WO2021208092A1

WO2021208092A1 - Method and device for processing stateful service

Info

Publication number: WO2021208092A1
Application number: PCT/CN2020/085418
Authority: WO
Inventors: 李君瑛
Original assignee: 华为技术有限公司
Priority date: 2020-04-17
Filing date: 2020-04-17
Publication date: 2021-10-21
Also published as: CN115349121A

Abstract

The present application relates to the technical field of communications, and provides a method and device for processing a stateful service, which are used to improve the processing performance of a single service when a network card processes the unloading of a stateful service. The method is applied in a network card, and the network card is connected to a host; the method comprises: receiving an identifier of a first connection from the host, the first connection being a connection in which a stateful service is located; obtaining the context of the first connection according to the identifier of the first connection, the context of the first connection being used to indicate related information of the first connection, and the context of the first connection being a context obtained by updating after the previous data block of the first connection is processed; generating a data processing instruction according to the context of the first connection, wherein the data processing instruction can be used to process data of a first data volume; downloading a data block of the first connection from the host by using a transmission bandwidth corresponding to the first data volume; processing the data block of the first connection according to the data processing instruction to obtain multiple messages; and sending the multiple messages.

Description

Method and device for processing stateful business

Technical field

This application relates to the field of communication technology, and in particular to a method and device for processing stateful services.

Background technique

With the continuous increase of business types and data volumes in cloud networks, the execution of network protocols or storage protocols has become a computationally intensive operation. The host (host) occupies a large amount of central processing unit (CPU) resources when executing a network protocol or a storage protocol, thereby bringing a large CPU load to the host. Smart network interface card (smart network interface card, smart NIC) is a high-performance network access card with a network processor as the core. It has a multi-core and multi-threaded network processor architecture that can be used to process various networks separated from the host. Protocol or storage protocol, which can greatly reduce the CPU load of the host. This way of separating the relevant protocol processing in the host and executing it by the smart network card can be called offload.

In the prior art, the host can communicate with the smart network card through a doorbell (DB) mechanism. For stateful service offloading, the smart network card usually includes three stages from receiving the DB sent by the host to sending the message to the network port. Stage 1 is DB processing: obtain the connection context (context) corresponding to the DB, obtain the working queue element (WQE) of the connection based on the context, and generate direct memory access (DMA) commands, The DMA command carries associated data with the path, and is used to guide how to forward and edit the subsequent download (download) of the message from the host; stage 2 is the execution of the DMA command: download the message and process with the path, the size of the downloaded message is usually It is the maximum transmission unit (MTU); Phase 3 is the editing of the message and the construction of the DMA command.

Among them, the above stage 1 is a stateful stage, that is, the processing of the DB strictly depends on the context of the same connection. Only the core or thread corresponding to the previous DB can update the context, and the core or thread corresponding to the next DB can be updated according to the updated context. Start processing in the context of, and the size of the message downloaded corresponding to each DB is only MTU, therefore, the processing performance of a single service is low.

Summary of the invention

The present application provides a method and device for processing a stateful service, which are used to improve the processing performance of a single service when the network card processes the stateful service unloading.

In order to achieve the above objectives, this application adopts the following technical solutions:

In the first aspect, a method for processing a stateful service is provided, which is applied to a network card, and the network card is connected to a host. For example, the network card can be connected to the host via a PCIe bus. The method includes: receiving an identifier of the first connection from the host For example, the identifier of the first connection can be carried in the first doorbell DB received from the host by the network card, the first connection is the connection where the stateful service is located; the context of the first connection is obtained according to the identifier of the first connection, and the first connection The context of is used to indicate the relevant information of the first connection. The context of the first connection is the context updated after the last data block of the first connection is processed; the data processing instruction is generated according to the context of the first connection, and the data processing instruction is available To process the data of the first data volume; use the sending bandwidth corresponding to the first data volume to download the data block of the first connection from the host; process the data block of the first connection according to the data processing instruction to obtain multiple messages; send multiple Message.

In the above technical solution, when the network card receives the first DB carrying the identifier of the first connection, the network card can generate a data processing instruction that can be used to process data of the first amount of data according to the context of the first connection, and from the host Download the data block of the first connection and process the data block of the first connection according to the data processing instruction to obtain multiple messages, so that the network card can obtain the context of the first connection once based on the first DB, and process based on the context Send multiple messages of the first connection, thereby improving the processing performance of a single stateful service.

In a possible implementation manner of the first aspect, the data processing instruction includes a direct memory access DMA instruction corresponding to the first data volume, and downloading the first connected data block from the host includes: corresponding to the first data volume DMA command to download the first connected data block from the host. In the foregoing possible implementation manners, the smart network card may generate a DMA instruction corresponding to the first data volume based on the first DB, thereby downloading the data block of the first connection from the host based on the DMA, so that the smart network card may be based on the first DB. The context of one connection processes multiple messages of the first connection, thereby improving the processing performance of a single stateful service.

