (eg, corresponds to the capacity layer).

When the value of the first bit is 1, it can indicate that the data on the logical sub-block pointed to by the subgroup is in the write cache; when the value of the first bit is 0, it can indicate that the data on the logical subblock pointed to by the subgroup is The data on the logical sub-block is not in the write cache. Similarly, when the value of the second bit is 1, it can indicate that the data on the logical sub-block pointed to by the subgroup is in the read cache; when the value of the second bit is 0, it can indicate that the subgroup The data on the logical sub-block pointed to is not in the read cache. Similarly, when the value of the third bit is 1, it can indicate that the data on the logical sub-block pointed to by the subgroup is located in the hard disk cache; when the value of the third bit is 0, it can indicate that the subgroup The data on the logical sub-block pointed to is not in the hard disk cache. Similarly, when the value of the fourth bit is 1 (the fourth bit is not shown in Figure 2), it may indicate that the data on the logical sub-block pointed to by the subgroup is located at the capacity layer, when the fourth bit When the value of is 0, it may indicate that the data on the logical sub-block pointed to by the sub-group is not in the capacity layer. In some scenarios, the bit corresponding to the last layer may not be set, for example, the fourth bit may be omitted.

The value of each bit in each subgroup on the bitmap changes as data is written, flowed, and eliminated.

Data flow refers to the inflow or outflow of data in each layer of the I/O stack, such as migrating data from the upper layer of the I/O stack to the next layer. Another example is that data flows out of one layer in the I/O stack, and the outflowing data flows into another layer in the I/O stack. Data flow mainly occurs in the disk flushing process for the write cache, garbage collection at the capacity layer, data loading when executing a read data request, data prefetch process (such as writing data with high read and write frequency to the read cache) or During data migration in the dynamic storage tiering feature.

Data writing refers to writing data to a certain layer of the I/O stack.

Data elimination refers to the deletion of data at a certain layer in the I/O stack. For example, when data in the read cache is eliminated, or when garbage collection is performed after the capacity layer, the overwritten data in the capacity layer needs to be deleted.

Taking the first bit in a subgroup as an example, the first bit corresponds to the first layer in the I/O stack, and the value of the first bit is 1, indicating that the logical subblock pointed to by the subgroup is The data is located at the first level in the I/O stack. When the amount of data in the first layer in the I/O stack reaches the threshold W, the data in the first layer can be migrated to the second layer. At this time, the value of the first bit will be reduced by one and become 0. The value 0 indicates The data in the logical subblock pointed to by this subgroup is not in the first layer. When data is written later, the written data is written into the logical sub-block pointed to by the subgroup, then, when the data is written, the value of the first bit will become 1. If data is written later, the written data is also written into the logical sub-block pointed to by the sub-group, and the data written in the logical sub-block before is overwritten. Then, when the data is written, the first A bit will still hold a value of 1. Taking the value of the first bit at this time as 1 as an example, when data is eliminated later, if the data written in the logical sub-block pointed to by this sub-group is found to be inactive data, the sub-block will be deleted. After the data in the logical sub-block pointed to by the group is deleted, the value of the first bit becomes 0. If two data writes occur in the logical sub-block pointed to by the sub-group, and the data written for the first time in the logical sub-block pointed to by the sub-group is the data that needs to be overwritten, delete the data pointed to by the sub-group The data written for the first time in the logical sub-block of , retains the data written last time, and the value of the first bit remains 1.

It can be seen that the value of a bit in each subgroup on the bitmap can only indicate whether the data on the logical subblock pointed to by the subgroup is located in this layer, but cannot specifically describe the data on the logical subblock pointed to by the subgroup. The data of is the data written for the first time.

It should be noted that, in practical applications, a subgroup may not set the bit corresponding to the last layer in the I/O stack. In this case, each bit in the subgroup corresponds to the other layers in the I/O stack except the last layer. When all the bits in the subgroup are 0, it means that the subgroup points to The data in the logical sub-block of the group is not located in the rest of the I/O stack except the last layer. It can be further explained that the data in the logical sub-block pointed to by the sub-group can only be located in the I/O stack in the last layer.

Because the logical block is a larger space in the logical storage space, and the indicated space of the logical address of the data may be a part of the logical block, the indicated space of the logical address may be a part of the logical sub-block ( For the convenience of expression, this part of the logical sub-block is referred to as the logical sub-block indicated by the logical address). For this reason, when searching for data, it is necessary to first determine the logical block to which the space indicated by the logical address belongs (for the convenience of expression, the logical block to which the space indicated by the logical address belongs can also be referred to as the logical block to which the logical address belongs), One or more groups of bits in the bitmap that point to the logical block are determined. Afterwards, the logical sub-block indicated by the logical address is determined from the logical block, and one or more subgroups pointing to the logical sub-block in the one or more groups of bits in the bitmap are determined.

In order to be able to determine one or more groups of bits pointing to the logical block to which the logical address belongs in the bitmap according to the logical address, a hash table can also be set in the storage system 100, and the logical address and the bit position are recorded in the hash table. The corresponding relationship of each group of bits in the figure. One or more groups of bits pointing to the logical block to which the logical address belongs can be determined through the logical address of the data and the hash table. Alternatively, a hash function can also be set in the storage system 100, and the logical address is used as the input of the hash function, and the obtained hash value can indicate one or more groups of bits in the bitmap, one or more groups of bits Points to the logical block to which this logical address belongs.

When determining one or more groups of bits in the bitmap, the logical address may be rounded according to the size of the logical block. The number obtained after rounding is used to query the hash table to determine one or more groups of bits pointing to the logical block to which the logical address belongs. Or apply a hash function to the number obtained after rounding to determine one or more groups of bits pointing to the logical block to which the logical address belongs.

After determining one or more groups of bits in the bitmap, one or more subgroups can be further determined from the one or more groups of bits, and the logical subblocks pointed to by the one or more subgroups are the logical The logical subblock indicated by the address. When determining the one or more subgroups, the logical subblock indicated by the logical address may be determined according to the offset of the logical address in the logical block and the data length of the target data.

The offset of the logical address in the logical block can be determined by the difference between the start address of the logical block and the logical address. Taking the size of each logical block as 256KB, the logical block includes 32 logical sub-blocks, the size of each logical sub-block is 8KB, and each group of bits includes 32 sub-groups as an example. If the location indicated by the logical block address (LBA) of the data is 1MB+520KB, the data length is 256KB. You can first round 1MB+520KB according to the size of the logical block (256K) to obtain 1MB+512KB. By applying a hash function to 1MB+512KB, two groups of bits pointing to the logical block to which the logical address belongs can be determined. 1MB+512KB is the starting address of the logical block, and the offset of the logical address in the logical block is the difference between 1MB+512KB and 1MB+520KB, that is, the offset is 8KB. The location indicated by the logical address is the location where the logical block is offset by 8KB. And because the size of each logical sub-block is 8KB, the logical sub-block indicated by the logical address is the logical sub-block offset by one logical sub-block after the starting position of the logical block, that is, the second logical sub-block . Because the data length is 256KB, the logical sub-block pointed to by this logical address is the first logical sub-block from the second logical sub-block to the 32nd logical sub-block and the next logical sub-block in this logical block, a total of 32 logical sub-blocks. The subgroups pointing to the 32 logical subblocks on the bitmap are the second subgroup to the 32nd subgroup subblock in a group of bits pointing to the logical block and the first subgroup in the next group of bits.

After finding each subgroup on the bitmap that points to the logical subblock indicated by the logical address, you can continue to determine the data on the logical subblock pointed to by the subgroup according to the value of each bit in each subgroup The layer in the I/O stack where it resides.

For each subgroup, after determining the layer of the data on the logical subblock pointed to by the subgroup in the I/O stack, the data on the logical subblock pointed to by the subgroup can be searched from the layer. Specifically for this layer, the data index in this layer can be searched according to the logical address of the data, the metadata of the data can be determined, and then the data can be read from the position indicated by the metadata.

For example, for the second subgroup in the first group of bits, if the subgroup includes three bits, they correspond to the first layer (write cache) and the second layer (read cache) in the I/O stack respectively. cache) and the third layer (hard disk cache). If the value of the three bits in the subgroup is 100, it means that the data on the logical subblock pointed to by the subgroup is in the write cache, and the data can be read from the write cache. If the value of the three bits in the subgroup is 010, it means that the data on the logical subblock pointed to by the subgroup is in the read cache, and the data can be read from the read cache. If the value of the three bits in the subgroup is 000, it means that the data on the logical subblock pointed to by the subgroup is not located in the first three layers of the I/O stack, but in the fourth layer capacity tier, the data can be read from the capacity tier. If the value of the three bits in the subgroup is 110, it means that the data on the logical subblock pointed to by the subgroup is stored in the first two layers of the I/O stack. When data is written into the storage system 100, it is preferentially written to the first layer of the I/O stack, and the latest data written on the logical sub-block is on the first layer of the I/O stack.

