A method and system for cross-data center RDMA network congestion control

Abstract

The application provides a cross-data center RDMA network congestion control method and system, which comprises the following steps: writing first data center congestion information into a packet header field of a data packet, sending the data packet to a first interconnection switch, transmitting the data packet to a second interconnection switch, and forwarding the data packet to a second data center after the second interconnection switch removes the congestion information. Receiving a first feedback message generated by the first interconnection switch based on the first data center congestion information, calculating a first congestion rate according to the first feedback message, and preliminarily adjusting a data transmission rate according to the first congestion rate. Receiving a second feedback message generated by the second data center based on the center congestion information, calculating a second congestion rate, and adjusting the data transmission rate again according to the first congestion rate and the second congestion rate. The application can adapt to different network conditions, more quickly adjust network congestion, and guarantee the fairness between flows.

Description

Technical Field

[0001] This invention relates to the field of network transmission control technology, and in particular to a method and system for congestion control in cross-data center RDMA networks. Background Technology

[0002] In recent years, with the increasing demand for data center networks, more and more service providers and cloud network vendors are choosing to deploy their businesses across multiple data center networks. Cross-data center RDMA networks utilize RDMA technology to connect different data centers, providing users with faster and more reliable high-performance network services.

[0003] RDMA is a novel network technology that offers high throughput, low latency, and ultra-low CPU load, and has gradually become a standard technology in data center networks. Deploying RDMA across data center networks requires lossless network support from the underlying network. Existing Priority Flow Control (PFC) is typically used to ensure lossless operation of switch ports. However, using PFC often leads to head-end congestion and PFC pause frame storms. Therefore, deploying an RDMA network usually also requires deploying appropriate congestion control algorithms to avoid triggering PFC.

[0004] Existing congestion control methods include end-to-end congestion control and switch-based congestion control. End-to-end congestion control requires a control loop that traverses long-distance networks across data centers, resulting in excessively long response times at the sending end. Furthermore, in scenarios where long-distance and short-distance traffic compete, different traffic types experience different congestion times, leading to unfair traffic control. Switch-based congestion control requires deploying corresponding algorithms on most switches within the data center, increasing deployment complexity and offering limited congestion control effectiveness. Summary of the Invention

[0005] In view of this, embodiments of the present invention provide a method and system for congestion control in cross-data center RDMA networks to solve the problems of excessively long congestion feedback loops, unfairness in controlling mixed traffic congestion, and inconvenience in deploying congestion control algorithms in the prior art.

[0006] One aspect of the present invention provides a congestion control method for a cross-datacenter RDMA network, the method being executed in a first data center, the method comprising the following steps:

[0007] The congestion information of the first data center is written into the header field of the data packet, the data packet is sent to the first interconnection switch, and the data packet is transmitted to the second interconnection switch through the first interconnection switch based on the optical link, so that the second interconnection switch removes the congestion information of the first data center and forwards it to the second data center;

[0008] The system receives a first feedback message generated by the first interconnection switch, carrying congestion information of the first data center; calculates a first congestion rate based on the first data center congestion information using a preset congestion control algorithm; and initially adjusts the data transmission rate based on the first congestion rate.

[0009] The system receives a second feedback message containing congestion information of the second data center, generated and returned by the second data center, and calculates a second congestion rate based on the congestion information of the second data center using a preset congestion control algorithm; wherein the second feedback message is generated based on the congestion information of the second data center written into the header field of the data packet from the second data center.

[0010] The data transmission rate is adjusted again based on the first congestion rate and the second congestion rate.

[0011] In some embodiments of the present invention, when the first interconnection switch receives a data packet from the second data center and forwards the data packet to the first data center, the number of flying bytes in the data packet forwarded to the first data center is set to be less than a preset data center bandwidth-delay product.

[0012] The first interconnecting switch is equipped with a scheduling table constructed from destination IPs to record the number of forwarded bytes for the same destination IP;

[0013] The first data center is equipped with a byte counter to record the number of bytes received; the number of bytes received is embedded in the ACK message returned to the first interconnection switch so that the first interconnection switch can calculate the number of flying bytes and the remaining window.