In a possible implementation of the first aspect, the data processing instruction further includes associated data, and processing the data block of the first connection according to the data processing instruction includes: comparing data of the first connection according to the associated data The block is fragmented to obtain multiple data fragments. In the foregoing possible implementation manners, the smart network card may generate associated data corresponding to the first data volume based on the first DB, so as to process the data block of the first connection based on the associated data, so that the smart network card can be based on The context of the first connection processes multiple packets of the first connection, thereby improving the processing performance of a single stateful service.

In a possible implementation of the first aspect, the fragmentation processing includes at least one of the following: inserting multiple markers, deleting part of the header data, deleting part of the tail data, determining the message header, and determining the cyclic redundancy of the payload. After checking the CRC, determine the checksum CS.

In a possible implementation of the first aspect, the data processing instruction further includes a message editing instruction, and processing the data block of the first connection according to the data processing instruction further includes: dividing a plurality of data according to the message editing instruction The film is edited and encapsulated to obtain multiple messages. In the above possible implementation manners, the smart network card can generate a message editing instruction at the same time when generating a DMA instruction, which can avoid subsequent editing and packaging of multiple data fragments to wake up the smart network card corresponding to the first DB. Cores/threads, thereby further improving processing performance.

In a possible implementation of the first aspect, the message editing instructions include: editing instructions for the head slice, editing instructions for the middle slice, and editing instructions for the tail slice. In the foregoing possible implementation manners, the complexity of the message editing instructions can be reduced.

In a possible implementation of the first aspect, before using the sending bandwidth corresponding to the first data volume to download the data block of the first connection from the host, the method further includes: allocating and sending the first DB from the available bus bandwidth. Bandwidth, the sending bandwidth is equal to the first amount of data. In the foregoing possible implementation manners, the smart network card can download multiple packets of the first connection from the host at one time, so that the multiple packets can be processed at the same time based on the context of the first connection subsequently, thereby increasing the number of individual packets. The processing performance of the state business.

In a possible implementation of the first aspect, after downloading the data block of the first connection from the host using the transmission bandwidth corresponding to the first data volume, the method further includes: according to the first data volume and the data of the first connection The difference in the actual data volume of the block adjusts the sending bandwidth. The foregoing possible implementation manners can improve the flexibility of bandwidth allocation and bandwidth utilization of the smart network card.

In a second aspect, a device for processing stateful services is provided, which is applied to a network card. The network card is connected to a host. For example, the network card is connected to the host via a PCIe bus. The identification of the connection, for example, the identification of the first connection can be carried in the first doorbell DB received from the host by the network card, and the first connection is the connection where the stateful service is located; the acquiring unit is also used to acquire according to the identification of the first connection The context of the first connection, the context of the first connection is used to indicate the relevant information of the first connection, the context of the first connection is the context updated after the last data block of the first connection is processed; the processing unit, according to the first connection The context of generating a data processing instruction, the data processing instruction can be used to process the data of the first amount of data; the acquiring unit is also used to download the data block of the first connection from the host using the sending bandwidth corresponding to the first amount of data; the processing unit, It is also used to process the data block of the first connection according to the data processing instruction to obtain multiple messages; the sending unit is used to send multiple messages.

In a possible implementation of the second aspect, the data processing instruction includes a direct memory access DMA instruction corresponding to the first amount of data, and the acquiring unit is further configured to: according to the DMA instruction corresponding to the first amount of data, from the host Download the data block of the first connection.

In a possible implementation of the second aspect, the data processing instruction further includes associated data, and the processing unit is further configured to: perform slicing processing on the data block of the first connection according to the associated data to obtain multiple Data fragmentation.

In a possible implementation of the second aspect, the fragmentation processing includes at least one of the following: inserting multiple markers, deleting part of the header data, deleting part of the tail data, determining the message header, and determining the cyclic redundancy of the payload. After checking the CRC, determine the checksum CS.

In a possible implementation manner of the second aspect, the data processing instruction further includes a message editing instruction, and the processing unit is further used for: editing and encapsulating multiple data fragments according to the message editing instruction to obtain multiple Message. Optionally, the message editing instructions include: editing instructions for the head slice, editing instructions for the middle slice, and editing instructions for the tail slice.

In a possible implementation manner of the second aspect, the device further includes: a bandwidth allocation unit, configured to allocate a transmission bandwidth to the first DB from the available bus bandwidth, and the transmission bandwidth is equal to the first data amount.

In a possible implementation manner of the second aspect, the bandwidth allocation unit is further configured to: adjust the transmission bandwidth according to the difference between the first data amount and the actual data amount of the data block of the first connection.

In another aspect of the present application, a device for processing stateful services is provided. The device is a network card or a chip built in the network card. The device includes a memory and a processor coupled with the memory, and the memory stores code And data, the processor runs the code in the memory to make the device execute the stateful service processing method provided by the first aspect or any one of the possible implementations of the first aspect.

In another aspect of the present application, a communication system is provided. The communication system includes a network card and a host, and the network card is connected to the host through a bus; wherein, the network card is a network card provided in any one of the above aspects, and is used to execute the first Aspect or the stateful service processing method provided by any possible implementation manner of the first aspect.