Form 2, the counter (counter) group.

As shown in Figure 3B, the schematic diagram of the counter group includes multiple groups (group) counters in the counter group, and each group of counters can point to at least one logic block, and each group of counters includes a plurality of subgroups (grain), in each subgroup Including multiple counters, each subgroup points to a logical subblock in the logical block, different subgroups in a set of counters point to different logical subblocks, and the number of counters in each subgroup can be equal to the total number of layers in the I/O stack , which can also be equal to the total number of layers in the I/O stack minus one.

A counter in a subgroup corresponds to a layer in the I/O stack, and different counters correspond to different layers in the I/O stack. For example, multiple counters in a subgroup correspond to a layer in the I/O stack in the top-down order of the I/O stack, that is, the first counter corresponds to the first layer in the I/O stack (such as corresponding to the write cache), the second counter corresponds to the second layer in the I/O stack (such as corresponding to the read cache), and the third counter corresponds to the third layer in the I/O stack (such as corresponding to the hard disk cache). The fourth counter corresponds to the fourth layer in the I/O stack (eg, corresponds to the capacity layer).

For a counter in this subgroup, when the value of the counter is null or 0, it can indicate that the data in the logical sub-block is not in this layer; when the value of the counter is non-null or a non-zero integer, it can indicate The data in the logical sub-block is located at this layer, and the specific value on the counter can represent the number of times the data in the logical sub-block is updated. For example, when the value of the first counter is not 0, it may indicate that the data on the logical sub-block pointed to by the subgroup is in the write cache; when the value of the first counter is 0 or null, it may indicate that the The data on the logical subblock pointed to by the subgroup is not in the write cache. Similarly, when the value of the second counter is non-zero, it can indicate that the data on the logical sub-block pointed to by the subgroup is in the read cache; when the value of the second counter is 0 or null, it can indicate The data on the logical subblock pointed to by this subgroup is not in the read cache. Similarly, when the value of the third counter is non-zero, it can indicate that the data on the logical sub-block pointed to by the subgroup is located in the hard disk cache; when the value of the third counter is 0 or null, it can indicate The data on the logical subblock pointed to by this subgroup is not cached on the hard disk. Similarly, when the value of the fourth counter is not 0, it can indicate that the data on the logical sub-block pointed to by the subgroup is located in the capacity layer; when the value of the fourth counter is 0 or null, it can indicate The data on the logical subblock pointed to by this subgroup is not in the capacity layer.

Because there will be data flow in the I/O stack, such as migrating data from the upper layer in the I/O stack to the next layer, or data flowing out from a layer in the I/O stack, the outflowing data will flow into the Another layer in the I/O stack. In the I/O stack, data may be written to the same logical address multiple times. For example, in the I/O stack, data is allowed to be written to the same logical address multiple times. The last written data will overwrite the previously written data. The data. There is also data elimination in the I/O stack, such as deleting inactive data in a certain layer in the I/O stack, or deleting overwritten data in a certain layer in the I/O stack.

The value of the counter can be used to record the data flow, data writing and data elimination of the I/O stack on the layer. Taking the first counter in a subgroup as an example, the first counter can correspond to the first layer in the I/O stack, and the value of the first counter is 1, indicating the logical subblock pointed to by the subgroup The data in is at the first level in the I/O stack. When the amount of data in the first layer in the I/O stack reaches the threshold W, the data in the first layer can be migrated to the second layer. At this time, the value of the first counter will be reduced by one, indicating that the subgroup points to The data in the logical sub-block is not in the first layer. Still taking the first counter in a subgroup as an example, the current value of the counter is 1, and then two data writing processes are performed, and the written data is written into the logical subblock pointed to by the subgroup , then, when the data is written for the first time, the value of the first counter will increase by one and become 2. When the data is written for the second time, the value of the first counter will be increased by one more to become 3. The value of the first counter changes to 2 or 3, which may indicate that data has been written 2 times or 3 times successively in the logical subblock pointed to by the subgroup. Taking the value of the first counter at this time as 3 as an example, when the subsequent data elimination is performed, if it is found that the data written in the previous two times in the logical sub-block pointed to by the sub-group is the data that needs to be overwritten, The data written in the previous two times in the logical sub-block pointed to by the sub-group will be deleted. After deleting the data written in the previous two times in the logical sub-block pointed to by the sub-group, the value of the first counter becomes 1.

It should be noted that, in practical applications, a subgroup may not set the counter corresponding to the last layer in the I/O stack. In this case, each counter in the subgroup corresponds to other layers in the I/O stack except the last layer. When all the counters in the subgroup are 0 or null, it means that the subgroup The data in the logical sub-block pointed to by the group is not located in the rest of the I/O stack except the last layer. It can be further explained that the data in the logical sub-block pointed to by the sub-group can only be located in the I/O stack. /O last layer in the stack.

Using a logical address to identify one or more groups of counters, and finding one or more subgroups from the group or group of counters, is the same as a method of using a logical address to determine one or more groups of bits, and finding one or more subgroups from the group The method of finding one or more subgroups in or group bits is similar, for details, please refer to the foregoing description, which will not be repeated here.

It can be seen from the above description that the information that can be described by the counter group and the bitmap is basically the same, and the information that the counter group can describe is more abundant. The value of each counter in the counter group can determine the data flow in the layer corresponding to the counter, Writing and elimination. Here are two simple examples to illustrate the advantages of counter groups:

Example 1: When determining one or more character blocks (such as one or more sets of bits, one or more sets of counters) of the logical block to which the logical address belongs in the global index, use hash calculation to map the logical address to the character piece. Different logical addresses, after hash calculation, may have the same hash value, resulting in a hash collision, which will map two different logical addresses to the same character block. In this case, the value of the character block mapped to (that is, the value of the character sub-block in each subgroup) actually needs to represent where the data in the logical block to which the two logical addresses belong is located in the I/O stack. layer.

For ease of understanding, here is an example where the two logical addresses are LBA1 and LBA2, and the character block to which the two logical addresses in the global index are mapped is character block A. Here, only the characters in each subgroup in the character block A are The value of the first character sub-block a will be described, and the value of other character sub-blocks a is similar to the value of the first character sub-block.

When data is written to the logical address LBA1, the data will be written to the first layer of the I/O stack first. If the character sub-block a is a bit, the value of this bit will become 1. If the character sub-block is a counter, the value of the counter will also become 1. Afterwards, if data is written to the logical address LBA2, the data will also be preferentially written to the first layer of the I/O stack. If the character sub-block a is a bit, the value of this bit remains 1. If the character sub-block a is a counter, the value of the counter will increase from 1 to 2. It can be seen from this that the value of the counter can clearly record the number of times data is written to the first layer of the I/O stack. In this case, the global index exists in the form of a bitmap or a counter group, both of which can accurately indicate that the data in the logical block to which the logical address LBA1 or LBA2 belongs is located at the first layer of the I/O stack.

However, later, if the data in the logic block to which the logical address LBA1 belongs has a data flow, it flows from the first layer of the I/O stack to the second layer of the I/O stack. If the character sub-block a is a bit, the bit The value of will change from 1 to 0, and the next bit of this bit (that is, the bit corresponding to the second layer of the I/O stack) will change from 0 to 1. If the character sub-block a is a counter, the value of the counter will change from 2 to 1, and the next counter of the counter (that is, the counter corresponding to the second layer of the I/O stack) will change from 0 to 1.

When it is necessary to query the data in the logical address LBA2, if the global index exists in the form of a bitmap, since the value of the bit corresponding to the first layer of the I/O stack is 0, the value of the next bit is 1 , by querying the global index, it is determined that the data in the logical address LBA2 is located in the second layer of the I/O stack. In fact, the data in the logical address LAB2 has not flowed, and is still located on the first layer of the I/O stack. This will easily cause problems in subsequent data reading and cannot be read from the second layer of the I/O stack. Data in LBA2. If the global index exists in the form of a counter group, since the value of the counter corresponding to the first layer of the I/O stack is 1, the data in the logical address LBA2 is determined to be located at the first layer of the I/O stack by querying the global index. It is consistent with the layer where the data in the logical address LAB2 in the I/O stack is located, and the data in the logical address LBA2 can be accurately read subsequently.

