CN106302201A - Flow control methods, equipment and system - Google Patents

Abstract

The embodiment of the invention discloses a flow control method, device and system. The flow control method includes: obtaining a first message data packet from a receiving queue, where the message data packet sent by the first device to the second device is stored; and deleting the obtained first message data packet in the receiving queue , if the number of message data packets stored in the receiving queue is less than or equal to the first threshold, sending a data import instruction carrying credit data to the first device. The second device directly updates the credit data in the memory of the first device through the data import command, and the first device controls the upper limit of the number of data packets it sends to the second device according to the credit data in the memory; the credit is completed from the hardware level The change of data saves the interaction time required for changing the credit data at the software level, advances the time for the first device to control sending data packets to the second device according to the credit data, and improves the utilization rate of network bandwidth.

Description

Flow control method, device and system

technical field

The embodiments of the present invention relate to the network field, and in particular to a flow control method, device and system.

Background technique

Whether the message communication between devices is based on Transmission Control Protocol/Internet Protocol (Transmission Control Protocol/Internet Protocol, referred to as TCP/IP) protocol or other message communication that supports communication protocols (such as socket interface protocol), it is necessary to Messages are encapsulated into data packets, and data packets are exchanged between devices to realize message communication; in order to ensure the probability of correctly receiving data packets, a flow control mechanism is determined, through which one device (sending end) is coordinated to another device (Receiver) The speed at which data is transmitted. In the flow control mechanism, when the receiving end is ready to receive data, it generates a message carrying credit data (CreditData), encapsulates the message into a data packet (flow control packet), and sends the flow control packet to the sending end; the sending end analyzes the The flow control packet obtains the message carrying the credit data, and then the sender determines the maximum number of bytes that can be sent before receiving the flow control packet next time according to the credit data. When the sending end sends a data packet, when the number of bytes of the sent data packet reaches the maximum number of bytes determined according to the credit data, the sending end must wait for the next flow control packet sent by the receiving end and parse out the flow control packet from the flow control packet. The next credit data can continue to send data packets to the receiving end according to the next credit data.

Taking Queue Pair (QP for short) as an example, when the sending end expects to provide services to the receiving end, it establishes a QP channel, encapsulates the data of the service in a message, encapsulates the message in a data packet, and passes the The QP channel sends the data packet to the receiving end; each time the receiving end receives a data packet from the sending end, it will consume a receiving buffer grid in a receiving queue, and the receiving queue is composed of multiple receiving buffer grids. After the data packet in the receiving buffer is taken away by the application program, it can be used to receive the data packet sent from the sending end again, and the receiving buffer that can be used to receive the data packet is called an available buffer. In addition, the sending end and the receiving end have also established a QP channel for flow control; the receiving end, whenever the number of its available buffer grids reaches half of the receiving queue, generates a data packet (flow control packet) carrying credit data , transmit the flow control packet to the sender through the QP channel and use the QP send and receive operation; then the sender parses the credit data from the flow control packet, and determines the maximum data packets that can be sent to the receiver according to the parsed credit data the number of .

Therefore, using the QP channel to transmit the flow control packet carrying credit data to realize the flow control between the receiving end and the sending end requires more software overhead (including the establishment of a new QP channel, data packet encapsulation at the receiving end, Data packet analysis) delays the time for the sender to obtain the credit data, thereby prolonging the interval between sending two control data packets according to the credit data, and reducing the utilization rate of network bandwidth.

Contents of the invention

In view of this, the embodiments of the present invention provide a flow control method, device, and system, which change the credit data used for flow control through remote memory access instructions, and complete the change of credit data from the hardware level, eliminating the need to modify the credit data at the software level. The interaction time required for the data.

In a first aspect, an embodiment of the present invention provides a flow control method, and the flow control method includes:

Obtaining the first message data packet from the receiving queue, where the message data packet sent by the first device to the second device is stored in the receiving queue;

Delete the obtained first message data packet in the receiving queue, and if the number of message data packets stored in the receiving queue is less than or equal to the first threshold, send a message containing A data import instruction for credit data, writing the credit data into a preset address in the internal memory of the first device, so that the first device determines to transfer to the second device according to the credit data recorded at the preset address 2. The upper limit of the number of message data packets sent by the device, wherein the credit data is the number of message data packets expected to be received by the receiving queue, and the data import instruction belongs to a remote memory access instruction.

With reference to the first aspect, in a first possible implementation manner, the sending a data import instruction carrying credit data to the first device writes the credit data into a preset address, specifically:

The memory resource controller MRC of the second device sends the data import instruction to the MRC of the first device, so that the MRC of the first device forwards the data import instruction to the memory controller of the first device MC, the MC of the first device writes the credit data into the preset address according to the data import instruction.

With reference to the first possible implementation of the first aspect or the second possible implementation of the first aspect, in a third possible implementation, the data import instruction includes: the identifier of the first device, A physical address corresponding to the preset address.