In yet another aspect of the present application, a computer-readable storage medium is provided. The computer-readable storage medium stores instructions, which when run on a computer, cause the computer to execute the first aspect or the first aspect. The stateful service processing method provided by any of the possible implementations.

In another aspect of the present application, a computer program product is provided. The computer program product includes computer-executable instructions, and the computer-executable instructions are stored in a computer-readable storage medium; at least one processor of the device can be read from the computer-readable storage medium. The computer-executable instruction is read, and at least one processor executes the computer-executed instruction to make the device execute the stateful service processing method provided by the first aspect or any one of the possible implementation manners of the first aspect.

It is understandable that any device, computer storage medium or computer program product for processing stateful services provided above is used to execute the corresponding method provided above. Therefore, the beneficial effects that can be achieved can refer to the above The beneficial effects of the corresponding methods provided in the article will not be repeated here.

Description of the drawings

FIG. 1 is a schematic structural diagram of a communication system provided by an embodiment of this application;

2 is a schematic flowchart of a method for processing a stateful service according to an embodiment of the application;

FIG. 3 is a schematic diagram of a fragmentation process provided by an embodiment of the application;

FIG. 4 is a schematic structural diagram of a smart network card provided by an embodiment of this application;

FIG. 5 is a schematic diagram of multi-message processing provided by an embodiment of this application;

FIG. 6 is a schematic structural diagram of a processing device for a stateful service provided by an embodiment of this application;

FIG. 7 is a schematic structural diagram of another device for processing a stateful service provided by an embodiment of the application.

Detailed ways

In this application, "at least one" refers to one or more, and "multiple" refers to two or more. "And/or" describes the association relationship of the associated objects, indicating that there can be three relationships, for example, A and/or B, which can mean: A alone exists, A and B exist at the same time, and B exists alone, where A, B can be singular or plural. The character "/" generally indicates that the associated objects before and after are in an "or" relationship. "The following at least one item (a)" or similar expressions refers to any combination of these items, including any combination of a single item (a) or a plurality of items (a). For example, at least one of a, b, or c can mean: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple . In addition, the embodiments of the present application use words such as "first" and "second" to distinguish the same items or similar items that have substantially the same function and effect. For example, the first threshold and the second threshold are only for distinguishing different thresholds, and the order of their order is not limited. Those skilled in the art can understand that words such as "first" and "second" do not limit the number and execution order.

It should be noted that in this application, words such as "exemplary" or "for example" are used to indicate examples, illustrations, or illustrations. Any embodiment or design solution described as "exemplary" or "for example" in this application should not be construed as being more preferable or advantageous than other embodiments or design solutions. To be precise, words such as "exemplary" or "for example" are used to present related concepts in a specific manner.

FIG. 1 is a schematic structural diagram of a communication system provided by an embodiment of the application. The communication system includes a host and a network card, and the host and the network card are connected through a bus. For example, the network card is a smart network interface card (smart NIC), and the smart network card is connected to the host through a peripheral component interconnect express (PCIe) bus. Optionally, the communication system may include one or more hosts, and the one or more hosts may all be connected to the smart network card. Hereinafter, the embodiment of the present application will be introduced and explained by taking the network card as a smart network card as an example.

Wherein, the host is provided with multiple virtual machines (virtual machines, VMs), and each VM can run one or more virtual functions (VF), and the one or more VFs can correspond to different functions. A VF may correspond to one or more queues, and the input or output mechanism of the VF is realized through the one or more queues. The multiple queues may include a sending queue and a receiving queue. The smart network card can be used to process various network protocols or storage protocols separated from the host, and it can also be called protocol offload. For example, network offloading can include virtualization I/O (virtualization I/O, Virt IO) offloading, single-root I/O virtualization (single-root I/O virtualization, SR-IOV) offloading, user datagram protocol ( user datagram protocol, UDP)/transmission control protocol (transmission control protocol, TCP)/Internet protocol (Internet Protocol, IP) checksum (checksum, CS) offloading, receiving side expansion (receive side scaling, RSS) offloading/TCP split TCP Segment Offload (TSO)/Large Receive Offload (LRO), Virtual Extensible Local Area Network (VxLAN)/Generic Network Virtualization Encapsulation (Geneve) Offloading, stateful (stateful) open virtual switch (open virtual switch, OVS) offloading, IP security (IP security, IPSec) offloading, TCP offloading (TCP offload engine, TOE) offloading, converged Ethernet remote DMA version 2( RDMA overconverged ethernet V2, RoCEv2) offloading, etc.; storage offloading can include erasure coding (EC) offloading, virtual block service (VBS) offloading, T10 data integrity field (DIF) )/Data Integrity Extension (DIX) Offload, Fibre Channel (FC) Offload, Non-volatile Memory Host Controller Interface Specification Non-volatile Memory Host Controller Interface Specification (nonvolatile memory express , NVMe) offloading and non-volatile memory host controller interface (NVME over fabric, NoF) offloading across the network, etc.