[0014] In some embodiments of the present invention, the method further includes:

[0015] After receiving the data packet, the first interconnecting switch uses multiple VOQ queues to separate the received data packets to different destination IPs;

[0016] When the first interconnection switch receives a data packet, it checks whether the destination IP corresponding to the data packet has been allocated a VOQ queue. If it has been allocated, the data packet is pushed into the allocated corresponding VOQ queue. If it has not been allocated, it checks whether the corresponding destination IP has a remaining window. If there is a remaining window, the data packet is forwarded directly. If there is no remaining window, a new VOQ queue is allocated for the data packet.

[0017] The first interconnecting switch updates the number of flying bytes and the remaining window based on the returned ACK message, and the VOQ queue begins transmitting data packets.

[0018] In some embodiments of the present invention, the preset congestion control algorithm is the DCQCN algorithm, the congestion information is the ECN signal, and the first feedback message and the second feedback message are in CNP message format.

[0019] In some embodiments of the present invention, the preset congestion control algorithm is the TIMELY algorithm, the congestion information is the RTT signal, and the first feedback message and the second feedback message are in ACK message format.

[0020] In some embodiments of the present invention, the preset congestion control algorithm is the HPCC algorithm, the congestion information includes port queue information and port rate, and the first feedback message and the second feedback message are in ACK message format.

[0021] In some embodiments of the present invention, the preset congestion control algorithm includes: BFC algorithm or RoCC algorithm.

[0022] Another aspect of the present invention provides a cross-data center RDMA network congestion control system, including a processor, a memory, and a computer program / instructions stored in the memory, wherein the processor is used to execute the computer program / instructions, and when the computer program / instructions are executed, the system implements the steps of the method described above.

[0023] Another aspect of the present invention provides a computer-readable storage medium having a computer program / instructions stored thereon, characterized in that the computer program / instructions, when executed by a processor, implement the steps of the method described above.

[0024] Another aspect of the present invention provides a computer program product, including a computer program / instructions, characterized in that the computer program / instructions, when executed by a processor, implement the steps of the method described above.

[0025] The beneficial effects of the present invention are at least as follows:

[0026] This invention provides a method and system for congestion control in a cross-data center RDMA network. The method includes: writing congestion information from a first data center into the header field of a data packet, sending it to a first interconnection switch, transmitting it to a second interconnection switch, and having the second interconnection switch clean the congestion information before forwarding it to the second data center. It also includes receiving a first feedback message generated by the first interconnection switch based on the congestion information from the first data center, calculating a first congestion rate based on the first feedback message, and initially adjusting the data transmission rate based on the first congestion rate. Finally, it includes receiving a second feedback message generated by the second data center based on its congestion information, calculating a second congestion rate, and further adjusting the data transmission rate by combining the first and second congestion rates. This method controls congestion occurring within each of the two data centers through near-end DCI switches. For congestion occurring within the data centers, it uses the congestion control algorithm already used by the data centers without requiring additional configuration, enabling faster adjustment of network congestion and ensuring fairness between flows.

[0027] Additional advantages, objects, and features of the invention will be set forth in part in the description which follows, and will also become apparent in part to those skilled in the art upon studying the description, or may be learned by practice of the invention. The objects and other advantages of the invention can be realized and obtained by means of the structures specifically pointed out in the description and drawings.

[0028] Those skilled in the art will understand that the objectives and advantages achievable with the present invention are not limited to those specifically described above, and that the above and other objectives achievable with the present invention will become clearer from the following detailed description. Attached Figure Description

[0029] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, are not intended to limit the scope of the invention. In the drawings:

[0030] Figure 1 This is a flowchart of a cross-data center RDMA network congestion control method according to an embodiment of the present invention.

[0031] Figure 2 This is a structural diagram of a cross-data center RDMA network congestion control system according to another embodiment of the present invention. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the embodiments and accompanying drawings. Here, the illustrative embodiments and descriptions of this invention are used to explain the invention, but are not intended to limit the invention.

[0033] It should also be noted that, in order to avoid obscuring the invention with unnecessary details, only the structures and / or processing steps closely related to the solution according to the invention are shown in the accompanying drawings, while other details that are not closely related to the invention are omitted.

[0034] It should be emphasized that the term "including / comprises" as used herein refers to the presence of a feature, element, step, or component, but does not exclude the presence or addition of one or more other features, elements, steps, or components.