When it is necessary to query the data in the logical address LBA1, if the global index exists in the form of a bitmap, since the value of the bit corresponding to the first layer of the I/O stack is 0, the value of the next bit is 1 , it is determined by querying the global index that the data in the logical address LBA1 is located in the second layer of the I/O stack. It is consistent with the layer where the data in the logical address LAB1 in the I/O stack is located, and the data can be read accurately. If the global index exists in the form of a counter group, since the value of the counter corresponding to the first layer of the I/O stack is 1, it is determined that the data in the logical address LBA1 is located at the first layer of the I/O stack by querying the global index. Although it is inconsistent with the layer where the data in the logical address LAB1 in the I/O stack is located, in the process of subsequent data reading, although the data in the logical address LBA1 cannot be queried on the first layer of the I/O stack, but later , can be traversed according to the order of the layers of the I/O stack, and the data in the logical address LBA1 can be queried at the second layer, and the data in the logical address LBA1 can still be read accurately.

It can be seen from this that the value of the counter can not only clearly record the flow status of data in the first layer of the I/O stack, but also solve the problem of hash conflicts to a certain extent and ensure the accuracy of data reading.

Example 2: Still taking the character sub-block a and LBA1 as an example, in the scenario of additional writing, when data is written to the logical address LBA1 for the first time, the data will be written to the first layer of the I/O stack first, if The character sub-block a is one bit, and the value of this bit is 1. If the character sub-block a is a counter, the value of the counter is also 1. Afterwards, if data is written to the logical address LBA1 again to cover the previously written data, the data will also be written to the first layer of the I/O stack first. If the character sub-block a is one bit, the bit The value remains 1. If the character sub-block a is a counter, the value of the counter will increase from 1 to 2. During data elimination, the data written for the first time will be deleted from the storage medium, and the global index can be updated during data elimination. If the character sub-block a is a bit, the value of this bit will change from 1 to 0. If the character sub-block a is a counter, the value of the counter will change from 2 to 1. In fact, the layer where the data in the logical address LBA1 is located is still the first layer. When the character sub-block a is a bit, there may be errors in the value. It can be seen that the global index in the form of a counter group can accurately Record the layer where the data in the logical address LBA1 in the I/O stack is located.

However, compared with the bitmap, one counter in the counter group will occupy multiple bits, and the multiple bits are used to represent different values, which also makes the space occupied by the counter group larger than that of the bitmap.

In this embodiment, the global index may be realized by means of a counter group, or may be realized by a bitmap, or both methods may be applicable. The following describes the data access method provided by the embodiment of the present application with reference to the accompanying drawings. In different scenarios, the execution subject of the data access method provided by the embodiment of the present application will be different. For example, the data access method may be executed by the processor 112 of the storage device 110 in the storage system 100 shown in FIG. 1 , or may be executed by the network card 114 of the storage device 110 in the storage system 100 . The two possible cases are described below:

Scenario 1: The processor 112 of the storage device 110 in the storage system 100 executes the data access method provided in the embodiment of the present application.

As shown in Figure 4, a data access method provided by this application, the method includes:

Step 401: The processor 112 receives a data write request, the data write request is used to request to write target data, and the data write request carries target data and a logical address of the target data. The logical address may include a start logical address and a data length (length). The starting logical address of the data can be represented by a logical block address (logic block address, LBA) and a logical unit number (logical unit number, LUN).

The data writing request may be directly sent by the client device 200 to the storage device 110 . The data writing request may also be sent to the storage device 110 by other storage devices 110 in the storage system 100. For example, there is a device for managing the storage device 110 in the storage system 100, and the device can allocate the storage device 110 for data, It is also possible to instruct the storage device 110 to write data into the storage device 110 . When the device determines that target data needs to be written in the storage device 110 , it may send a data writing request to the storage device 110 .

Step 402: The processor 112 of the storage device 110 writes the target data to the location indicated by the logical address according to the data writing request.

When the processor 112 writes the target data to the location indicated by the logical address, it may preferentially write the target data to the first layer in the I/O stack, for example, it may preferentially write the target data into the cache.

The processor 112 creates an index for the target data when writing to the first layer of the I/O stack, and the index of the target data can indicate the correspondence between the logical address of the target data and the metadata of the target data. The metadata of the object data can indicate the physical address of the object data in the layer.

Step 403: the processor 112 updates the global index, and after the update, the global index can indicate the layer where the target data is located in the I/O stack.

The processor 112 may determine, according to the logical address of the target data, a character block in the global index pointing to the logical block of the logical address and a subgroup indicating a logical sub-block in the logical block. For example, the processor 112 may query the character block corresponding to the logical address of the target data and the subgroup in the character block in the hash table. Because the processor 112 will preferentially store the target data in the first layer in the I/O stack, the processor 112 can set the character corresponding to the first layer in the I/O stack in the subgroup of the logical sub-block in the logical block The specific value of the sub-block is such that the set value of the character sub-block can indicate that data is stored in the first layer of the I/O stack in the sub-group.

When the global index exists in the form of a bitmap, when the processor 112 updates the global index, the processor 112 can determine one or more groups of bits indicating the logical block to which the logical address belongs according to the logical address of the target data, and then , and then determine the logical sub-block indicated by the logical address according to the offset of the logical address in the logical block and the data length of the target data, and then determine each subgroup pointing to the logical sub-block in the bitmap. Since the processor 112 preferentially stores the target data in the first layer in the I/O stack, the processor 112 can point to the first bit in each subgroup of the logical subblock (that is, the corresponding I/O stack The value of the bit of the first layer) is set to 1 to indicate that the target data is stored in the first layer of the I/O stack.

When the global index exists in the form of a counter group, when the processor 112 updates the global index, the processor 112 can determine one or more groups of counters indicating the logical block to which the logical address belongs according to the logical address of the target data, and then, Then determine the logical sub-block indicated by the logical address according to the offset of the logical address in the logical block and the data length of the target data, and then determine each subgroup pointing to the logical sub-block in the bitmap. Since the processor 112 preferentially stores the target data in the first layer in the I/O stack, the processor 112 can point to the first counter in each subgroup of the logical subblock (that is, the corresponding I/O stack The value of the counter of the first layer) is increased by 1 to indicate that the target data is written in the first layer of the I/O stack.

The embodiment of the present application does not limit the order in which step 402 and step 403 are performed, that is, the processor 112 may first write the target data. After writing the target data, processor 112 updates the global index. The processor 112 may also update the global index first, and then write the target data. Whether it is the writing of target data or the updating of the global index, the processor 112 needs to realize the logical address of the target data. The writing of the target data and the updating of the global index are two relatively independent processes, and there is no Therefore, in the embodiment of the present application, the sequence of execution of step 402 and step 403 is not particularly emphasized.

In the embodiment of the present application, the processor 112 may update the global index first (step 403 is executed first), and then writes the target data into the storage location indicated by the logical address (step 402 is executed again). The processor 112 first updates the global index to update the layer of the target data in the I/O stack to the global index in advance, so that if the processor 112 receives the data for requesting the target data within a short time after receiving the data write request For the read request, because the update of the global index is performed before, the processor 112 can accurately determine the layer where the target data in the I/O stack is located according to the updated global index.

Step 404: The processor 112 feeds back a data writing response, indicating that the target data has been successfully written.

Steps 401 to 404 are the data writing process. In addition to the data writing process, the processor 112 may also execute other data processing processes, such as data flow, data elimination, and the like.

For example, when it is necessary to migrate the data stored in the storage medium of a certain layer in the I/O stack to the storage medium of the next layer (that is, to refresh the storage medium of a certain layer in the I/O stack) , the processor 112 may migrate the data in the storage medium of the layer to the storage medium of the next layer, and after the data migration is completed, the processor 112 may update the global index.

When the global index exists in the form of a bitmap, when the processor 112 updates the global index, the processor 112 can set the bit in the global index corresponding to the layer in the I/O stack to 0 to represent the layer If there is no data in the storage medium of the I/O stack, the bit corresponding to the lower layer of the I/O stack in the global index is set to 1 to indicate that data is stored in the storage medium of the lower layer of the layer.

When the global index exists in the form of a counter group, when the processor 112 updates the global index, the processor 112 may decrement the value of the counter in the global index corresponding to the layer in the I/O stack by 1 to represent the The data in the storage medium of the layer has been migrated once, and the value of the counter corresponding to the next layer of the layer in the I/O stack in the global index is increased by 1 to represent the storage medium of the next layer of the layer The new data has been moved into .

For another example, the processor 112 may also migrate data with a higher reading frequency in a certain layer of the I/O stack to a higher layer in the I/O stack, such as the second layer where the read cache is located. The processor 112 may control the outflow of data with a higher reading frequency in the storage medium in this layer, and control the inflow of the data into the storage medium in the second layer. Processor 112 may also update the global index. The operation of the processor 112 to update the global index may be performed after the data flows out and before the data flows into the second layer.

When the global index exists in the form of a bitmap, when the processor 112 updates the global index, the processor 112 can associate the subgroup pointed to by the logical address of the data in the global index with the layer in the I/O stack The corresponding bit is set to 0 to indicate that the data does not exist in the storage medium of this layer, and the bit corresponding to the second layer in the I/O stack in the subgroup pointed to by the logical address of the data in the global index is set to 1, indicating that the data is stored in the storage medium of the second layer in the I/O stack.