In addition, when the host needs to send a message to the Ethernet (ethernet, Eth) through the smart network card, the host can communicate with the smart network card through the doorbell (DB) mechanism, and the smart network card will process the relevant messages, and Send the processed message to Eth. In a possible embodiment, the smart network card may include: a transmit bandwidth provision (TX bandwidth provision) module, a receive bandwidth provision (RX bandwidth provision) module, a transmit processing (TX processing) module, and a receive processing (RX processing) module , Scheduler, processor pool including multiple processor cores, traffic manager, TX port for sending messages to Eth, and The receiving virtual machine (RX VM) used to send messages to the host. Optionally, the processor pool in the smart network card may be an application specific integrated circuit (ASIC) or a field programmable gate array (FPGA), etc. The embodiment of the application does not specifically limit this .

FIG. 2 is a schematic flowchart of a method for processing a stateful service according to an embodiment of the application. The method can be applied to the communication system including the host and the smart network card shown in FIG. 1. The method includes the following steps.

S201: The smart network card receives the identifier of the first connection from the host, and the first connection is the connection where the stateful service is located.

Among them, a connection (connection) refers to a logical link established by a session at both ends. For example, the connection may be a TCP connection, a UDP connection, or a ROCE queue pair (QP) connection. Multiple connections may be established between the host and the network (Eth) through the smart network card, the first connection may be any one of the multiple connections, and the identifier of the first connection may be used to identify the first connection. Optionally, the identification of the first connection may be carried in the first DB. For example, the smart network card receives the first DB from the host, and the first DB carries the identification of the first connection of the host.

In addition, the above-mentioned stateful services correspond to stateless services. A stateless service may mean that the processing of a single message of the service can be processed based on the header of the message itself, and there is no correlation between the message and the message. A stateful service can mean that a single message of the service cannot decide how to process the message. The processing of the message needs to depend on the state of the "connection" where the message is located, and the information of the message itself, etc., before the decision can be made. The processing behavior of the message, that is, there is an association between the message of the stateful service and the message. The state information of the above "connection" includes but is not limited to: the expected sequence number of the next message and the response (ACK) The serial number, receiving window update and statistical information, etc. For ease of understanding, the following uses firewalls as an example to illustrate stateful firewalls and stateless firewalls. The firewall mentioned here can include firewalls, and can also include firewalls at different levels such as security groups in OpenStack. Among them, the stateless firewall refers to filtering or blocking network data packets based on static values, for example, based on addresses, ports, and protocols, etc. That is, the stateless firewall itself does not care about the current network connection status. A stateful firewall can distinguish the state of a network connection. For example, a stateful firewall can distinguish between a TCP connection and which stage of the TCP connection is currently in. Filter or block network data packets.

When the host needs to send the message of the first connection to the network through the smart network card, the host can send the first DB carrying the identifier of the first connection to the smart network card, which can be run on the first VM in the host. The first VF is sent, and the first VF may be a VF used to perform tasks related to the first connection. Specifically, when the first VF running on the first VM in the host needs to send the message of the first connection to the network through the smart network card, the first VF may push the corresponding working queue element (WQE) ( post) to the first sending queue corresponding to the first VF. The first sending queue can be a queue for sending messages for the first connection. This WQE can be used to describe the size of the sent message and the sending of the message on the host. The storage address in the memory, etc., the WQE can also be called a descriptor. After that, the first VF can send the first DB to the smart network card, so that the smart network card can receive the first DB, and the first DB carries the identifier of the first connection.

Optionally, the first DB may also carry an identifier of the first VF, and the identifier of the first VF may be used to uniquely identify the first VF among multiple VFs in the host. When the first DB only carries the identification of the first connection, the identification of the first connection can be used to uniquely identify the first connection corresponding to the first VF among the multiple VFs of the host; when the first DB carries the first connection When the identification of a connection and the identification of the first VF are used, the identification of the first VF can be used to uniquely identify the first VF among the multiple VFs in the host, and the identification of the first connection can be used for multiple connections corresponding to the first VF. Uniquely identifies the first connection in.

S202: The smart network card obtains the context of the first connection according to the identifier of the first connection.

Specifically, the context of the first connection may be stored in the cache of the smart network card, or the context of the first connection may not be stored. When the context of the first connection is not stored in the smart network card, the smart network card can obtain the context of the first connection from the host according to the identifier of the first connection; when the context of the first connection is stored in the smart network card, the smart network card can Get the context of the first connection from its own cache. Among them, the context of the first connection is used to record related information about the first connection, such as the address of the first sending queue in the memory of the host, the current sending position in the first sending queue, the sending window, and the information of the sent message Serial number, etc. In addition, the context of the first connection is the updated context obtained after the last data block of the first connection is processed.

S203: The smart network card generates a data processing instruction according to the context of the first connection, and the data processing instruction can be used to process data of the first amount of data.

The first data amount may be set in advance, and the first data amount is greater than the maximum transmission unit (MTU). The specific value of the first data volume can be fixed or variable. It can be set by those skilled in the art based on experience or actual conditions. For example, the first data volume can be 64KB. This is not specifically limited.