[0035] It should also be noted that, unless otherwise specified, the term "connection" in this article can refer not only to a direct connection, but also to an indirect connection involving an intermediary.

[0036] In the following description, embodiments of the invention will be illustrated with reference to the accompanying drawings. In the drawings, the same reference numerals represent the same or similar parts, or the same or similar steps.

[0037] One embodiment of the present invention provides a cross-datacenter RDMA network congestion control method, which is used to perform in a first data center, such as... Figure 1 As shown, the method includes the following steps S101 to S104:

[0038] Step S101: Write the congestion information of the first data center into the header field of the data packet, send the data packet to the first interconnection switch, and transmit the data packet to the second interconnection switch through the first interconnection switch based on the optical link, so that the second interconnection switch can remove the congestion information of the first data center and forward it to the second data center.

[0039] Step S102: Receive the first feedback message generated by the first interconnection switch carrying the congestion information of the first data center, calculate the first congestion rate using a preset congestion control algorithm based on the congestion information of the first data center, and initially adjust the data transmission rate based on the first congestion rate.

[0040] Step S103: Receive the second feedback message carrying congestion information of the second data center, generated and returned by the second data center, and calculate the second congestion rate using a preset congestion control algorithm based on the second data center congestion information. The second feedback message is generated based on the second data center congestion information written into the header field of the data packet from the second data center.

[0041] Step S104: Adjust the data transmission rate again based on the first congestion rate and the second congestion rate.

[0042] In some embodiments of the present invention, when the first interconnection switch receives a data packet from the second data center and forwards the data packet to the first data center, the number of flying bytes in the data packet forwarded to the first data center is set to be less than a preset data center bandwidth-delay product.

[0043] The first interconnecting switch is equipped with a scheduling table built from destination IPs to record the number of bytes forwarded for the same destination IP.

[0044] The first data center is equipped with a byte counter to record the number of bytes received. The number of received bytes is embedded in the ACK message returned to the first interconnection switch, so that the first interconnection switch can calculate the number of bytes flown and the remaining window.

[0045] In this context, "flying bytes" refers to the total number of bytes in a data transmission that have been sent but not yet acknowledged. For example, in the TCP protocol, the sender maintains a sending window to control the number of data packets sent. The flying bytes are the total number of bytes in the sending window that have been sent but not yet acknowledged. The purpose of flying bytes is to help the sender control the data transmission rate to adapt to network congestion and the receiver's processing capacity.

[0046] The bandwidth-delay product (BDP) is the product of the maximum number of bits a network can hold and the transmission delay. It represents the amount of unacknowledged bits that can be stored on a network link at any given time and is used to estimate the maximum amount of data on the network link. By calculating the BDP, the maximum amount of unacknowledged data on the network link can be determined. This helps adjust the TCP window size to fully utilize the link's bandwidth while avoiding the sender sending too much unacknowledged data due to network latency. In practical applications, the BDP is often used to determine the TCP window size, i.e., the maximum amount of unacknowledged data the sender can send.

[0047] In some embodiments of the present invention, the method further includes:

[0048] After receiving the data packet, the first interconnection switch uses multiple VOQ queues to separate the received data packets to different destination IPs.

[0049] When the first interconnecting switch receives a data packet, it checks whether the destination IP corresponding to the data packet has been allocated a VOQ queue. If it has been allocated, the data packet is pushed into the corresponding allocated VOQ queue. If it has not been allocated, it checks whether the corresponding destination IP has a remaining window. If there is a remaining window, the data packet is forwarded directly. If there is no remaining window, a new VOQ queue is allocated for the data packet.

[0050] The first interconnecting switch updates the number of flying bytes and the remaining window based on the returned ACK message, and the VOQ queue begins transmitting data packets.

[0051] The ACK (Acknowledgment) message is a control message in the TCP protocol used to acknowledge received data packets. In TCP, the receiver notifies the sender that it has successfully received the data packet by sending an ACK message. An ACK message typically includes an acknowledgment number, the sequence number of the received data packet, and information such as the possible window size. By sending an ACK message, the sender knows that the data packet has successfully arrived at the receiver and can then continue sending the next data packet.