When the global index exists in the form of a counter group, when the processor 112 updates the global index, the processor 112 can point to the subgroup corresponding to the layer in the I/O stack in the subgroup pointed to by the logical address of the data in the global index. The value of the counter is decremented by 1 to indicate that the data in the storage medium of this layer has been outflowed once, and the value of the counter corresponding to the first layer in the I/O stack in the subgroup pointed to by the logical address of the data in the global index is Add 1 to the value to indicate that new data has flowed into the storage medium of the first layer in the I/O stack.

For another example, the processor 112 may also delete invalid data in a certain layer in the I/O stack, where the invalid data may be overwritten data in an additional write scenario, or some data with a low reading frequency . The processor 112 can move out and delete the data in the storage medium in the layer, and the processor 112 can also update the global index.

When the global index exists in the form of a bitmap, when the processor 112 updates the global index, the processor 112 can correspond to the layer in the I/O stack in the subgroup pointed to by the logical address of the data in the global index The bit of is set to 0 to indicate that the data does not exist in the storage medium of this layer.

When the global index exists in the form of a counter group, when the processor 112 updates the global index, the processor 112 can point to the subgroup corresponding to the layer in the I/O stack in the subgroup pointed to by the logical address of the data in the global index. The value of the counter is decremented by 1, which indicates that the data in the storage medium of this layer has been eliminated once.

In addition to the data writing process, data flow, and data elimination, the processor 112 may also execute the data reading process, see steps 405 to 408 for details.

Step 405: The processor 112 receives a data read request, the data read request is used to request to read the target data, and the data write request carries the logical address of the target data.

Step 406: The processor 112 queries the global index according to the logical address of the target data, and determines the layer of the target data in the I/O stack.

The processor 112 can determine the character block in the global index that executes the logical block to which the logical address belongs according to the logical address of the target data, and then determine according to the offset of the logical address in the logical block and the data length of the target data The character block points to the subgroup of the logical subblock indicated by the logical address. The processor 112 determines the layer where the target data is located according to the value of each character sub-block in the subgroup.

When the global index exists in the form of a bitmap, the processor 112 determines, according to the logical address of the target data, one or more groups of bits indicating the logical block to which the logical address belongs, and then according to the logical address in the logical block The offset in and the data length of the target data determine the logical sub-block indicated by the logical address, and then determine each subgroup pointing to the logical sub-block in the bitmap. The processor 112 may determine the value of each bit in each subgroup pointing to the logical subblock, and determine that the layer corresponding to the bit value of 1 is the layer where the target data is located.

It should be noted that, in the case that the bit corresponding to the last layer in the I/O stack is not set in each subgroup in the bitmap, if the value of each bit in each subgroup pointing to the logical subblock is 0, the processor 112 may determine that the layer where the target data is located is the last layer in the I/O stack.

When the global index exists in the form of a counter group, the processor 112 determines according to the logical address of the target data one or more groups of counters indicating the logical block to which the logical address belongs, and then according to the logical address in the logical block The offset of the target data and the data length of the target data determine the logical sub-block indicated by the logical address, and then determine each subgroup pointing to the logical sub-block in the bitmap. The processor 112 can determine the value of each counter in each subgroup pointing to the logical subblock, and determine that the layer corresponding to the counter value other than 0 is the layer where the target data is located. If there are multiple counters in a subgroup The value of is not 0, indicating that the data in the logical sub-block has been written multiple times, and the layer where the latest written target data is located is the highest layer among the layers corresponding to multiple non-zero counters.

It should be noted that, in the case that the counter corresponding to the last layer in the I/O stack is not set in each subgroup in the counter group, if the values of each counter in each subgroup pointing to the logical subblock are 0 , the processor 112 may determine that the layer where the target data resides is the last layer in the I/O stack.

Step 407: After the processor 112 determines the layer of the target data in the I/O stack, it can directly read the target data from the layer according to the logical address of the target data.

The processor 112 may query the index of the target data in the layer according to the logical address of the target data, determine the metadata of the target data, and then read the target data from the location indicated by the metadata.

Step 408: The processor 112 feeds back a data read response, where the data read response includes the target data.

Scenario 2: The network card 114 of the node in the storage system 100 executes the data access method provided in the embodiment of the present application.

As shown in Figure 5, it is a data access method provided by the present application, in which the network card 114 of the storage device 110 shown in Figure 1 executes the data writing process and the data reading process, and the network card 114 performs data writing The process and the way of the data reading process are similar to the way that the processor 112 executes the data writing process and the data reading process. Let me repeat.

Step 501: The network card 114 of the storage device 110 receives a data writing request.

Step 502: According to the data writing request, the network card 114 writes the target data into the storage location indicated by the logical address.

When the network card 114 writes the target data into the storage location indicated by the logical address, it may preferentially write the target data into the first layer in the I/O stack, for example, it may preferentially write the target data into the cache.

Step 503: The network card 114 updates the global index, and the updated global index can indicate the layer where the target data is located in the I/O stack.

Step 504: The network card 114 feeds back a data writing response, indicating that the target data has been successfully written.

Steps 501 to 504 are the data writing process. In addition to the data writing process, the network card 114 can also execute the data reading process. Refer to steps 505 to 408 for details.

Step 505: The network card 114 receives a data read request, the data read request is used to request to read the target data, and the data write request carries the logical address of the target data.

Step 506: The network card 114 queries the global index according to the logical address of the target data, and determines the layer of the target data in the I/O stack.

Step 507: After determining the layer of the target data in the I/O stack, the network card 114 can directly read the target data from the layer according to the logical address of the target data.

The network card 114 may query the index of the target data in the layer according to the logical address of the target data, determine the metadata of the target data, and then read the target data from the location indicated by the metadata.

Step 508: The network card 114 feeds back a data read response, and the data read response includes the target data.

The data writing request and the data reading request are processed by the network card 114 of the storage device 110, which can effectively reduce the occupation of the processor 112 of the storage device 110, and can also effectively improve the efficiency of the data writing process and the data reading process.

In this scenario, the network card 114 may also be provided with a cache, and the network card 114 may preferentially write the target data into the cache in the network card 114 when processing the data write request. In this case, the cache in the network card 114 may be added to the I/O stack as a layer of the I/O stack. For example, the cache of the network card 114 is used as the first layer of the I/O stack, and the write cache, read cache, hard disk cache, and capacity layer in the storage device 110 are sequentially used as the second, third, and third layers of the I/O stack. Fourth floor, fifth floor. The writing and flow sequence of data in the I/O stack is still carried out in a top-down direction. The difference from the I/O stack mentioned in the previous description is that a new layer is added to the I/O stack.

In the second scenario, the network card 114 may execute the data writing process and the data reading process, and the processor 112 may execute other data processing processes, such as data flow, data elimination, and the like. For the manner in which the processor 112 executes other data processing procedures, reference may be made to the description in the embodiment shown in FIG. 4 , which will not be repeated here.

In other scenarios, the index of data in one or more layers of the I/O stack of the storage system 100 supports unilateral RDMA access or pass-through access, that is, the client device 200 can use unilateral RDMA access or pass-through access method to read the metadata for the data in that tier or tiers.

Supporting unilateral RDMA access means that the index of the data in the layer or layers is stored in the memory of the storage device 110, and the client device 200 records the start address of the index of the data in the layer or layers. When the client device 200 needs to read the metadata of a certain layer of data, the client device 200 can calculate the memory address of the metadata of the data according to the starting address of the index of the data in the layer and the logical address of the data , the client device 200 may obtain the metadata of the data through unilateral RDMA based on the memory address of the metadata.

Supporting pass-through access means that the storage medium in one or more layers is a persistent storage. Here, the persistent storage is a hard disk as an example. Indexes of data stored in one or more layers are stored in the hard disk of the storage device 110 . The client device 200 side records the starting address of the data index in the layer or layers. When the client device 200 needs to read the metadata of a certain layer of data, the client device 200 can calculate the metadata of the data in the hard disk according to the starting address of the index of the data in the layer and the logical address of the data Based on the storage address of the metadata, the client device 200 can obtain the metadata of the data from the hard disk through direct access. The so-called direct access here indicates that the client device 200 directly reads the data stored in the hard disk through the network card 114 of the storage device 110 and the controller in the hard disk, and the processor 112 of the storage device 110 does not need to participate in the direct access process.