Specifically, the smart network card may obtain WQE from the first sending queue of the host according to the context of the first connection, for example, based on the address of the first sending queue indicated by the context of the first connection, from the first sending queue of the host Acquire the WQE, and then generate a data processing instruction corresponding to the first data volume based on information such as the size of the sent message indicated by the WQE (for example, the first data volume) and the storage address of the sent message in the memory of the host. Optionally, the data processing instruction may also include a direct memory access (DMA) instruction (command), associated data, and packet edit instruction (packet edit, PE) corresponding to the first data volume. At least one; the path-associated data is used to indicate segmentation information; the message editing instruction can be a three-stage PE instruction, which can specifically include an editing instruction for a head slice, an editing instruction for at least one middle slice, and an editing instruction for a tail slice.

Further, the smart network card may also update the context of the first connection after performing step S203, and store the updated context of the first connection in the buffer, or may send the updated context of the first connection to the host .

S204: The smart network card uses the sending bandwidth corresponding to the first data volume to download the data block of the first connection from the host.

Wherein, the transmission bandwidth corresponding to the first data amount may be equal to the first data amount. Before downloading the data block of the first connection from the host, the smart network card may allocate the sending bandwidth to the first DB from the available bus bandwidth (available bus bandwidth corresponding to the PCIe bus). When the smart network card needs to download the data block of the first connection from the host, the smart network card can use the sending bandwidth to download the data block of the first connection from the host, and the data block of the first connection can include multiple packets of data. data. Optionally, the data processing instruction may include a DMA instruction, and the DMA instruction carries the storage address of the data block of the first connection in the memory of the host, so that the smart network card can download the first connection from the host through the DMA instruction Data block.

In addition, the actual data volume of the first connected data block may be equal to or smaller than the first data volume. When the actual data volume is less than the first data volume, the smart network card can reduce the transmission bandwidth according to the difference between the first data volume and the actual data volume, thereby saving the available bus bandwidth corresponding to the PCIe bus.

S205: The smart network card processes the data block of the first connection according to the data processing instruction to obtain multiple messages.

Wherein, the data processing instruction may include associated data, and the associated data may be used to guide the smart network card to perform fragmentation according to a maximum segment size (MSS) and calculate CS. For example, the associated data may include information indicating the size of each data fragment (ie MSS), information indicating the calculation of the checksum of each data fragment, and information indicating the interval or length of inserting markers , And instructions to delete some data, etc. Therefore, the smart network card can perform fragmentation processing on the data block of the first connection according to the associated associated data to obtain multiple data fragments. The fragmentation processing may include at least one of the following processing: inserting multiple markers, deleting parts Header data, delete part of the tail data, determine the header, determine the cyclic redundancy check (CRC) of the payload, and determine the CS. Optionally, the marker can be DIF, or DIX, etc.; the deleted part of the header data can be redundant data, for example, the part of the header data has been sent before; the deleted part of the tail data can be redundant data, For example, this part of the tail data is less than the payload of one packet.

Exemplarily, as shown in Fig. 3, when the first connected data block is a small computer system interface (Internet, iSCSI) protocol data unit (PDU) from the CPI interface and the iSCSI PDU When it includes iSCSI header (HDR) and iSCSI PDU payload (also called iSCSI PDU data (data)), the smart network card can insert multiple DIFs in the iSCSI PDU payload, determine that the header is iSCSI HDR, and determine The CRC of the iSCSI HDR (represented as HDR CRC in Figure 3) and the CRC of the payload (represented as data CRC in Figure 3), and then the iSCSI PDU payload can be divided into multiple data fragments according to MSS. The PAD in FIG. 3 is the abbreviation of padding, which can be used to indicate the data part of the padding, which can be specifically indicated by the associated data.

In addition, the data processing instruction may also include a message editing instruction, and the smart network card may edit and encapsulate multiple data fragments obtained by the fragment processing to obtain multiple messages. Specifically, when the message editing instruction is a three-segment PE instruction, the smart network card can perform data segmentation on the first data segment of the multiple data segments (also called the first data segment) according to the editing instruction of the header segment. Or the first slice) for encoding and packaging, and according to the editing instructions of at least one intermediate slice, the second data slice of the multiple data slices to the penultimate data slice (also called intermediate data slice or intermediate slice) ) Performs encoding and encapsulation, and encodes and encapsulates the penultimate data slice (also referred to as a tail data slice or a trailer) among the multiple data slices according to the editing instruction of the tail slice to obtain multiple messages. Optionally, the edit instruction of at least one intermediate slice can also be used to indicate the change rule of the header (header) from the second data slice to the penultimate data slice when encoding and encapsulating, for example, the TCP sequence number ( Sequence number, SN) is an incremental method, and IP ID is the same.

S206: The smart network card sends multiple messages.

After the smart network card has edited multiple messages, the smart network card can send multiple messages to the network through the network port. Optionally, after the smart network card sends multiple packets to the network, it may also send feedback information to the host, and the feedback information may be used to indicate that the host has successfully sent multiple packets.

