Congestion control method, apparatus and device based on credit value

Abstract

The application provides a congestion control method, device and equipment based on credit value, which is used for solving the technical problem that the credit value issued by the CBRC algorithm of UET protocol is poor in effectiveness. The application utilizes the ECN mechanism supported by the switch equipment and the evaluation of the link RTT by the end-side equipment, so that the receiver judges the link quality based on the two factors, and dynamically adjusts the credit value according to the RTT or the packet loss rate when the link quality is poor, to reduce the credit value issued by the receiver to the sender on the link.

Description

Technical Field

[0001] The present invention relates to the technical field of communications and cloud computing, and in particular to a congestion control method, apparatus and device based on credit value. Background Art

[0002] Artificial intelligence and high-performance computing (HPC) bring new challenges to networks, such as the need for greater scale, higher bandwidth density, multipathing, rapid response to congestion, and interdependence on the performance of individual data flows (where tail latency is a key consideration). Existing protocols may address some aspects of these issues, but because they are designed for general-purpose networks, they lack support for features critical to AI and HPC, such as multipathing and easy configuration.

[0003] The Ultra Ethernet Consortium (UEC) is a new organization sponsored by the Linux Foundation and its Joint Development Foundation Initiative. The goal of UEC is to go beyond existing Ethernet capabilities, such as Remote Direct Memory Access (RDMA) and RDMA over ConvergedEthernet (RoCE) technology to provide a high-performance, distributed and lossless transport layer optimized for high-performance computing and artificial intelligence. Compared to some of the problems with existing Ethernet functions mentioned above, the design of the UEC specification will fill these gaps and provide the larger-scale networking required for these workloads. The goal of UEC is to provide a complete communication stack that solves technical problems across multiple protocol layers and provides easy-to-configure and manage features to achieve higher network utilization and lower tail latency, both of which are critical to reducing the completion time of Artificial Intelligence (AI) and High Performance Computing (HPC) jobs. Specifically, the goals that UEC wants to achieve in the future are:

[0004] 1) Open protocol specifications based on existing IP and Ethernet protocols.

[0005] 2) Multi-path, packet-spraying transmission fully utilizes the AI ​​network without causing congestion or head-of-line blocking, and eliminates the need for centralized load balancing algorithms and routing controllers.

[0006] 3) Incast management mechanism to control fan-in on the final link to the target host with minimal packet loss.

[0007] 4) An efficient rate control algorithm allows transmission to quickly increase to line speed without causing performance loss to competing flows.

[0008] 5) API for out-of-order packet delivery, supporting message completion in both out-of-order and in-order modes, maximizing network and application concurrency and minimizing message latency.

[0009] 6) Scalable future networks to support millions of endpoints.

[0010] 7) Performance and optimal network utilization without the need for network and workload specific congestion algorithm parameter tuning.

[0011] 8) Aims to achieve line-speed performance of 800G, 1.6T, and future faster Ethernet on commodity hardware.

[0012] The transmission protocols used in traditional high-performance computing / AI scenarios include TCP, Data Center TCP, and RDMA. However, the congestion control algorithms involved in these protocols can only guide the control loop at the receiving end based on the interactive information generated by the process of switch queues regulating different protocol data flows. This may lead to problems such as flow conflicts (unreasonable load balancing), unreasonable bandwidth allocation, incast congestion (referring to the incast congestion caused by multiple working nodes simultaneously sending message groups to the same aggregation node), and high latency.

[0013] To implement efficient incast management, the Ultra Ethernet Transport (UET) protocol supports the use of a credit-based rate control (CBRC) algorithm (also known as a credit-based congestion control algorithm). This algorithm draws heavily on the principles of the edge-queued datagram service (EQDS), shifting queue-related operations and adjustments from the data center network to the host side. The core idea of ​​CBRC is to distribute credits on the UET host receiving end and establish a reasonable allocation strategy to ensure that bandwidth is fairly shared between different traffic flows while preventing local traffic from being overloaded. This method, which achieves end-to-end congestion control without requiring switch involvement, can effectively optimize bandwidth allocation, reduce latency, and achieve traffic load balancing.