In this scenario, the embodiment of the present application provides a data access method, in which the processor 112 of the storage device 110 does not need to participate. When the client device 200 needs to read the target data, it can read the metadata of the target data from the one or more layers through unilateral RDMA access or direct access, and the client device 200 can also access the target data through unilateral RDMA The global index points to some or all character sub-blocks in the target sub-group of the target logical sub-block, wherein the target logical sub-block is the logical sub-block indicated by the logical address of the target data. The client device 200 can determine the layer of the target data in the I/O stack according to the specific values of some or all of the character sub-blocks accessed, and then determine whether the metadata of the read target data is valid, that is, Whether the target data is stored at the storage address indicated by the metadata of the target data. In the case where it is determined that the metadata of the read target data is valid, if the metadata of the target data indicates that the target data is located in memory, the target data can be read through unilateral RDMA access; if the target data When the metadata indicates that the target data is located in the hard disk, the target data can be read through a direct access method. The data access method in this scenario is described below.

Scenario 3: The client device 200 accesses target data through unilateral RDAM or direct access.

In the embodiment of this application, the client accesses the target data through unilateral RDMA. The difference in unilateral RDAM or direct access is only due to the difference in the storage medium where the target data or data index is located. , whether the target data is read from the storage system 100 through unilateral RDAM, or the target data is accessed from the storage system 100 through direct access, the basic process is the same, the difference is only when the target data is read metadata, or target data. For convenience of description, the layer in the I/O stack of the storage system 100 that supports unilateral RDMA access or pass-through access is called a pass-through layer, that is, there may be one or more pass-through layers in the I/O stack of the storage system 100 .

As shown in Figure 6, a data access method provided by this application, the method includes:

Step 601: the storage device 110 notifies the client device 200 of the memory address of the global index.

On the storage device 110 side, the global index may be stored in the memory of the storage device 110, and the storage device 110 may notify the client device 200 of the starting address of the global index in the memory and the length of the global index as the memory address of the global index .

Step 602: the client device 200 may initiate a first one-sided RDMA to the storage device 110, and the first one-sided RDMA is used to read the metadata of the target data in the target pass-through layer in the I/O stack of the storage system 100. The target pass-through layer is one or more layers of one or more pass-through layers.

Step 603: the client device 200 may initiate a second unilateral RDMA to the storage device 110, and the second unilateral RDMA is used to obtain from the storage device 110 all or part of the characters in each subgroup in the global index pointing to the target logical subblock subblock.

Here, some character sub-blocks may not include character sub-blocks corresponding to the target direct layer. For example, the client device 200 may obtain character sub-blocks corresponding to layers above the target direct layer in each subgroup.

The embodiment of the present application does not limit the order in which the client device 200 initiates the first unilateral RDMA and the second unilateral RDMA to the storage device 110 . The client device 200 can initiate the first unilateral RDMA and the second unilateral RDMA to the storage device 110 within a relatively short period of time, that is, can initiate the first unilateral RDMA and the second unilateral RDMA synchronously.

The first unilateral RDMA and the second unilateral RDMA are described below:

(1) First unilateral RDMA—reading the metadata of the target data in the target pass-through layer in the I/O stack of the storage system 100 .

The client device 200 side records the starting address of the data index in the target pass-through layer. When the client device 200 needs to read the metadata of the target data in the target pass-through layer, the client device 200 can calculate the metadata of the data according to the starting address of the index of the data in the target pass-through layer and the logical address of the data. memory address. This does not limit the manner in which the client device 200 calculates the memory address of the metadata of the target data according to the start address of the data index in the target pass-through layer and the logical address of the target data. For example, the logical address of the target data can be Calculate the memory address of the metadata of the target data by querying the hash table, or acting on the hash function, and performing a learning index.

After determining the memory address of the metadata of the target data, the client device 200 may initiate a first request to the network card 114 of the storage device 110 based on RDMA, and the first request is used to request to read the metadata of the target data. The first request carries the memory address of the metadata of the target data.

After the network card 114 of the storage device 110 receives the first request, the network card 114 of the storage device 110 can process the first request, obtain the metadata of the target data according to the memory address of the metadata of the target data, and obtain the metadata of the target data. The metadata is carried in the first response, and the first response is fed back to the client device 200 .

According to the description of the composition of the I/O stack, it can be known that for the same logical sub-block, data can be stored in different layers in the I/O stack. And because the data in the same logical sub-block can flow in and out at different layers in the I/O stack, this may cause the data in the logical sub-block to be indexed in different layers in the I/O stack, and also That is, the metadata of the data in the logical sub-block is reserved in different layers in the I/O stack. Therefore, the embodiment of the present application does not limit the number of target direct layers, and allows the client device 200 to acquire metadata of target data in multiple target direct layers through the first unicast RDMA. When the number of target pass-through layers is one, the client device 200 may initiate a first unilateral RDMA once to acquire metadata of target data in the target pass-through layer. When there are multiple target pass-through layers, the client device 200 may initiate the first one-sided RDMA multiple times, and each time the first one-sided RDMA acquires the metadata of the target data in one of the target pass-through layers.

However, the metadata of the target data in the target pass-through layer may be invalid, that is, the data stored at the physical address indicated by the metadata of the target data is not the latest written data. In order to be able to verify the validity of the metadata of the target data in the target pass-through layer, the client device 200 may perform a second one-sided RDMA.

(2) The second unilateral RDMA—the storage device 110 acquires each subgroup of the logical subblock indicated by the logical address pointing to the target data in the global index.

The client device 200 may determine the position in the global index of each subgroup of the logical subblock indicated by the logical address pointing to the target data in the global index according to the logical address of the target data.

After determining the position of each subgroup in the global index, since the storage device 110 has notified the client device 200 of the memory address of the global index in the storage device 110, the client device 200 can determine according to the memory address of the global index The memory addresses of the respective subgroups.

Taking the size of each logical block as 256KB, the logical block includes 32 logical sub-blocks, the size of each logical sub-block is 8KB, and each group of bits includes 32 sub-groups as an example. If the location indicated by the LBA of the target data is 1MB+520KB, two character blocks pointing to the logical block to which the logical address belongs can be determined. For example, the two character blocks are the third and fourth character blocks in the global index. Then determine the 32 logical sub-blocks indicated by the logical address according to the offset of the logical address in the logical block and the data length of the target data, for example, determine the 32 sub-blocks pointed to in the third and fourth character blocks 32 subgroups of logical subblocks. In the global index, the subgroups pointing to the 32 logical subblocks are the second subgroup to the 32nd subgroup subblock in the third character block pointing to the logical block and the first in the fourth character block subgroup. The starting positions of the 32 subgroups are positions offset by two character blocks and the length of one subgroup in the global index, and the length of the 32 subgroups is the length of one character block.

The total number of character sub-blocks corresponding to the layer of the I/O layer is set to N in each subgroup in the global index. If the global index exists in the form of a bitmap, the size of a subgroup is equal to N bits. A character block is a group of bits, and the size of a group of bits is 32*N bits. The 32 subgroups are located in the global index at positions offset by 32*N+N bits from the start address and with a length of 32*N bits.

If the global index exists in the form of a counter group, and if each counter occupies M bits, the size of a subgroup is equal to N*M bits. A character block is a group of counters, and the size of a group of counters is 32*N*M bits. The 32 subgroups are located in the global index at positions offset from the start address of 32*N*M+N*M bits and with a length of 32*N*M bits.

The positions of the respective subgroups in the global index are thus determined. Afterwards, the client device 200 can determine the memory address of each subgroup according to the memory address of the global index. For example, the client device 200 may offset the start address of the global index by two character blocks and the length of one subgroup as the start address of each subgroup, and the length of each subgroup is the length of one character block. The start address of each subgroup and the length of each subgroup can be used as the memory address of each subgroup. For another example, the client device 200 may also offset the starting address of the global index by two character blocks and the length of one subgroup as the starting address of each subgroup, and offset the starting address of the global index by three characters After the block and the length of a subgroup are used as the end address of each subgroup, the start address and end address of each subgroup can be used as the memory address of each subgroup.

If the client device 200 only needs to obtain part of the character sub-blocks in each subgroup, and obtains the rest of the character sub-blocks in each subgroup except the character sub-block corresponding to the target direct layer, the client device 200 can further The memory address of each group is processed, and the memory address of the character sub-block corresponding to the target through layer is removed from the memory address of each sub-group to obtain the memory addresses of some character sub-blocks in each sub-group.

Still taking the location indicated by the LBA of the target data as 1MB+520KB, and the data length as 256KB as an example, if each subgroup includes P character subblocks, the character subblock corresponding to the target direct layer in each subgroup is the last one Character sub-blocks, the memory addresses of some character sub-blocks in 32 subgroups can be 32 address segments, the starting address of the 32 address segments is the starting address of the global index offset by two character blocks and a subgroup length address, and the length of each address segment is the length of a sub-block of P-1 characters. Each address segment is separated by the length of a character sub-block.

If the global index exists in the form of a bitmap, the starting position of the memory address of some character sub-blocks in the 32 subgroups is located at the offset starting address of 32*N+N bits in the global index, a total of 32 address segments, each The length of the address segment is P-1 bits, and each address segment is separated by 1 bit.