For ease of understanding, the following takes the structure of the smart network card shown in FIG. 4 as an example to illustrate the solutions provided in the embodiments of the present application.

Among them, the TX bandwidth allocation module of the smart network card may include multiple DB queues (DB Q), queue mapping (QM) modules, bandwidth allocation nodes (represented as vNIC in Figure 4), and round robin (RR). ) Scheduler for scheduling (represented as RR in Figure 4). In FIG. 4, description is made by taking multiple DBQs corresponding to multiple hosts (that is, the smart network card is connected to multiple hosts (for example, H0 to H3)) as an example.

Specifically, after the host sends the first DB to the smart network card, the first DB can be queued in multiple DB queues according to certain rules and mapped to the corresponding bandwidth allocation node through the queue mapping module to complete the allocation of the sending bandwidth. Then, the RR allocates processor cores or threads in the processor cores from the processor pool to the first DB, and the allocated processor cores or threads may be referred to as cores/threads in the following. The core/thread can obtain the context of the first connection based on the identifier of the first connection carried by the first DB, and generate the DMA instruction of the first data amount, the associated data and the message editing instruction according to the context of the first connection, and Download the data block of the first connection from the host through a DMA instruction; optionally, the core/thread may also send the difference between the first data volume and the actual data volume of the data block of the first connection to the corresponding bandwidth allocation node, So that the bandwidth allocation node adjusts the sending bandwidth of the first DB. The TX processing module can complete the fragmentation processing of the data block of the first connection according to the associated data, and the obtained multiple data fragments can be stored in the memory of the smart network card. The traffic manager can schedule multiple data fragments from the memory and edit and encapsulate the data fragments according to the message editing instructions, and the resulting message can be sent to the network through the TX port.

Further, the context of the first connection may be decomposed into multiple sub-contexts, so that different processor cores or threads of the smart network card concurrently process different DBs according to different sub-contexts, thereby improving the throughput of the first connection's message. Exemplarily, as shown in Figure 5, the context of the first connection can be decomposed into four sub-contexts and represented as S0, S1, S2, and S3 respectively, so that the smart network card can process the 4 DBs of the first connection at the same time and represent them respectively These are DB1, DB2, DB3 and DB4. Assuming that the processor cores allocated by the scheduler to the 4 DBs are Core1, Core2, Core3, and Core4, after Core1 finishes processing DB1S0, Core1 can continue to process DB1S1, and Core2 can process DB2S0 at this time; when Core1 finishes processing DB1S1, Core1 can continue processing DB1S2, at this time Core2 has finished processing DB2S0 and can start processing DB2S1, while Core3 can start processing DB3S0,..., and so on. Further, if the first connection also corresponds to DB5, when Core1 completes the processing of DB1S3, Core1 can start processing DB5S0, Core2 starts processing DB2S3, Core3 starts processing DB3S2, and Core4 starts processing DB4S1, so that 4 smart network cards can be processed. The cores concurrently participate in the context processing of the first connection, and each core can process a data block of the first connection, so that the throughput of the first connection can be improved.

In the embodiment of the present application, when the smart network card receives the first DB carrying the identifier of the first connection, the smart network card may generate a data processing instruction of the first amount of data according to the context of the first connection, and from the host Download the data block of the first connection, and process the data block of the first connection according to the data processing instruction to obtain multiple messages, so that the smart network card can obtain the context of the first connection once based on the first DB, and process based on the context Send multiple messages of the first connection, thereby improving the processing performance of a single stateful service. In addition, the smart network card can generate message editing instructions at the same time when generating DMA instructions, which can avoid subsequent editing and packaging of multiple data fragments to wake up the core/thread corresponding to the first DB in the smart network card again, thereby Further improve the processing performance.

The foregoing mainly introduces the solutions provided by the embodiments of the present application from the perspective of interaction between various devices. It can be understood that various devices, such as the host and the smart network card. In order to realize the above-mentioned functions, it includes hardware structures and/or software modules corresponding to the respective functions. Those skilled in the art should easily realize that in combination with the units and algorithm steps of the examples described in the embodiments disclosed herein, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a certain function is executed by hardware or computer software-driven hardware depends on the specific application and design constraint conditions of the technical solution. Professionals and technicians can use different methods for each specific application to implement the described functions, but such implementation should not be considered beyond the scope of this application.

The embodiments of the present application may divide the functional modules of the smart network card according to the foregoing method examples. For example, each functional module may be divided corresponding to each function, or two or more functions may be integrated into one module. The above-mentioned integrated modules can be implemented in the form of hardware or software functional modules. It should be noted that the division of modules in the embodiments of the present application is illustrative, and is only a logical function division, and there may be other division methods in actual implementation. The following is an example of dividing each function module corresponding to each function.