[0052] Here, "window" refers to the sliding window in the TCP protocol. In this embodiment, it is used to control the rate at which the first interconnection switch forwards data packets to the first data center, so as to adapt to the processing capacity of the first data center and alleviate the queue backlog problem caused by mixed traffic congestion in the first data center.

[0053] In some embodiments of the present invention, the preset congestion control algorithm is the DCQCN algorithm, the congestion information is the ECN signal, and the first feedback message and the second feedback message are in CNP message format.

[0054] Among them, DCQCN (Distributed Congestion Control for Quantum Network) rapidly alleviates congestion problems within data centers in a distributed manner, using ECN signals as congestion information at the switch level. ECN (Explicit Congestion Notification) is a congestion control mechanism for TCP / IP networks. It allows network devices to send explicit notifications about network congestion to both communicating parties, rather than relying on packet loss to implicitly indicate network congestion.

[0055] In this embodiment, the first and second feedback messages are in CNP message format. CNP (Congestion Notification Packet) messages are a message format used in the DCQCN algorithm to notify the sender of network congestion.

[0056] Furthermore, when the preset congestion control algorithm is the DCQCN algorithm, the method of initially adjusting the data transmission rate according to the first congestion rate includes: reducing the queue length of the switch ports in the first data center that are experiencing congestion according to the first congestion rate.

[0057] In some embodiments of the present invention, the preset congestion control algorithm is the TIMELY algorithm, the congestion information is the RTT signal, and the first feedback message and the second feedback message are in ACK message format.

[0058] The RTT (Round-Trip Time) signal represents the time it takes for a data packet to travel from the sender to the receiver and back to the sender.

[0059] The first and second feedback messages are in ACK message format. An ACK (Acknowledgment) message is a control message in the TCP protocol used to acknowledge received data packets. In this embodiment, the first and second feedback messages will use the ACK message format to acknowledge received data packets and convey congestion information.

[0060] In some embodiments of the present invention, the preset congestion control algorithm is the HPCC algorithm, the congestion information includes port queue information and port rate, and the first feedback message and the second feedback message are in ACK message format.

[0061] The HPCC algorithm uses INT technology to add information such as the port queue and port rate of each switch to the packet header as it passes through. When the packet is sent back to the sending end, the sending end calculates the link utilization based on the link information carried in the packet and adjusts the sending rate accordingly.

[0062] In some embodiments of the present invention, the preset congestion control algorithm further includes: BFC algorithm or RoCC algorithm.

[0063] The BFC (Block Formatting Frame) algorithm controls the network hop-by-hop. Within the data center, switches control packet transmission between themselves using credit frames (currently available transmission credit frames in the queue). Once a packet enters a switch, it is queued. Only after receiving a credit frame from a downstream switch is the corresponding packet sent out, making network traffic more predictable.

[0064] The RoCC algorithm controls the queue length at the switch port by using a PID (Proportional-Integral-Derivative) control algorithm at the switch and feeding back the rate allocation message to the sender in advance.

[0065] Another embodiment of the present invention proposes a congestion control method and system for cross-data center RDMA networks, the overall structure of which is as follows: Figure 2As shown, it includes a first data center, a DCI switch (first interconnection switch) located near the first data center, a second data center, and a DCI switch (second interconnection switch) located near the second data center. The method is divided into three parts, as follows: Figure 2 The three loops are shown: the near-source control loop, the end-to-end control loop, and the near-destination control loop. This method can be summarized as follows:

[0066] The near-source control loop controls congestion occurring in the upstream data center by transmitting near-source feedback from the sending DCI switch (first interconnection switch) to the upstream data center (first data center). The end-to-end control loop controls mixed traffic congestion in the downstream data center based on notification messages generated at the receiving end that contain congestion information from the downstream data center (second data center). The near-destination control loop controls congestion occurring in the downstream data center by employing a forwarding scheduling mechanism at the receiving DCI switch (second interconnection switch). Overall, this method controls congestion occurring in the two data centers separately through the DCI switches at both ends (i.e., the near ends of the two data centers). For congestion occurring within the data center itself, the congestion control algorithm originally used by the data center is employed.