If the global index exists in the form of a counter group, and if each counter occupies M bits, the size of a subgroup is equal to N*M bits. A character block is a group of counters, and the size of a group of counters is 32*N*M bits. The starting position of the memory address of some character sub-blocks in the 32 subgroups is located at the offset starting address of 32*N*M+N*M bits in the global index, and the length of each address segment is (P-1)*M Bits, each address segment is separated by M bits.

After determining the memory addresses of the subgroups, the client device 200 may initiate a second unilateral RDMA to acquire the subgroups from the storage device 110 . The client device 200 may initiate a second request to the network card 114 of the storage device 110 based on RDMA, the second request is used to request to obtain the subgroups in the global index, and the second request carries the memory address of each subgroup.

After receiving the second request, the network card 114 of the storage device 110 can read the subgroups according to the memory addresses of the subgroups, carry the subgroups in the second response, and send the second response to Client device 200.

After the client device 200 determines the memory address of some of the character sub-blocks in each sub-group, it may initiate a second unilateral RDMA to acquire some of the character sub-blocks in each of the sub-groups from the storage device 110 . The client device 200 may initiate a third request to the network card 114 of the storage device 110 based on RDMA, the second request is used to request to obtain some character sub-blocks in each sub-group in the global index, and the second request carries the character sub-blocks of each sub-group The memory address of the partial character subblock in the group.

After receiving the third request, the network card 114 of the storage device 110 can read some character sub-blocks in each sub-group according to the memory addresses of some character sub-blocks in each sub-group, and read some character sub-blocks in each sub-group. The sub-block is carried in the third response, and the third response is sent to the client device 200 .

Step 604: The client device 200 checks the validity of the metadata of the target data in the target pass-through layer according to the specific values of each subgroup or some character subblocks in each subgroup.

The client device 200 may determine whether the layer where the target data resides is the target pass-through layer according to the respective subgroups.

When the global index exists in the form of a bitmap, the client device 200 can first determine whether the bits other than the bit corresponding to the target direct layer in each subgroup are 1, and the layer corresponding to the bit with a value of 1 Whether to be above this target passthrough layer. If there is a bit of 1 in each subgroup except the bit corresponding to the target pass-through layer, and the layer corresponding to the bit with a value of 1 is located above the target pass-through layer, it means that the latest write in The target data of the logical address is located in the layer corresponding to the bit whose value is 1, and the metadata of the target data in the target pass-through layer is invalid.

If there is a bit with a value of 1 in each subgroup, and the layer corresponding to the bit with a value of 1 is the target pass-through layer. Optionally, there is also a bit with a value of 1, and the layer corresponding to this bit is located under the target pass-through layer, which means that the target data newly written at the logical address is located in the target pass-through layer, and the target data in the target pass-through layer The metadata for is valid.

If there is no bit of 1 in each subgroup except the bits corresponding to the target pass-through layer, it means that the latest target data written in the logical address is located in the target pass-through layer, and the target data in the target pass-through layer Metadata is valid.

When the global index exists in the form of a counter group, the client device 200 can first determine whether the counters in each subgroup other than the counter corresponding to the target direct layer are non-zero, and whether the layer corresponding to the non-zero counter is above the target passthrough layer. If there is a non-zero counter in each subgroup except the counter corresponding to the target pass-through layer, and the layer corresponding to the non-zero counter is above the target pass-through layer, it means that the latest write in The target data of the logical address is located in the layer corresponding to the counter whose value is 1, and the metadata of the target data in the target pass-through layer is invalid.

If there are non-zero counters in each subgroup, and the layer corresponding to the non-zero counter is located in the target direct layer, optionally, there are other counters with a value of 1, and the layer corresponding to the other counter If it is located under the target pass-through layer, it means that the latest target data written in the logical address is located in the target pass-through layer, and the metadata of the target data in the target pass-through layer is valid.

If there is no non-zero counter in each subgroup except the counter corresponding to the target pass-through layer, it means that the latest target data written in the logical address is located in the target pass-through layer, and the target data in the target pass-through layer Metadata is valid.

The client device 200 may determine whether the layer where the target data resides is the target through layer according to specific values of some character sub-blocks in each subgroup.

When the global index exists in the form of a bitmap, the client device 200 can first determine whether some of the bits in each subgroup (that is, some of the character subblocks) are 1, and whether the layer corresponding to the bit with a value of 1 is located in the above the target passthrough layer. If there is a bit of 1 in each subgroup except the bit corresponding to the target pass-through layer, and the layer corresponding to the bit with a value of 1 is located above the target pass-through layer, it means that the latest write in The target data of the logical address is located in the layer corresponding to the bit whose value is 1, and the metadata of the target data in the target pass-through layer is invalid.

If some bits in each subgroup have 1 bits, and the layer corresponding to the bit with a value of 1 is located under the target pass-through layer, it means that the latest target data written in the logical address is located in the target pass-through layer, The metadata of the target data in the target passthrough layer is valid.

If some bits in each subgroup do not have 1 bits, it means that the latest target data written in the logical address is located in the target pass-through layer, and the metadata of the target data in the target pass-through layer is valid.

When the global index exists in the form of a counter group, the client device 200 can first determine whether the counters other than some counters (that is, some character sub-blocks) in each subgroup are non-zero, and the layers corresponding to the non-zero counters Whether to be above this target passthrough layer. If there is a non-zero counter in each subgroup except the counter corresponding to the target pass-through layer, and the layer corresponding to the non-zero counter is above the target pass-through layer, it means that the latest write in The target data of the logical address is located in the layer corresponding to the counter whose value is 1, and the metadata of the target data in the target pass-through layer is invalid.

If there are non-zero counters in some counters in each subgroup, and the layer corresponding to the non-zero counter is located under the target pass-through layer, it means that the latest target data written in the logical address is located in the target pass-through layer, The metadata of the target data in the target passthrough layer is valid.

If there are no non-zero counters in some of the subgroups, it means that the latest target data written in the logical address is located in the target pass-through layer, and the metadata of the target data in the target pass-through layer is valid.

Step 605: If the metadata of the target data in the target pass-through layer is valid, the client device 200 obtains the target data from the storage device 110 by using the metadata of the target data.

That is, the client device 200 may send a fourth request to the network card 114 of the storage device 110 based on RDMA, where the fourth request carries metadata of the target data in the target pass-through layer. After receiving the fourth request, the network card 114 of the storage device 110 can determine the physical address of the target data according to the metadata of the target data in the target pass-through layer, read the target data, and feed back the target data through the fourth response to the client device 200.

If the metadata of the target data in the target pass-through layer is invalid, the client device 200 may acquire the target data from the storage device 110 using the embodiment shown in FIG. 4 or 5 . The client device 200 may also acquire target data from the storage device 110 by using bilateral RDMA.

Figure 5 and Figure 6 respectively show that the data access method provided by this embodiment is executed by the processor 112 and the network card, and in this embodiment, the method can also be executed by other chips different from the processor 112 or the network card . For example, the chip may be a data processing unit (data processing unit, DPU).

In order to further understand the embodiment shown in FIG. 6 , a specific implementation manner of a data access method in scenario three is introduced. Referring to FIG. 7 , in FIG. 7 , the I/O stack in the storage system 100 includes at least two layers, which are respectively a write cache and a read cache, and the global index in the storage device 110 can be stored in memory. Each subgroup of the global index includes two character subblocks, the first character subblock corresponds to the write cache, and the second character subblock corresponds to the read cache. The global index is represented in two forms in memory, one is a bitmap, and the other is a counter group.

Step 701: the storage device 110 notifies the client device 200 of the memory address of the bitmap.

Since the bitmap occupies a relatively small space, when the subsequent storage device 110 needs to read some subgroups or bits of the bitmap from the client device 200, it only needs to read a small space, which can effectively improve the performance of the global index. read efficiency.

Step 702: The client device 200 may initiate a first unilateral RDMA to the storage device 110, and the first unilateral RDMA is used to read metadata of the target data in the read cache in the I/O stack of the storage system 100.

Step 703: The client device 200 may initiate a second unilateral RDMA to the storage device 110, and the second unilateral RDMA is used to obtain from the storage device 110 the first one of the subgroups pointing to the target logical subblock in the bitmap bit.

Step 704: The client device 200 checks the validity of the metadata of the target data in the read cache according to the specific value of the first bit in each subgroup.

The client device 200 first determines whether the first bit in each subgroup is 1. If the first bit is 1, it means that the target data is in the write cache, and the metadata of the target data in the read cache is invalid.

If the first bit in each subgroup is 0, it means that the latest target data written in the logical address is in the read cache, and the metadata of the target data in the target pass-through layer is valid.