In the case of using an integrated unit, FIG. 6 shows a possible structural schematic diagram of the stateful service processing apparatus involved in the foregoing embodiment. The device may be a smart network card or a chip built in the smart network card, and the device includes: an acquiring unit 601, a processing unit 602, and a sending unit 603. Wherein, the acquiring unit 601 is used to support the device to perform S201, S202, and S204 in the above method embodiment; the processing unit 602 is used to support the device to perform S203 and S205 in the above method embodiment; the sending unit 603 is used to support the device Perform S206 in the foregoing method embodiment. Further, the device may further include: a bandwidth allocating unit 604, configured to support the device to perform the step of allocating the sending bandwidth in the foregoing method embodiment.

In practical applications, the processing unit 602 may be the integration of the TX processing module, scheduler, processor pool, and traffic manager in the smart network card described in the foregoing method embodiment, and the bandwidth allocation unit 604 may be the foregoing method embodiment. In the described TX bandwidth allocation module in the smart network card, the sending unit 603 may be the TX port in the smart network card described in the foregoing method embodiment.

It should be noted that all relevant content of each step involved in the foregoing method embodiment can be cited in the functional description of the corresponding functional module. For details, please refer to the description in the foregoing method embodiment, and the description of the embodiment of this application will not be repeated here.

On the basis of hardware implementation, the processing unit 602 in this application may be the processor of the device, the acquiring unit 601 may be the receiver of the device, and the sending unit 603 may be the transmitter, receiver, and transmitter of the device. Usually can be integrated together as the communication interface of the device.

As shown in FIG. 7, it is a schematic diagram of a possible logical structure of the apparatus for processing a stateful service involved in the foregoing embodiment provided by the embodiment of this application. The device may be a smart network card or a chip built in the smart network card, and the device includes a processor 702 and a communication interface 703. The processor 702 is used to control and manage the actions of the device. For example, the processor 702 is used to support the device to execute S203 and S205 in the foregoing method embodiments, and/or other processes used in the technology described herein. In addition, the device may also include a memory 701 and a bus 704. The processor 702, a communication interface 703, and the memory 701 are connected to each other through the bus 704; the communication interface 703 is used to support the device to communicate, for example, to support the device to communicate with a host or a network. Communication; The memory 701 is used to store the program code and data of the device.

Among them, the processor 702 may be a central processing unit, a general-purpose processor, a baseband processor, a digital signal processor, an application specific integrated circuit, a field programmable gate array or other programmable logic devices, transistor logic devices, hardware components, or any of them combination. It can implement or execute various exemplary logical blocks, modules, and circuits described in conjunction with the disclosure of this application. The processor may also be a combination that implements computing functions, for example, a combination of one or more microprocessors, a combination of a digital signal processor and a microprocessor, and so on. The bus 704 may be a peripheral component interconnect standard (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus 704 can be divided into an address bus, a data bus, a control bus, and the like. For ease of representation, only one thick line is used in FIG. 7, but it does not mean that there is only one bus or one type of bus.

In another aspect of the present application, a communication system is provided. The communication system includes a network card and a host, and the network card is connected to the host via a bus; wherein, the network card is any one of the network cards provided above, and is used to perform the above-mentioned method implementation. The steps of the network card in the example.

In the several embodiments provided in this application, it should be understood that the disclosed device and method may be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the modules or units is only a logical function division. In actual implementation, there may be other division methods, for example, multiple units or components may be It can be combined or integrated into another device, or some features can be omitted or not implemented.

The units described as separate parts may or may not be physically separate. The parts displayed as units may be one physical unit or multiple physical units, that is, they may be located in one place, or they may be distributed to multiple different places. . Some or all of the units may be selected according to actual needs to achieve the objectives of the solutions of the embodiments.

In addition, the functional units in the various embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units may be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or software functional unit.

If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. The readable storage medium may include: U disk, mobile hard disk, read-only Various media that can store program codes such as memory, random access memory, magnetic disk or optical disk. Based on this understanding, the technical solutions of the embodiments of the present application can be embodied in the form of software products in essence or a part that contributes to the prior art, or all or part of the technical solutions.

In another embodiment of the present application, a readable storage medium is also provided. The readable storage medium stores computer execution instructions. When a device (may be a single-chip microcomputer, a chip, etc.) or a processor executes the above method embodiments The stateful service processing method provided.

In another embodiment of the present application, a computer program product is also provided. The computer program product includes computer-executable instructions, and the computer-executable instructions are stored in a computer-readable storage medium; at least one processor of the device can be accessed from a computer. The reading storage medium reads the computer-executable instruction, and at least one processor executes the computer-executable instruction to make the device execute the stateful service processing method provided by the foregoing method embodiment.