[0067] The near-source flow control algorithm aims to quickly alleviate mixed traffic congestion within upstream data centers. When a data packet is sent from the sender in the upstream data center to the receiver in the downstream data center, it is first forwarded through multiple switches in the upstream data center, with congestion information (such as ECN tags and INT information) recorded in the packet header during the forwarding process. Then, the data packet reaches the upstream DCI switch, which generates a corresponding near-source feedback message based on the congestion information in the packet and sends it back to the sender. To ensure traffic fairness within the data center, the near-source feedback message format generated by the upstream DCI switch is consistent with the notification message format used by the congestion control algorithm within the data center. For example, if the data center uses the DCQCN algorithm, the near-source feedback message also uses the CNP message format. If the data center uses the TIMELY algorithm or the HPCC algorithm, the near-source feedback message uses the ACK message format. After receiving the feedback, the sender calculates the congestion rate according to the deployed congestion control algorithm. For example, if the data center uses the DCQCN algorithm, the sender responds to the received CNP message according to the rate update algorithm in the DCQCN algorithm. If the data center uses the TIMELY algorithm, then the sending end adjusts its rate according to the received near-source feedback messages, following the TIMELY algorithm.

[0068] The purpose of end-to-end flow control algorithms is to alleviate congestion caused by mixed traffic within downstream data centers. Data packets, after being forwarded through the upstream data center, are sent to the downstream data center. They first reach the downstream DCI switch, which clears the congestion information recorded in the upstream data center. Subsequently, the packets are forwarded through multiple switches within the downstream data center, with congestion information recorded in the packet header during this process. Finally, the receiving end generates a notification message based on the downstream data center congestion information recorded in the data packet and sends it back to the sending end. The message format remains the same as the congestion notification message format used within the data center. Upon receiving the end-to-end feedback, the sending end also calculates the congestion rate based on the deployed congestion control algorithm. Finally, the sending end combines these two congestion rates to calculate the final transmission rate.

[0069] To further shorten the control loop and alleviate queue congestion caused by mixed traffic in the downstream data center, this method limits the forwarding of packets arriving at the downstream DCI switch. The algorithm is window-based, limiting the number of flight bytes between the downstream DCI switch and the receiver to below the bandwidth-delay product within the data center. The method maintains a scheduling table on the downstream DCI switch, aggregated by destination IP, recording the number of forwarded bytes to the same destination IP. The receiver also maintains a byte counter, recording the number of bytes received over long distances, and embeds this counter value in the ACK sent to the downstream DCI switch. The downstream DCI switch uses this information to calculate the number of flight bytes and the remaining window size.

[0070] To avoid header congestion issues in downstream DCI switches, this method uses dynamically allocated VOQ queues on the downstream DCI switches to separate traffic from different destination IPs. When a downstream DCI switch receives a data packet, it first checks whether a VOQ has already been allocated to the destination IP of the packet. If it has, it indicates that mixed traffic congestion has occurred in the downstream data center, and the data packet is pushed into the corresponding VOQ. If not, it checks whether there is a remaining window on the port. If there is a remaining window, it forwards the packet directly; if there is no remaining window, it allocates a new VOQ queue. After waiting for an ACK confirmation that the receiving end has received enough bytes to make the sending window available, the VOQ begins transmitting data packets.

[0071] In summary, this invention provides a method and system for congestion control in a cross-data center RDMA network. The method includes: writing congestion information from a first data center into the header field of a data packet, sending it to a first interconnection switch, transmitting it to a second interconnection switch, and having the second interconnection switch remove the congestion information before forwarding it to the second data center. It also includes receiving a first feedback message generated by the first interconnection switch based on the congestion information from the first data center, calculating a first congestion rate based on the first feedback message, and initially adjusting the data transmission rate based on the first congestion rate. Finally, it includes receiving a second feedback message generated by the second data center based on its congestion information, calculating a second congestion rate, and further adjusting the data transmission rate based on the first and second congestion rates. This method controls congestion occurring in both data centers by using near-end DCI switches at both ends of the data centers, adapting to complex network environments, adjusting network congestion more quickly, and ensuring fairness between flows.

[0072] Corresponding to the above method, the present invention also provides a system comprising a computer device including a processor and a memory, wherein the memory stores computer instructions, and the processor is configured to execute the computer instructions stored in the memory, wherein when the computer instructions are executed by the processor, the system implements the steps of the method as described above.