[0014] However, the current CBRC algorithm relies solely on the receiver to publish credit values ​​based on its own receiving capabilities to drive the sender, and is unable to perceive link congestion or overload (overscription). The fixed credit value publishing coefficient used in the credit value sending timer will cause it to blindly publish credit values, further exacerbating the deterioration of link quality. Summary of the Invention

[0015] In view of this, the present invention provides a congestion control method, apparatus and device based on credit value, which are used to solve the technical problem that the validity of credit value published by the CBRC algorithm of the UET protocol is poor.

[0016] According to one aspect of an embodiment of the present invention, the present invention provides a congestion control method based on a credit value, which is applied to a receiver device of a Ultra Ethernet Transport (UET) protocol. The method includes:

[0017] The receiver's packet transport sublayer (PDS) receives the message sent by the first sender and measures the current one-way delay (OTT) from the first sender to the receiver. cur (i);

[0018] The receiver is based on the current one-way delay OTT cur (i) judging the network congestion status based on the ECN carried in the message, and dynamically adjusting the credit value issued to the first sender according to the network congestion status.

[0019] Furthermore, the method for dynamically adjusting the credit value issued to the first sender according to the network congestion condition is:

[0020] When the target one-way delay is OTT obj (i) If the ratio of the target one-way delay to the current one-way delay is greater than 1 and no ECN-marked packets are received, the link is considered good. The receiver increases the credit value MTU increment in the credit value calculation method by the ratio of the target one-way delay to the current one-way delay:

[0021] When the ratio of the target one-way delay to the current one-way delay is less than or equal to 1 and no ECN-marked packets are received, packets are considered to be backlogged on the link. The receiver reduces the credit value MTU increment in the credit value calculation method by the ratio of the target one-way delay to the current one-way delay:

[0022] If the ratio of the target one-way delay to the current one-way delay is less than or equal to 1 and an ECN-marked packet is received, the link is considered severely congested. The receiver reduces the credit value MTU increment in the credit value calculation method based on the ratio of the target one-way delay to the current one-way delay and the proportion of ECN packets.

[0023] The target one-way delay is the inherent round-trip delay RTT between the first sender and the receiver. fix (i) The product of the one-way delay OTT gain coefficient α and divided by 2.

[0024] Furthermore, while dynamically adjusting the credit value MTU increment, the timing interval of the credit value issuance timer is synchronously modified to the currently calculated credit value issuance quantity.

[0025] Furthermore, when the link is determined to be in good condition or packets are accumulated on the link, the receiver dynamically adjusts the credit value MTU increment in the credit value calculation method according to the ratio of the target one-way delay to the current one-way delay as follows:

[0026] When the first sender is in the active sender queue ASL, the credit value issued by the receiver to the first sender is the minimum of the first value and the second value; the first value is the credit value most recently requested by the sender from the receiver; the second value is the credit value most recently sent by the receiver to the sender, the sum of the ratio of the target one-way delay to the current one-way delay and the maximum transmission unit (MTU);

[0027] When the first sender is in the idle sender queue ISL, the credit value published by the receiver to the first sender is: the credit value most recently sent by the receiver to the sender, the ratio of the target one-way delay to the current one-way delay, and the product of the credit value publication coefficient k and the maximum transmission unit MTU.

[0028] Furthermore, if the link is determined to be severely congested, the receiver reduces the credit value MTU increment in the credit value calculation method based on the ratio of the target one-way delay to the current one-way delay and the proportion of ECN packets as follows:

[0029] When the first sender is in the active sender queue (ASL), the credit value issued by the receiver to the first sender is the minimum of the first and third values. The first value is the credit value most recently requested by the sender from the receiver. The third value is the credit value most recently sent by the receiver to the sender, the sum of the absolute value of the difference between the ratio of the target one-way delay to the current one-way delay and the ECN message ratio, multiplied by the maximum transmission unit (MTU).

[0030] When the first sender is in the idle sender queue (ISL), the credit value issued by the receiver to the first sender is: the credit value the receiver recently sent to the sender, the absolute value of the difference between the ratio of the target one-way delay to the current one-way delay and the ECN packet ratio, and the product of the credit value issuance coefficient k and the maximum transmission unit (MTU).