In a second aspect, an embodiment of the present invention provides a second device, where the second device includes:

A data processing module, configured to acquire a first message data packet from a receiving queue, where the message data packet sent by the first device to the second device is stored in the receiving queue;

The data processing module is further configured to delete the acquired first message data packets in the receiving queue, if the number of message data packets stored in the receiving queue is less than or equal to the first threshold, Determining credit data, the credit data being the number of message packets expected to be received by the receiving queue;

An instruction sending module, configured to send a data import instruction carrying credit data to the first device, and write the credit data into a preset address in the memory of the first device, so that the first device The credit data recorded at the preset address determines the upper limit of the number of message data packets sent to the second device, and the data import instruction belongs to a remote memory access instruction.

With reference to the second aspect, in a first possible implementation manner, the instruction sending module is configured to send a data import instruction carrying credit data to the first device, and write the credit data into the first device. The preset address in the device memory, specifically:

The instruction sending module is the memory resource controller MRC of the second device, configured to send the data import instruction to the MRC of the first device, so that the MRC of the first device forwards the data import instruction to The memory controller MC of the first device writes the credit data into the preset address by the MC of the first device according to the data import instruction.

With reference to the first possible implementation of the second aspect or the second possible implementation of the second aspect, in a third possible implementation, the data import instruction includes: the identifier of the first device, A physical address corresponding to the preset address.

In a third aspect, an embodiment of the present invention provides a flow control method, the flow control method comprising:

receiving the data import command sent by the second device to the first device, writing credit data to the preset address pointed to by the data import command in the memory, wherein the data import command carries the credit data, and the data Import instructions are remote memory access instructions;

The upper limit of the number of message data packets sent to the second device is determined according to the credit data recorded at the preset address.

With reference to the third aspect, in a first possible implementation manner, the receiving a data import instruction sent by the second device to the first device writes the credit data to a preset address pointed to by the data import instruction in the memory, Specifically include:

The MRC of the first device receives the data import instruction sent from the MRC of the second device, and forwards the data import instruction to the memory controller MC of the first device;

The MC of the first device writes the credit data to the preset address according to the data import instruction.

With reference to the first possible implementation of the third aspect or the second possible implementation of the third aspect, in a third possible implementation, the data import instruction includes: the identifier of the first device, A physical address corresponding to the preset address.

In a fourth aspect, an embodiment of the present invention provides a first device, where the first device includes:

An instruction receiving module, configured to receive a data import instruction sent by the second device to the first device, wherein the data import instruction carries credit data, and the data import instruction belongs to a remote memory access instruction;

An instruction response module, configured to write the credit data to a preset address pointed to by the data import instruction in the memory;

A message data packet limiting module, configured to determine the upper limit of the number of message data packets sent to the second device according to the credit data recorded in the preset address.

With reference to the fourth aspect, in a first possible implementation manner, the instruction receiving module is a memory resource controller MRC of the first device, configured to receive a data import instruction sent from the MRC of the second device, and forward the data importing instructions into the memory controller MC of the first device;

The instruction response module is the MC of the first device, configured to write the credit data to the preset address according to the data import instruction.

With reference to the first possible implementation of the fourth aspect or the second possible implementation of the fourth aspect, in a third possible implementation, the data import instruction includes: the identifier of the first device, A physical address corresponding to the preset address.

In a fifth aspect, an embodiment of the present invention provides a system, the system includes a first device and a second device, and the first device and the second device are connected in communication;

The second device is configured to obtain a first message data packet from a receiving queue, where the message data packet sent by the first device to the second device is stored in the receiving queue;

The second device is configured to delete the acquired first message data packets in the receiving queue, and if the number of message data packets stored in the receiving queue is less than or equal to the first threshold, send The first device sends a data import command carrying credit data, the credit data is the number of message packets expected to be received by the receiving queue, and the data import command belongs to a remote memory access command;

The first device is configured to write the credit data into a preset address in its memory according to the data import instruction, and determine the credit data to be sent to the second device according to the credit data recorded at the preset address. The upper limit of the number of message packets.

With reference to the fifth aspect, in a first possible implementation manner, the first device is configured to write the credit data to a preset address in its internal memory according to the data import instruction, specifically:

The first device is configured to have the memory resource controller MRC of the first device receive the data import instruction sent from the MRC of the second device, and forward the data import instruction to the first device's A memory controller MC, wherein the MC of the first device writes the credit data to the preset address according to the data import instruction.

With reference to the first possible implementation of the fifth aspect or the second possible implementation of the fifth aspect, in a third possible implementation, the data import instruction includes: the identifier of the first device, A physical address corresponding to the preset address.

Through the above scheme, the second device directly updates the credit data in the memory of the first device through the data import command, and the first device controls the maximum number of data packets it sends to the second device according to the credit data in the memory; The change of credit data is completed at the hardware level, which saves the interaction time required for changing the credit data at the software level, advances the time for the first device to control the sending of data packets to the second device based on the credit data, and improves the utilization of network bandwidth .