Finally, it should be noted that the above are only specific implementations of this application, but the scope of protection of this application is not limited to this. Any changes or substitutions within the technical scope disclosed in this application shall be covered by this application. Within the scope of protection applied for. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

Claims

A method for processing stateful services, which is characterized in that it is applied to a network card, and the network card is connected to a host, and the method includes:

Receiving an identifier of a first connection from the host, where the first connection is a connection where a stateful service is located;

Acquire the context of the first connection according to the identifier of the first connection, the context of the first connection is used to indicate related information of the first connection, and the context of the first connection is that of the first connection The updated context obtained after the last data block was processed;

Generating a data processing instruction according to the context of the first connection, where the data processing instruction can be used to process data of a first amount of data;

Download the data block of the first connection from the host by using the transmission bandwidth corresponding to the first data volume;

Process the data block of the first connection according to the data processing instruction to obtain multiple messages;

Sending the multiple messages.
The method according to claim 1, wherein the data processing instruction includes a direct memory access DMA instruction corresponding to the first data volume, and the data block of the first connection is downloaded from the host ,include:

Downloading the data block of the first connection from the host according to the DMA instruction corresponding to the first data volume.
The method according to claim 1 or 2, wherein the data processing instruction further includes associated data, and the processing of the data block of the first connection according to the data processing instruction includes:

Fragment processing is performed on the data block of the first connection according to the path associated data to obtain multiple data fragments.
The method according to claim 3, wherein the fragmentation processing includes at least one of the following:

Insert multiple markers, delete part of the header data, delete part of the tail data, determine the message header, determine the cyclic redundancy check CRC of the payload, and determine the checksum CS.
The method according to claim 3 or 4, wherein the data processing instruction further comprises a message editing instruction, and the processing of the data block of the first connection according to the data processing instruction further comprises:

According to the message editing instruction, the multiple data fragments are edited and packaged to obtain the multiple messages.
The method according to claim 5, wherein the message editing instruction includes: an editing instruction for a head slice, an editing instruction for a middle slice, and an editing instruction for a tail slice.
The method according to any one of claims 1-6, wherein the receiving the identifier of the first connection from the host comprises:

Receive a first doorbell DB from the host, where the first DB carries an identifier of the first connection in the host.
7. The method according to claim 7, wherein before the downloading the data block of the first connection from the host by using the sending bandwidth corresponding to the first data amount, the method further comprises:

The transmission bandwidth is allocated to the first DB from the available bus bandwidth, where the transmission bandwidth is equal to the first data amount.
8. The method according to claim 8, wherein after said downloading the data block of the first connection from the host using the sending bandwidth corresponding to the first data amount, the method further comprises:

Adjust the sending bandwidth according to the difference between the first data volume and the actual data volume of the data block of the first connection.
A processing device for stateful services, characterized in that it is applied to a network card, the network card is connected to a host, and the device includes:

An obtaining unit, configured to receive an identifier of a first connection from the host, where the first connection is a connection where a stateful service is located;

The acquiring unit is further configured to acquire the context of the first connection according to the identifier of the first connection, and the context of the first connection is used to indicate related information of the first connection. The context is the context obtained by updating after the last data block of the first connection is processed;

A processing unit, generating a data processing instruction according to the context of the first connection, the data processing instruction may be used to process data of a first amount of data;

The acquiring unit is further configured to download the data block of the first connection from the host by using the transmission bandwidth corresponding to the first data amount;

The processing unit is further configured to process the data block of the first connection according to the data processing instruction to obtain multiple messages;

The sending unit is configured to send the multiple messages.
11. The device according to claim 10, wherein the data processing instruction comprises a direct memory access DMA instruction corresponding to the first amount of data, and the acquiring unit is further configured to:

Downloading the data block of the first connection from the host according to the DMA instruction corresponding to the first data volume.
The device according to claim 10 or 11, wherein the data processing instruction further includes associated data, and the processing unit is further configured to:

Fragment processing is performed on the data block of the first connection according to the path associated data to obtain multiple data fragments.
The device according to claim 12, wherein the fragmentation processing includes at least one of the following: inserting multiple markers, deleting part of the header data, deleting part of the tail data, determining the header of the message, and determining the cycle of the load Redundancy check CRC, determine the checksum CS.
The device according to claim 12 or 13, wherein the data processing instruction further comprises a message editing instruction, and the processing unit is further configured to:

According to the message editing instruction, the multiple data fragments are edited and packaged to obtain the multiple messages.
The device according to claim 14, wherein the message editing instructions include: editing instructions for the head slice, editing instructions for the middle slice, and editing instructions for the tail slice.
The device according to any one of claims 10-15, wherein the acquiring unit is further configured to:

Receive a first doorbell DB from the host, where the first DB carries an identifier of the first connection in the host.
The device according to claim 16, wherein the device further comprises:

The bandwidth allocation unit is configured to allocate the sending bandwidth to the first DB from the available bus bandwidth, where the sending bandwidth is equal to the first data amount.
The apparatus according to claim 17, wherein the bandwidth allocation unit is further configured to:

Adjust the sending bandwidth according to the difference between the first data volume and the actual data volume of the data block of the first connection.
A processing device for stateful services, characterized in that the processing device for stateful services is a network card or a chip built into the network card, and the device includes a memory and a processor coupled with the memory. Stores code and data, and the processor runs the code in the memory to enable the device to execute the method for processing a stateful service according to any one of claims 1-9.
A computer-readable storage medium having instructions stored in the computer-readable storage medium, which when run on a computer, causes the computer to execute the method for processing a stateful service according to any one of claims 1-9 .
A computer program product, characterized in that, when the computer program product runs on a device, the device is caused to execute the method for processing a stateful service according to any one of claims 1-9.