Step 705: If the metadata of the target data in the read cache is valid, the client device 200 uses the metadata of the target data to obtain the target data from the storage device 110 through unilateral RDMA. If the metadata of the target data in the read cache is invalid, the client device 200 may acquire the target data from the storage device 110 by using the embodiment shown in FIG. 4 or 5 . The client device 200 may also acquire target data from the storage device 110 by using bilateral RDMA.

Based on the same inventive concept as the method embodiment, the embodiment of the present application also provides a data access device, which is used to execute the processor described in the method embodiment shown in Figures 4, 5, 6, and 7 above. Or the method executed by the network card, related features can refer to the above method embodiment, and will not be repeated here. As shown in FIG. 8, the data access device 800 includes a transmission module 801 and a reading module 802;

The transmission module 801 is configured to receive a data read request, and the data read request is used to request to read the target data stored in the storage device. The transmission module 801 may execute step 405 shown in FIG. 4 or step 505 shown in FIG. 5 .

The reading module 802 is configured to query the global index based on the data reading request, and the global index is used to indicate the storage layer where the target data in the I/O stack is located; read the target data according to the storage layer indicated by the global index. The reading module 802 may execute steps 406 to 408 shown in FIG. 4 or steps 506 to 508 shown in FIG. 5 .

In a possible implementation manner, the data access device 800 further includes a writing module 803 .

The transmission module 801 may receive a data write request, where the data write request is used to request to write target data in the storage system. The transmission module 801 may execute step 401 shown in FIG. 4 or step 501 shown in FIG. 5 .

The writing module 803 may update the global index according to the target data written in the data writing request, and the updated global index is used to indicate the storage layer of the target data in the I/O stack. The reading module 802 may execute steps 402 to 404 shown in FIG. 4 or steps 502 to 504 shown in FIG. 5 .

In a possible implementation, the data read request includes the logical address of the target data, and when the read module 802 queries the global index based on the data read request, it can determine in the global index the target data according to the logical address of the target data Multiple character sub-blocks of the logical address; determine the storage layer where the target data in the I/O stack is located according to the non-zero character sub-blocks in the multiple character sub-blocks.

In a possible implementation manner, the character sub-block is a bit, and the value of the bit includes 0 or 1. 1 indicates that the target data is located in the storage layer corresponding to the character sub-block, and 0 indicates that the target data is located in the storage layer corresponding to the character sub-block. layer.

In a possible implementation manner, the character sub-block is a counter, and the counter is 0 or a non-zero integer, and the non-zero integer is used to indicate that the target data is located in the storage layer corresponding to the character block, and the The non-zero integer is also used to indicate the number of times data is written into the storage layer corresponding to the character sub-block, and the data includes the target data.

In a possible implementation, when the reading module 802 determines a plurality of character sub-blocks pointing to the logical address of the target data in the global index according to the logical address of the target data, it may perform a hash operation according to the logical address of the target data The result of determining a plurality of character sub-blocks pointing to the logical address of the target data, and the hash operation is to query a hash table or act on a hash function.

In a possible implementation, when the reading module 802 determines a plurality of character sub-blocks pointing to the logical address of the target data according to the result of the hash operation on the logical address of the target data, it can be determined according to the result that the global index points to the target A character block of the logical block to which the logical address of the data belongs; and then determine a plurality of character sub-blocks pointing to the logical address of the target data from the character block according to the logical address of the target data.

In a possible implementation manner, the character block includes multiple subgroups, each subgroup corresponds to a logical subblock in the logical block, and each subgroup includes multiple character subblocks, and the reading module 802 can The offset between the logical address and the logical block to which the logical address of the target data belongs determines the target subgroup in multiple subgroups in the character block, and the character subblocks in the target subgroup are multiple characters pointing to the logical address of the target data subblock.

In a possible implementation manner, the metadata of the global index and the target data are located in the memory of the device in the storage device, and the transmission module 801 may feed back the global index and the metadata of the target data to the client device under the first instruction of the client device. Metadata, the first indication is based on RDMA transport. The first indication may be the first request and the second request in the embodiment shown in FIG. 6 , or the first request and the third request in the embodiment shown in FIG. 6 .

In a possible implementation, the global index is located in the memory of the device in the storage device, and the metadata of the target data is located in the persistent storage of the device in the storage device, and the transmission module 801 may send the The end device feeds back the global index, and the second indication is based on RDMA transmission; and the metadata of the target data is obtained from the persistent storage under the third indication of the client device, and the metadata of the target data is fed back to the client device.

In a possible implementation manner, the target data is located in the memory of the device in the storage device, and the transmission module 801 may feed back the target data to the client device under a fourth instruction of the client device, and the fourth instruction is based on the metadata of the target data Data-initiated, based on RDMA transmission. The fourth indication may be the fourth request in the embodiment shown in FIG. 6 .

In a possible implementation manner, the target data is located in the persistent memory of the device in the storage device, and the transmission module 801 can obtain the target data from the persistent memory under the fifth instruction of the client device, and send the target data to the client device The target data is fed back, and the fifth indication is initiated according to the metadata of the target data.

In a possible implementation, the data access device 800 further includes a control module 804, the control module 804 can control the data flow and data elimination in the I/O stack, and update the data according to the data flow and data elimination in the I/O stack global index.

It should be noted that the division of modules in the embodiment of the present application is schematic, and is only a logical function division, and there may be other division methods in actual implementation. Each functional module in the embodiment of the present application may be integrated into one processing module, each module may exist separately physically, or two or more modules may be integrated into one module. The above-mentioned integrated modules can be implemented in the form of hardware or in the form of software function modules.

The above-mentioned embodiments may be implemented in whole or in part by software, hardware, firmware or other arbitrary combinations. When implemented using software, the above-described embodiments may be implemented in whole or in part in the form of computer program products. The computer program product includes one or more computer instructions. When the computer program instructions are loaded or executed on the computer, the processes or functions according to the embodiments of the present invention will be generated in whole or in part. The computer may be a general purpose computer, a special purpose computer, a computer network, or other programmable devices. The computer instructions may be stored in or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from a website, computer, server or data center Transmission to another website site, computer, server, or data center by wired (eg, coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (eg, infrared, wireless, microwave, etc.). The computer-readable storage medium may be any available medium that can be accessed by a computer, or a data storage device such as a server or a data center that includes one or more sets of available media. The available media may be magnetic media (eg, floppy disk, hard disk, magnetic tape), optical media (eg, DVD), or semiconductor media. The semiconductor medium may be a solid state drive (SSD).

Those skilled in the art should understand that the embodiments of the present application may be provided as methods, systems, or computer program products. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.

The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to the present application. It should be understood that each procedure and/or block in the flowchart and/or block diagram, and a combination of procedures and/or blocks in the flowchart and/or block diagram can be realized by computer program instructions. These computer program instructions may be provided to a general purpose computer, special purpose computer, embedded processor, or processor of other programmable data processing equipment to produce a machine such that the instructions executed by the processor of the computer or other programmable data processing equipment produce a An apparatus for realizing the functions specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.

These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to operate in a specific manner, such that the instructions stored in the computer-readable memory produce an article of manufacture comprising instruction means, the instructions The device realizes the function specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.

These computer program instructions can also be loaded onto a computer or other programmable data processing device, causing a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process, thereby The instructions provide steps for implementing the functions specified in the flow chart or blocks of the flowchart and/or the block or blocks of the block diagrams.

Obviously, those skilled in the art can make various changes and modifications to the application without departing from the scope of the application. In this way, if these modifications and variations of the application fall within the scope of the claims of the application and their equivalent technologies, the application also intends to include these modifications and variations.