Description of drawings

FIG. 1 is a schematic diagram of a system logical structure of an application scenario of a flow control method;

FIG. 2 is an exemplary flowchart of a flow control method provided according to an embodiment of the present invention;

FIG. 3 is a schematic diagram of a logical structure provided by a second device 300 according to an embodiment of the present invention;

FIG. 4 is an exemplary flowchart of a flow control method provided according to an embodiment of the present invention;

FIG. 5 is an optional detailed flowchart of step S401 in FIG. 4;

FIG. 6 is a schematic diagram of a logical structure provided by a first device 600 according to an embodiment of the present invention.

detailed description

The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some of the embodiments of the present invention, but not all of them. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

The system 100 shown in FIG. 1 includes a first device 101 and a second device 102 . It should be noted that the "first" in the "first device" and the "second" in the "second device" refer to each other and are used to distinguish devices.

For the first device 101 in FIG. 1 , the first device 101 has a processor 1013 , a memory resource controller (Memory Resources Controller, MRC for short) 1011 , a memory controller (Memory Controller, MC for short) 1012 and a memory 1014 . In the first device 101, the processor 1013 is connected to the memory resource controller 1011 and/or the memory controller 1012 via a bus, or the memory resource controller 1011 and/or the memory controller 1012 is directly electrically connected to the processor 1013; Optionally, the memory resource controller 1011 and/or the memory controller 1012 are integrated into the processor 1013, the memory resource controller 1011 integrated into the processor 1013 is directly electrically connected to the processor 1013, and integrated into the processor 1013 The internal memory controller 1012 is directly electrically connected to the processor 1013 . The memory controller 1012 reads or writes data to the local memory (that is, the memory 1014 ) under the control of the processor 1013 . The memory resource controller 1011 reads or writes data to the remote memory under the control of the processor 1013 . For example, the memory 1024 of the second device 102 is a remote memory relative to the first device 101 . Both local memory and remote memory belong to memory and are used to store data.

For the second device 102 in FIG. 1 , there are a processor 1023 , a memory resource controller 1021 , a memory controller 1022 and a memory 1024 . In the second device 102, the processor 1023 is connected to the memory resource controller 1021 and/or the memory controller 1022 through a bus, or the memory resource controller 1021 and/or the memory controller 1022 is directly electrically connected to the processor 1023; Optionally, the memory resource controller 1021 integrated into the processor 1023 is directly electrically connected to the processor 1023 , and the memory controller 1022 integrated into the processor 1023 is directly electrically connected to the processor 1023 . The memory controller 1022 reads or writes data to the local memory (that is, the memory 1024 ) under the control of the processor 1023 . The memory resource controller 1021 reads or writes data to the remote memory under the control of the processor 1023 . For example, the memory 1014 in the first device 101 is a remote memory relative to the second device 102 . Both local memory and remote memory belong to memory and are used to store data.

In the system 100, the first device 101 is communicatively connected with the second device 102, and the first device 101 and the second device 102 realize data interaction through the interaction of data packets. Wherein, for service interaction, the first device 101 and the second device 102 implement service interaction through message interaction.

Taking the first device 101 expecting to send service data to the second device 102 as an example, the processor 1013 of the first device 101 encapsulates the service data in a message, and then encapsulates the message in a data packet, and stores the service data in the sending buffer Store encapsulated data packets in sequence; if it is desired to perform flow control on the data packets sent from the first device 101 to the second device 102, then determine the latest credit data, and the first device 101 controls the hardware transceiver module according to the credit data. The number of the data packets sent by the second device 102 is at most the credit data; correspondingly, the hardware transceiver module of the second device 102 receives the data packets sent by the first device 101, and in the second The received data packet is stored in the receiving queue of the device 102, and the data packet is parsed to obtain a message carrying service data, and the processor 1023 of the second device 102 parses the message to obtain service data. Similarly, it may be implemented that the second device 102 sends service data to the first device 101 , which will not be repeated here.

In one embodiment of the present invention, a queue pair (Queue Pair, QP) channel is established for the service, and service data interacted between the first device and the second device is transmitted through the QP channel, and the service data is defined as service data. Taking the service provided by the first device to the second device as an example, the data packet carrying the service data is transmitted through the QP channel; The number is controlled to achieve flow control.

In order to realize the flow control, a section of memory space is allocated in the memory of the first device, and the section of memory space is used to store credit data, and the physical address of the section of memory space is defined as the first physical address; A physical address is subjected to address mapping, and the address mapping obtains the virtual address of the memory space, and the processor of the first device can access the credit data according to the virtual address (for example, read the credit data according to the virtual address); optionally, The first device has a memory management unit (Memory Management Unit, MMU). When the processor accesses the credit data, the MMU completes the address mapping from the virtual address to the first physical address. In addition, the first physical address of the segment of memory space and the identifier of the first device are encrypted to generate a data packet of encrypted data, and send the data packet to the second device; the second device parses the encrypted data packet from the data packet. The first physical address and the identifier of the first device are decrypted to obtain: the first physical address and the identifier of the first device.