[0073] This invention also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the aforementioned edge computing server deployment method. The computer-readable storage medium can be a tangible storage medium, such as random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, floppy disks, hard disks, removable storage disks, CD-ROMs, or any other form of storage medium known in the art.

[0074] Those skilled in the art will understand that the exemplary components, systems, and methods described in conjunction with the embodiments disclosed herein can be implemented in hardware, software, or a combination of both. Whether implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this invention. When implemented in hardware, it can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this invention are programs or code segments used to perform the desired tasks. The programs or code segments can be stored in a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried in a carrier wave.

[0075] It should be clarified that the present invention is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of the present invention.

[0076] In this invention, features described and / or illustrated for one embodiment may be used in the same or similar manner in one or more other embodiments, and / or combined with or in place of features of other embodiments.

[0077] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, various modifications and variations of the embodiments of the present invention are possible. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A congestion control method for cross-data center RDMA networks, characterized in that, The method is used to execute in a first data center, and the method includes the following steps: The congestion information of the first data center is written into the header field of the data packet, the data packet is sent to the first interconnection switch, and the data packet is transmitted to the second interconnection switch through the first interconnection switch based on the optical link, so that the second interconnection switch removes the congestion information of the first data center and forwards it to the second data center; The system receives a first feedback message generated by the first interconnection switch, carrying congestion information of the first data center; calculates a first congestion rate based on the first data center congestion information using a preset congestion control algorithm; and initially adjusts the data transmission rate based on the first congestion rate. The system receives a second feedback message containing congestion information of the second data center, generated and returned by the second data center, and calculates a second congestion rate based on the congestion information of the second data center using a preset congestion control algorithm; wherein the second feedback message is generated based on the congestion information of the second data center written into the header field of the data packet from the second data center. The data transmission rate is adjusted again based on the first congestion rate and the second congestion rate; When the first interconnect switch receives a data packet from the second data center and forwards the data packet to the first data center, the number of flying bytes in the data packet forwarded to the first data center is set to be less than a preset bandwidth-delay product within the data center; the first interconnect switch is equipped with a scheduling table constructed from destination IPs to record the number of forwarded bytes for the same destination IP; the first data center is equipped with a byte counter to record the number of received bytes; the number of received bytes is embedded in the ACK packet returned to the first interconnect switch so that the first interconnect switch can calculate the number of flying bytes and the remaining window; After receiving a data packet, the first interconnecting switch uses multiple VOQ queues to separate the received data packets to different destination IPs. When the first interconnecting switch receives a data packet, it checks whether the destination IP corresponding to the data packet has been allocated a VOQ queue. If it has been allocated, the data packet is pushed into the allocated corresponding VOQ queue. If it has not been allocated, it checks whether the corresponding destination IP has a remaining window. If there is a remaining window, the data packet is forwarded directly. If there is no remaining window, a new VOQ queue is allocated for the data packet. The first interconnecting switch updates the number of flight bytes and the remaining window according to the returned ACK message, and the VOQ queue begins to transmit data packets.

2. The cross-data center RDMA network congestion control method according to claim 1, characterized in that, The preset congestion control algorithm is the DCQCN algorithm, the congestion information is the ECN signal, and the first feedback message and the second feedback message are in CNP message format.

3. The cross-data center RDMA network congestion control method according to claim 1, characterized in that, The preset congestion control algorithm is the TIMELY algorithm, the congestion information is the RTT signal, and the first feedback message and the second feedback message are in ACK message format.

4. The cross-data center RDMA network congestion control method according to claim 1, characterized in that, The preset congestion control algorithm is the HPCC algorithm, the congestion information includes port queue information and port rate, and the first feedback message and the second feedback message are in ACK message format.

5. The cross-data center RDMA network congestion control method according to claim 1, characterized in that, The preset congestion control algorithm includes either the BFC algorithm or the RoCC algorithm.

6. A cross-data center RDMA network congestion control system, comprising a processor, a memory, and computer programs / instructions stored in the memory, characterized in that, The processor is used to execute the computer program / instructions, and when the computer program / instructions are executed, the system implements the steps of the method as described in any one of claims 1 to 5.

7. A computer-readable storage medium having a computer program / instructions stored thereon, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method as described in any one of claims 1 to 5.

8. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method according to any one of claims 1 to 5.

Images