Claims

A storage system, characterized in that the storage system includes multiple storage layers and processing units, and the time delays for reading data between each storage layer are different,

The processing unit is used for:

receiving a data read request to request to read target data stored in the storage system;

Querying a global index based on the data read request, where the first index item in the global index is used to indicate the storage layer where the target data is located among the multiple storage layers;

Read the target data according to the storage layer indicated by the first index item.
The storage system according to claim 1, wherein the processing unit is further configured to:

receiving a data write request to request to write the target data in the storage system;

Record the first index item in the global index according to the data writing request.
The storage system according to claim 1 or 2, wherein the data read request includes the logical address of the target data, and when the processing unit queries the global index based on the data read request, it specifically uses At:

determining a plurality of character sub-blocks pointing to the logical address of the target data in the global index according to the logical address, the plurality of character sub-blocks belonging to the first index item;

The storage layer where the target data is located in the multiple storage layers is determined according to the values of the multiple character sub-blocks.
The storage system according to claim 3, wherein each character sub-block is used to describe whether data is stored in a storage layer in the plurality of storage layers, a character sub-block is a bit, and the bit The value includes 0 or 1, the 1 indicates that the target data is located in the storage layer corresponding to the one character sub-block, and the 0 indicates that the target data is not located in the storage layer corresponding to the one character sub-block.
The storage system according to claim 3, wherein each character sub-block is used to describe whether data is stored in a storage layer in the plurality of storage layers, and a character sub-block is a counter, and the counter is The value includes 0 or a non-zero integer, the 0 indicates that the target data is not located in the storage layer corresponding to the one character sub-block, and the non-zero integer is used to indicate that the target data is located in the one character sub-block In the corresponding storage layer, the non-zero integer is also used to indicate the number of times data is written into the storage layer corresponding to the one character sub-block, and the data includes the target data.
The storage system according to any one of claims 3 to 5, wherein the processing unit determines a plurality of characters pointing to the logical address of the target data in the global index according to the logical address of the target data Subblocks, specifically for:

A plurality of character sub-blocks pointing to the logical address of the target data are determined according to a result of a hash operation on the logical address of the target data, where the hash operation is to query a hash table or apply a hash function.
The storage system according to claim 6, wherein the processing unit determines a plurality of character sub-blocks pointing to the logical address of the target data according to the result of a hash operation on the logical address of the target data, specifically Used for:

determining a character block in the global index pointing to a logical block to which the logical address of the target data belongs according to the result;

A plurality of character sub-blocks pointing to the logical address of the target data are determined from the character block according to the logical address of the target data.
The storage system according to any one of claims 1-7, wherein the processing unit is a data processor (DPU).
The storage system according to any one of claims 1-8, wherein the processing unit is located in a network card in the storage system or in a central processing unit.
The storage system according to any one of claims 1 to 9, wherein the processing unit is further configured to control data flow and data elimination in the multiple storage layers, and Data flow and data eviction updates the global index.
The storage system according to any one of claims 1-10, wherein the multiple storage layers include a performance layer and a capacity layer, and the performance layer includes one or more of a write cache, a read cache, and a hard disk cache. multiple items, and the capacity layer includes one or more items of solid state disks and mechanical hard disks.
A data access method, characterized in that the method is applied to a storage system, the storage system includes a plurality of storage layers and processing units, and the time delays for reading data between each storage layer are different, and the method includes :

The processing unit receives a data read request to read target data stored in the storage system;

The processing unit queries a global index based on the data read request, and the first index item in the global index is used to indicate the storage layer where the target data is located among the multiple storage layers;

The processing unit reads the target data according to the storage layer indicated by the first index item.
The method of claim 12, further comprising:

The processing unit receives a data write request to request to write the target data in the storage system;

The processing unit records the first index item in the global index according to the data write request.
The method according to claim 12 or 13, wherein the data read request includes the logical address of the target data, and the processing unit queries the global index based on the data read request, comprising:

The processing unit determines a plurality of character sub-blocks pointing to the logical address of the target data in the global index according to the logical address, and the plurality of character sub-blocks belong to the first index item;

The processing unit determines the storage layer where the target data is located in the multiple storage layers according to the values of the multiple character sub-blocks.
The method according to claim 14, wherein each character sub-block is used to describe whether to store data in a storage layer in the plurality of storage layers, a character sub-block is a bit, and the selection of the bit The value includes 0 or 1, the 1 indicates that the target data is located in the storage layer corresponding to the one character sub-block, and the 0 indicates that the target data is not located in the storage layer corresponding to the one character sub-block.
The method according to claim 14, wherein each character sub-block is used to describe whether data is stored in a storage layer in the plurality of storage layers, and a character sub-block is a counter, and the fetching of the counter is The value includes 0 or a non-zero integer, the 0 indicates that the target data is not located in the storage layer corresponding to the one character sub-block, and the non-zero integer is used to indicate that the target data is located in the storage layer corresponding to the one character sub-block In the storage layer, the non-zero integer is also used to indicate the number of times data is written into the storage layer corresponding to the one character sub-block, and the data includes the target data.
The method according to any one of claims 14-16, characterized in that, the processing unit determines a plurality of character substrings pointing to the logical address of the target data in the global index according to the logical address of the target data blocks, including:

The processing unit determines a plurality of character sub-blocks pointing to the logical address of the target data according to a result of a hash operation on the logical address of the target data, and the hash operation is to query a hash table or act on a hash function .
The method according to claim 17, wherein the processing unit determines a plurality of character sub-blocks pointing to the logical address of the target data according to a result of a hash operation on the logical address of the target data, comprising:

The processing unit determines, according to the result, a character block in the global index pointing to a logical block to which the logical address of the target data belongs;

The processing unit determines a plurality of character sub-blocks pointing to the logical address of the target data from the character block according to the logical address of the target data.
The method according to any one of claims 12-18, characterized in that the method further comprises:

the processing unit controls data flow and data retirement in the plurality of storage tiers;

The processing unit updates the global index according to data flow and data elimination in the plurality of storage layers.
A network card, characterized in that the network card is located in the storage system, and the network card is used to execute the method according to any one of claims 12-19.
A processor, wherein the processor is located in the storage system, and the processor is configured to execute the method according to any one of claims 12-19.
A data access device, characterized in that the data processing device is located in a storage system, the storage system further includes a plurality of storage layers, and the time delays for reading data between each storage layer are different, and the data access device Including transmission module and reading module;

The transmission module is configured to receive a data reading request to read the target data stored in the storage system;

The reading module is configured to query a global index based on the data read request, the first index item in the global index is used to indicate the storage layer where the target data is located in the multiple storage layers; according to the The storage layer indicated by the first index item is used to read the target data.
The device according to claim 22, wherein the data access device further comprises a writing module;

The transmission module is further configured to receive a data writing request, requesting to write the target data in the storage system;

The writing module is configured to record the first index item in the global index according to the data writing request.
The device according to claim 22 or 23, wherein the data read request includes the logical address of the target data, and when the read module queries the global index based on the data read request, it specifically uses At:

Determine a plurality of character sub-blocks pointing to the logical address of the target data in the global index according to the logical address, the plurality of character sub-blocks belong to the first index item, and according to the plurality of character sub-blocks The value of determines the storage tier where the target data is located in the multiple storage tiers.
The device according to claim 24, wherein each character sub-block is used to describe whether data is stored in a storage layer in the plurality of storage layers, a character sub-block is a bit, and the selection of the bit The value includes 0 or 1, the 1 indicates that the target data is located in the storage layer corresponding to the character sub-block, and the 0 indicates that the target data is not located in the storage layer corresponding to the character sub-block.
The device according to claim 24, wherein each character sub-block is used to describe whether data is stored in a storage layer in the plurality of storage layers, and a character sub-block is a counter, and the fetching of the counter is The value includes 0 or a non-zero integer, the 0 indicates that the target data is not located in the storage layer corresponding to the character sub-block, and the non-zero integer is used to indicate that the target data is located in the storage layer corresponding to a character sub-block In the layer, the non-zero integer is also used to indicate the number of times data is written into the storage layer corresponding to the one character sub-block, and the data includes the target data.
The device according to any one of claims 24-26, characterized in that, the reading module determines a plurality of logical addresses pointing to the logical address of the target data in the global index according to the logical address of the target data When a character sub-block is used, it is specifically used for:

A plurality of character sub-blocks pointing to the logical address of the target data are determined according to a result of a hash operation on the logical address of the target data, where the hash operation is to query a hash table or apply a hash function.
A data access system, characterized in that the data access system includes a storage system and a client device as described in any one of claims 1-11;

The client device is configured to send a data read request to the storage system to read target data stored in the storage system.
The system of claim 28, wherein the global index and the metadata of the target data are located in memory in the storage system;

The client device is further configured to initiate a first indication to the storage system based on Remote Direct Memory Access (RDMA), where the first indication is used to request the global index and metadata of the target data;

The storage system is further configured to send the global index and metadata of the target data to the client device based on the first indication.
The system according to claim 28, wherein the global index is located in a memory in the storage system, and the metadata of the target data is located in a persistent storage in the storage system;

The client device is further configured to initiate a second indication to the storage system based on Remote Direct Memory Access (RDMA), the second indication is used to request the global index; and initiate a third indication to the storage system, the said third indication is used to request metadata of said target data;

The storage system is further configured to send the global index to the client device based on the second indication; acquire metadata of the target data from the persistent storage based on the third indication, and send the The client device sends metadata of the target data.
A system as claimed in claim 29 or 30 wherein,

The client device is further configured to check whether the metadata of the target data is valid according to the global index; if it is determined that the metadata of the target data is valid, send the The storage system initiates the data read request;

The storage system is further configured to send the target data to the client device based on the data read request.
A computer-readable storage medium, characterized in that instructions are stored in the computer-readable storage medium, and when the computer-readable storage medium is run on a computer, the computer is made to execute the method described in any one of claims 12-19.
A computer program product containing instructions, characterized in that, when it is run on a computer, it causes the computer to execute the method described in any one of claims 12-19.