In addition, a section of physical addresses is allocated in advance in the second device, and the allocated section of physical addresses is defined as the second physical address. After obtaining the first physical address through decryption, the second device determines a mapping relationship between the first physical address and the second physical address.

During the process of the first device sending data packets carrying services to the second device, the second device sequentially stores the data packets received from the first device in the receiving queue, and each time a data packet is received from the first device, the Store the data packet in a receiving buffer grid of the receiving queue, that is, a data packet needs a receiving buffer grid to store; in addition, the application of the service can read and store the data packet from the receiving buffer grid of the receiving queue Take away the data packet of the service; the receiving buffer grid of the taken data packet can continue to be used to receive the data packet sent from the first device; when the number of message data packets in the receiving queue is greater than the first threshold When it is reduced to be less than or equal to the first threshold, that is, the receiving buffer grid that can be used to receive data packets in the receiving queue increases from less than the second threshold to greater than or equal to the second threshold (optionally, the first threshold and the second When the sum of the thresholds is equal to the total number of receiving buffer grids contained in the receiving queue), the credit data is determined once, and a data import instruction is generated, and the data import instruction includes the second physical address and the identification of the first device, and the data import instruction It also includes determined credit data, in which it is recorded: the number of message data packets that the second device expects to receive by its receiving queue; the second device determines the data according to the second physical address in the data import instruction The import instruction does not access the memory of the second device but the memory of the first device, and replaces the second physical address in the data import instruction with: the first physical address mapped with the second physical address; then, directly to the second physical address The credit data determined this time is written into the first physical address pointed to by the data import instruction in the memory of a device. Optionally, when the memory controller (memory controller, MC for short) of the second device recognizes that the data import instruction is a remote memory access instruction pointing to the second physical address, the data import instruction is forwarded to the second device's The memory resource controller (MemoryResources Controller, referred to as MRC) processing; the MRC of the second device replaces the second physical address in the data import instruction with the corresponding first physical address, and then sends the data import instruction after the address replacement to The MRC of the first device; when the MRC of the first device recognizes that the first physical address pointed to by the data import instruction is the physical address of the local memory of the first device, it forwards the data import instruction to the MC of the first device; The MC of a device writes the currently determined credit data into the first physical address according to the data import instruction to update the credit data originally stored in the first physical address.

Optionally, the remote memory access instruction includes a load instruction and a store instruction, and the data import instruction is a store instruction. According to the store instruction pointing to the preset address of the internal memory of the first device, the second device can directly perform remote data writing to the preset address, and update the credit data stored in the preset address.

In this embodiment, when the first device sends a data packet carrying a service to the second device, in order to control the number of data packets sent to realize flow control, a section of memory space is allocated in the memory of the first device. A section of memory space used to store credit data; after notifying the second device of the first physical address of the section of memory space and the identifier of the first device, the second device can directly access the section of memory space through remote memory access instructions , change the credit data stored in this memory space. When the second device changes the credit data stored in the memory space, it belongs to the data transmission at the hardware level. Compared with the QP, it needs to allocate a buffer to receive the data packet carrying the credit data, parse the credit data from the data packet, and then The processor of the first device writes the credit data in this section of memory space, which reduces the time delay for updating credit data in this section of memory space of the first device, and advances the time to continue sending loads to the second device based on the credit data. The sending time of data packets with business can make more effective use of network bandwidth.

An embodiment of the present invention provides a flow control method. FIG. 2 shows the implementation process of the method according to this embodiment. However, for ease of description, FIG. 2 only shows the parts related to the embodiment of the present invention. The flow control method includes step S201 and step S202.

The message data packet described in this embodiment is: a data packet sent by the first device to the second device during the process of the first device providing a service to the second device; the data packet carries the data of the service.

A section of memory space is pre-allocated in the memory of the first device, and this section of memory space is used to store credit data. Here, the address that can access this section of memory space is collectively referred to as a preset address, for example, the preset address is the section The physical address of the memory space (that is, the above-mentioned first physical address), for another example, the preset address is a virtual address obtained after performing address mapping on the first physical address.

If the preset address is the first physical address of the segment memory space, address mapping is performed on the preset address to obtain the virtual address of the segment memory space; the processor of the first device can access the segment memory space according to the virtual address .

In addition, determine the identity of the first device, so that the second device can access the first device according to the identity of the first device; the first device, in its communication with the second device, will carry the first physical address and the first The identification of the device notifies the second device; then, the second device can directly access the segment of memory space in the memory of the first device according to the remote memory access instruction pointing to the first physical address.

Step S201, obtaining a first message data packet from a receiving queue, where the message data packet sent by the first device to the second device is stored. Wherein, the first message data packet is a data packet that has been stored in the receiving queue and is being taken from the receiving queue by the application of the service.

Step S202, delete the obtained first message data packet in the receiving queue, if the number of message data packets stored in the receiving queue is less than or equal to the first threshold, send a message to the first device Sending a data import instruction carrying credit data, writing the credit data into a preset address in the internal memory of the first device, so that the first device determines the credit data recorded at the preset address to The upper limit of the number of message data packets sent by the second device, wherein the credit data is the number of message data packets expected to be received by the receiving queue, and the data import instruction belongs to a remote memory access instruction.

In the process of the first device providing a service to the second device, the first device sends a message data packet carrying the service to the second device; for example, using QP to send and receive a message data packet between the first device and the second device, Adding the message data packets expected to be sent from the first device to the second device into the sending queue of the first device, the sending queue is composed of a plurality of sending buffer grids, each of which can cache a data packet to be sent, The message data packets to be sent can be sequentially buffered in the sending buffer grid; according to the credit data stored in the preset address of the internal memory of the first device, the maximum number of the message data packets that can be sent to the second device is determined, according to The determined number controls the sending of message packets in the sending queue.

Correspondingly, the second device has a receiving queue, and the receiving queue is composed of a plurality of receiving buffer cells, and each receiving buffer cell can buffer a received data packet, so when the first device provides services to the second device, The second device sequentially buffers the received message data packets in the receiving buffer grid.

In the second device, the application of the service can take the message data packet from the receiving queue (the message data packet being taken is the first data packet), and the action of taking the first message data packet includes Obtain the first message data packet in the receiving buffer grid, delete the first message data packet stored in the receiving buffer grid; every time the application of this service takes the first message data packet from a receiving buffer grid, the receiving buffer grid can be used again to cache the received message data packets; therefore, in the process of continuously taking message data packets from the receiving buffer grid, the number of message data packets in the receiving queue may change from greater than the first threshold Decrease to be less than or equal to the event of the first threshold, once this event occurs, when the number of message packets in the receiving queue is reduced to less than or equal to the first threshold, data import is triggered Instruction, that is, the number of receiving buffer grids that can be used to receive message packets again increases from less than the second threshold to greater than or equal to the second threshold (the sum of the first threshold and the second threshold is equal to the receiving buffer grid included in the receiving queue The data import instruction is triggered when the data import instruction is triggered; the parameters of the data import instruction include: the first physical address and the identifier of the first device. Therefore, the second device can directly write the credit data to the preset address of the memory of the first device according to the data import instruction, and the first device determines according to the credit data written this time: when the second device updates the Before the credit data, the first device can send at most the number of message data packets to the second device. By analogy, the second device, each time the number of message data packets in the receiving queue decreases from greater than the first threshold to less than or equal to the first threshold, triggers a data import instruction, and passes the data The import command updates the credit data stored in the preset address of the memory of the first device; after the credit data is updated each time, the first device can determine the number of message data packets that can be sent to the second device each time according to the updated credit data . According to the update of the credit data, the number of message data packets that the first device can send to the second device in different time periods is controlled, and the flow control when the first device provides services to the second device is realized.

In addition, when the QP channel is initially established between the first device and the second device, the number of receiving buffer boxes available for receiving data packets in the receiving queue of the second device is less than or equal to the first threshold, step S202 triggers data An import instruction, through which the credit data stored in the preset address of the internal memory of the first device is reset.

In addition, the QP channel established between the first device and the second device is restarted after being interrupted, if the number of receiving buffer grids available for receiving data packets in the receiving queue of the second device is less than or equal to the first threshold, the step S202 triggers a data import instruction, through which the credit data stored in the preset address of the internal memory of the first device is updated.

In this embodiment, there is no limitation on the way the second device sends the data import instruction to the first device; for example, the second device directly sends the data import instruction to the first device; another example, through other forwarding devices (such as routers, switches, etc.) forward the data import instruction sent by the second device to the first device.

In addition, this embodiment does not limit which device in the first device responds to the data import instruction and writes the credit data into the preset address of the memory; for example, the memory resource controller responds to the data import instruction and forwards the data import instruction. Command to the memory controller of the first device, and the memory controller of the first device writes the credit data into the preset address of the memory; for example, the processor of the first device responds to the data import instruction and controls the memory controller Write the credit data to the preset address of the memory.

Optionally, an optional refinement is made to the flow control method provided in FIG. 2 , the sending a data import command carrying credit data to the first device, and writing the credit data into the memory of the first device The default address in , specifically:

The memory resource controller MRC of the second device sends the data import instruction to the MRC of the first device, so that the MRC of the first device forwards the data import instruction to the memory controller of the first device MC, the MC of the first device writes the credit data into the preset address according to the data import instruction.

Specifically, a section of physical addresses is allocated in advance in the second device, and the allocated section of physical addresses is defined as the second physical address. The first device first sends the identifier of the first device and the first physical address to the second device; after the second device obtains the first physical address, it determines the mapping relationship between the first physical address and the second physical address.

Every time the second device expects to update the credit data in the internal memory of the first device, it triggers the data import instruction, and the address pointed to by the triggered data import instruction is the second physical address, where the triggering of the data import instruction The scene is not described in detail, and for details, refer to the description of the foregoing embodiments. When the second device recognizes that the second physical address is not a physical address pointing to the local memory, it replaces the second physical address in the data import instruction with the corresponding first physical address, and sends the data import instruction after the address replacement To the memory resource controller of the first device, when the memory resource controller of the first device recognizes that the first physical address pointed to by the data import instruction is the address of the local memory of the first device, forward it to the data Importing the instruction to the memory controller of the first device; the memory controller of the first device writes the credit data carried in the data importing instruction to the first physical address (corresponding to the preset address) of the memory according to the data importing instruction.

Optionally, the data import instruction includes: an identifier of the first device, and a physical address corresponding to the preset address. Wherein, the physical address corresponding to the preset address is the above-mentioned first physical address. That is, if the preset address is the first physical address of the first device, the preset address is the same as the physical address included in the data import instruction, and if the preset address is mapped to the first physical address virtual address, the preset address has a corresponding relationship with the physical address included in the data import instruction.

Wherein, the identifier of the first device is: when the first device communicates with the second device, an address or an identity used by the second device to find the first device.

Specifically, the second device can designate the first device to execute the data import instruction according to the identifier of the first device pointed to by the data import instruction; further, according to the first physical address corresponding to the preset address, it can point to the second A preset address of the internal memory of a device, and write the credit data to the preset address.

An embodiment of the present invention provides a second device 300. FIG. 3 shows a schematic diagram of the logic structure of the second device 300 provided according to this embodiment. However, for ease of description, FIG. 3 only shows example related part.

The second device 300 includes: a data processing module 301, configured to obtain a first message data packet from a receiving queue, and the receiving queue stores a message data packet sent by the first device to the second device;

The data processing module 301 is further configured to delete the acquired first message data packets in the receiving queue, if the number of message data packets stored in the receiving queue is less than or equal to the first threshold , determining credit data, where the credit data is the number of message packets expected to be received by the receiving queue;

An instruction sending module 302, configured to send a data import instruction carrying credit data to the first device, and write the credit data into a preset address in the memory of the first device, so that the first device according to The credit data recorded at the preset address determines the upper limit of the number of message data packets sent to the second device, and the data import instruction belongs to a remote memory access instruction.

Optionally, the instruction sending module 302 is configured to send a data import instruction carrying credit data to the first device, and write the credit data into a preset address in the memory of the first device, specifically :

The instruction sending module 302 is the memory resource controller MRC of the second device, configured to send the data import instruction to the MRC of the first device, so that the MRC of the first device forwards the data import instruction To the memory controller MC of the first device, the MC of the first device writes the credit data into the preset address according to the data import instruction.

Optionally, the data import instruction includes: an identifier of the first device, and a physical address corresponding to the preset address.

An embodiment of the present invention provides a flow control method. FIG. 4 shows the implementation process of the method according to this embodiment. However, for ease of description, FIG. 4 only shows the parts related to the embodiment of the present invention. The flow control method includes step S401 and step S402.

A section of memory space is pre-allocated in the memory of the first device, and this section of memory space is used to store credit data. Here, the address that can access this section of memory space is collectively referred to as a preset address, for example, the preset address is the section The physical address of the memory space (that is, the above-mentioned first physical address), for another example, the preset address is a virtual address obtained after performing address mapping on the first physical address. ;

If the preset address is the first physical address of the segment memory space, address mapping is performed on the preset address to obtain the virtual address of the segment memory space; the processor of the first device can access the segment memory space according to the virtual address .

In addition, determine the identity of the first device, so that the second device can access the first device according to the identity of the first device; the first device, in its communication with the second device, will carry the first physical address and the first The identifier of the device notifies the second device; then, the second device can directly access the segment of memory space in the memory of the first device according to the remote memory access instruction pointing to the first physical address.

Step S401, receiving a data import command sent by the second device to the first device, writing the credit data to a preset address pointed to by the data import command in the memory, wherein the data import command carries credit data, The data import instruction belongs to the remote memory access instruction;

Step S402, determining the upper limit of the number of message data packets sent to the second device according to the credit data recorded in the preset address.

In the second device, when the application of the service takes the message data packets from the receiving buffer grid, each time the number of message data packets in the receiving queue decreases from greater than the first threshold to less than or equal to the first threshold When a threshold value is reached, the data import instruction is triggered once; that is, the data import instruction is triggered when the number of receiving buffer cells that can be used to receive message data packets again increases from less than the second threshold value to greater than or equal to the second threshold value, wherein, The sum of the first threshold and the second threshold is equal to the total number of receiving buffer boxes included in the receiving queue. According to the credit data carried in the data import instruction, the first device updates the credit data stored in the preset address of the internal memory of the first device; after the credit data is updated each time, the first device can determine the credit data according to the updated credit data each time. The number of message packets sent to the second device. According to the update of the credit data, the number of message data packets that the first device can send to the second device in different time periods is controlled, and the flow control when the first device provides services to the second device is realized.

In addition, when the QP channel is initially established between the first device and the second device, the number of receiving buffer slots available for receiving data packets in the receiving queue of the second device is less than or equal to the first threshold, and the second device triggers Data import instruction: Step S401 receives a data import instruction sent from the second device to the first device, and resets the credit data stored in the preset address of the internal memory of the first device through the data import instruction.

In addition, after the QP channel established between the first device and the second device is interrupted and re-enabled, if the number of receiving buffer boxes available for receiving data packets in the receiving queue of the second device is less than or equal to the first threshold, the second The second device triggers a data import command; Step S401 receives the data import command sent from the second device to the first device, and updates the credit data stored in the preset address of the internal memory of the first device through the data import command.

Optionally, in step S401, a specific optional refinement is made, see FIG. 5, the receiving the data import command sent by the second device to the first device, and writing to the preset address pointed to by the data import command in the memory The credit data specifically includes steps S4011 and S4012.

Step S4011, the MRC of the first device receives the data import command sent from the MRC of the second device, and forwards the data import command to the memory controller MC of the first device;

Step S4012, the MC of the first device writes the credit data to the preset address according to the data import instruction.

Optionally, the data import instruction includes: an identifier of the first device, and a physical address corresponding to the preset address.

An embodiment of the present invention provides a first device 600. FIG. 6 shows a schematic diagram of the logical structure of the first device 600 provided according to this embodiment. However, for ease of description, FIG. 6 only shows example related part.

The first device 600 includes:

An instruction receiving module 601, configured to receive a data import instruction sent by the second device to the first device, wherein the data import instruction carries credit data, and the data import instruction belongs to a remote memory access instruction;

An instruction response module 602, configured to write the credit data into the preset address pointed to by the data import instruction in the memory;

A message data packet limiting module 603, configured to determine the upper limit of the number of message data packets sent to the second device according to the credit data recorded in the preset address.

Optionally, the instruction receiving module 601 is the memory resource controller MRC of the first device, configured to receive the data import instruction sent from the MRC of the second device, and forward the data import instruction to the memory of the first device Controller MC;

The command response module 602 is the MC of the first device, configured to write the credit data to the preset address according to the data import command.

Optionally, the data import instruction includes: an identifier of the first device, and a physical address corresponding to the preset address.

An embodiment of the present invention provides a system 100, the system 100 includes a first device 101 and a second device 102, and the first device 101 and the second device 102 are connected in communication;

The second device 102 is configured to obtain a first message data packet from a receiving queue, and the receiving queue stores the message data packet sent by the first device 101 to the second device 102; wherein, the The first message data packet is a data packet that has been stored in the receiving queue and is being taken away by the application of the service from the receiving queue;

The second device 102 is configured to delete the acquired first message data packets in the receiving queue, if the number of message data packets stored in the receiving queue is less than or equal to the first threshold, Sending a data import command carrying credit data to the first device 101, where the credit data is the number of message data packets expected to be received by the receiving queue, and the data import command belongs to a remote memory access command;

The first device 101 is configured to write the credit data into a preset address in its memory according to the data import instruction, and determine to send the credit data to the second device 102 according to the credit data recorded at the preset address. The upper limit of the number of message packets sent.

Optionally, the first device 101 is configured to write the credit data into a preset address in its memory according to the data import instruction, specifically:

The first device 101 is configured to have the memory resource controller 1011 of the first device 101 receive the data import instruction sent from the memory resource controller 1021 of the second device, and forward the data import instruction to The memory controller 1022 of the first device writes the credit data to the preset address according to the data import instruction.

Optionally, the data import instruction includes: an identifier of the first device, and a physical address corresponding to the preset address.

In the several embodiments provided in this application, it should be understood that the disclosed system, device and method can be implemented in other ways. For example, the device embodiments described above are only illustrative. For example, the division of modules and units is only a logical function division, and there may be other division methods during implementation, such as multiple modules or units or components. May be combined or may be integrated into another system, or some features may be omitted, or not implemented. In another point, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces, and the indirect coupling or communication connection of devices or modules may be in electrical, mechanical or other forms.

The modules described as separate components may or may not be physically separated, and the components as modules may or may not be physical modules, that is, they may be located in one place, or may be distributed to multiple network modules. Part or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional module in each embodiment of the present invention 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 hardware plus software function modules.

The aforementioned integrated modules in the form of software function modules may be stored in a computer-readable storage medium. The above-mentioned software function modules are stored in a storage medium, and include several instructions to enable a computer device (which may be a personal computer, server, or network device, etc.) to execute some steps of the methods described in various embodiments of the present invention. The aforementioned storage media include: removable hard disk, read-only memory (English: Read-Only Memory, ROM for short), random access memory (English: Random Access Memory, RAM for short), magnetic disks or optical discs, etc., which can store programs. The medium of the code.

Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: it can still be Modifications are made to the technical solutions described in the foregoing embodiments, or equivalent replacements are made to some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions depart from the protection scope of the technical solutions of the various embodiments of the present invention.

Claims (15)

1. A flow control method, characterized in that, the flow control method comprises: Obtaining the first message data packet from the receiving queue, where the message data packet sent by the first device to the second device is stored in the receiving queue; Delete the obtained first message data packet in the receiving queue, and if the number of message data packets stored in the receiving queue is less than or equal to the first threshold, send a message containing A data import instruction for credit data, writing the credit data into a preset address in the internal memory of the first device, so that the first device determines to transfer to the second device according to the credit data recorded at the preset address 2. The upper limit of the number of message data packets sent by the device, wherein the credit data is the number of message data packets expected to be received by the receiving queue, and the data import instruction belongs to a remote memory access instruction. 2. The method according to claim 1, characterized in that, sending a data import command carrying credit data to the first device, and writing the credit data into a preset in the internal memory of the first device address, specifically: The memory resource controller MRC of the second device sends the data import instruction to the MRC of the first device, so that the MRC of the first device forwards the data import instruction to the memory controller of the first device MC, the MC of the first device writes the credit data into the preset address according to the data import instruction. 3. The method according to claim 1 or 2, wherein the data import instruction includes: an identifier of the first device and a physical address corresponding to the preset address. 4. A second device, characterized in that the second device comprises: A data processing module, configured to acquire a first message data packet from a receiving queue, where the message data packet sent by the first device to the second device is stored in the receiving queue; The data processing module is further configured to delete the acquired first message data packets in the receiving queue, if the number of message data packets stored in the receiving queue is less than or equal to the first threshold, Determining credit data, where the credit data is the number of message packets expected to be received by the receiving queue; An instruction sending module, configured to send a data import instruction carrying credit data to the first device, and write the credit data into a preset address in the memory of the first device, so that the first device The credit data recorded at the preset address determines the upper limit of the number of message data packets sent to the second device, and the data import instruction belongs to a remote memory access instruction. 5. The second device according to claim 4, wherein the instruction sending module is configured to send a data import instruction carrying credit data to the first device, and write the credit data into the The preset address in the memory of the first device, specifically: The instruction sending module is the memory resource controller MRC of the second device, configured to send the data import instruction to the MRC of the first device, so that the MRC of the first device forwards the data import instruction to The memory controller MC of the first device writes the credit data into the preset address by the MC of the first device according to the data import instruction. 6. The second device according to claim 4 or 5, wherein the data import instruction includes: an identifier of the first device and a physical address corresponding to the preset address. 7. A flow control method, characterized in that, the flow control method comprises: receiving the data import command sent by the second device to the first device, writing credit data to the preset address pointed to by the data import command in the memory, wherein the data import command carries the credit data, and the data Import instructions are remote memory access instructions; The upper limit of the number of message data packets sent to the second device is determined according to the credit data recorded at the preset address. 8. The flow control method according to claim 7, characterized in that, receiving the data import instruction sent by the second device to the first device, writing credit to the preset address pointed to by the data import instruction in the memory Data, specifically: The MRC of the first device receives the data import instruction sent from the MRC of the second device, and forwards the data import instruction to the memory controller MC of the first device; The MC of the first device writes the credit data to the preset address according to the data import instruction. 9. The flow control method according to claim 7 or 8, wherein the data import instruction includes: an identifier of the first device and a physical address corresponding to the preset address. 10. A first device, characterized in that the first device comprises: An instruction receiving module, configured to receive a data import instruction sent by the second device to the first device, wherein the data import instruction carries credit data, and the data import instruction belongs to a remote memory access instruction; An instruction response module, configured to write the credit data to a preset address pointed to by the data import instruction in the memory; A message data packet limiting module, configured to determine the upper limit of the number of message data packets sent to the second device according to the credit data recorded in the preset address. 11. The first device according to claim 10, characterized in that, The instruction receiving module is the memory resource controller MRC of the first device, which is used to receive the data import instruction sent from the MRC of the second device, and forward the data import instruction to the memory controller MC of the first device; The instruction response module is the MC of the first device, configured to write the credit data to the preset address according to the data import instruction. 12. The first device according to claim 10, wherein the data import instruction comprises: an identifier of the first device and a physical address corresponding to the preset address. 13. A system, the system comprising a first device and a second device, the first device and the second device are connected in communication; characterized in that, The second device is configured to obtain a first message data packet from a receiving queue, where the message data packet sent by the first device to the second device is stored in the receiving queue; The second device is configured to delete the acquired first message data packets in the receiving queue, and if the number of message data packets stored in the receiving queue is less than or equal to the first threshold, send The first device sends a data import command carrying credit data, the credit data is the number of message packets expected to be received by the receiving queue, and the data import command belongs to a remote memory access command; The first device is configured to write the credit data into a preset address in its memory according to the data import instruction, and determine the credit data to be sent to the second device according to the credit data recorded at the preset address. The upper limit of the number of message packets. 14. The system according to claim 13, wherein the first device is configured to write the credit data into a preset address in its internal memory according to the data import instruction, specifically: The first device is configured to have the memory resource controller MRC of the first device receive the data import instruction sent from the MRC of the second device, and forward the data import instruction to the first device's A memory controller MC, wherein the MC of the first device writes the credit data into the preset address according to the data import instruction. 15. The method according to claim 13 or 14, wherein the data import instruction includes: an identifier of the first device and a physical address corresponding to the preset address.