[0031] The ECN message ratio is the ratio of messages with ECN marking received by the receiver within the most recent inherent round-trip delay of the sender to the total messages received within the inherent round-trip delay.

[0032] Furthermore, the first sender and the receiver use a precise time synchronization protocol to perform clock synchronization.

[0033] According to one aspect of an embodiment of the present invention, the present invention further provides a congestion control device based on a credit value, which is applied to a receiver device of a UET protocol for Ultra Ethernet Transport. The device includes:

[0034] The receiving and measuring module is used to receive the message sent by the first sender at the receiving side packet transmission sublayer PDS and measure the current one-way delay OTT from the first sender to the receiver. cur (i);

[0035] Credit value publishing module, used by the receiver to calculate the OTT value based on the current one-way delay cur (i) judging the network congestion status based on the ECN carried in the message, and dynamically adjusting the credit value issued to the first sender according to the network congestion status.

[0036] Furthermore, the credit value issuing module includes:

[0037] Congestion status judgment unit, used to determine the congestion status based on the current one-way delay OTT cur (i) The network congestion status is judged by the ECN message; when the target one-way delay OTT obj (i) The link is considered to be in good condition when the ratio of the target one-way delay to the current one-way delay is greater than 1 and no ECN-marked message is received; the link is considered to be in a good condition when the ratio of the target one-way delay to the current one-way delay is less than or equal to 1 and no ECN-marked message is received; the link is considered to be severely congested when the ratio of the target one-way delay to the current one-way delay is less than or equal to 1 and an ECN-marked message is received;

[0038] The first adjustment unit is configured to increase the credit value MTU increment in the credit value calculation method according to the ratio of the target one-way delay to the current one-way delay when the link status is good:

[0039] The second adjustment unit is configured to reduce the credit value MTU increment in the credit value calculation method according to the ratio of the target one-way delay to the current one-way delay when packets accumulate in the link:

[0040] The third adjustment unit is configured to reduce the credit value MTU increment in the credit value calculation method according to the ratio of the target one-way delay to the current one-way delay and the proportion of ECN packets when the link is severely congested;

[0041] The target one-way delay is the inherent round-trip delay RTT between the first sender and the receiver. fix (i) The product of the one-way delay OTT gain coefficient α and divided by 2.

[0042] Furthermore, the first adjustment unit, the second adjustment unit and the third adjustment unit are further configured to simultaneously modify the timing interval of the credit value issuance timer to the currently calculated credit value issuance quantity while dynamically adjusting the credit value MTU increment portion.

[0043] Furthermore, the method in which the first adjustment unit and the second adjustment unit dynamically adjust the credit value MTU increment part in the credit value calculation method according to the ratio of the target one-way delay to the current one-way delay is:

[0044] When the first sender is in the active sender queue ASL, the credit value issued to the first sender is the minimum of the first value and the second value; the first value is the credit value the sender most recently requested from the receiver; the second value is the credit value the receiver most recently sent to the sender, the ratio of the target one-way delay to the current one-way delay, and the sum of the maximum transmission unit (MTU).

[0045] When the first sender is in the idle sender queue (ISL), the credit value issued to the first sender is: the credit value most recently sent by the receiver to the sender, the ratio of the target one-way delay to the current one-way delay, the credit value issuance coefficient k, and the maximum transmission unit (MTU).

[0046] The method in which the third adjustment unit reduces the credit value MTU increment in the credit value calculation method according to the ratio of the target one-way delay to the current one-way delay and the proportion of ECN messages is:

[0047] When the first sender is in the active sender queue (ASL), the credit value issued to the first sender is the minimum of the first and third values. The first value is the credit value most recently requested by the sender from the receiver. The third value is the credit value most recently sent by the receiver to the sender, the sum of the absolute value of the difference between the ratio of the target one-way delay to the current one-way delay and the ECN message ratio, multiplied by the maximum transmission unit (MTU).

[0048] When the first sender is in the idle sender queue (ISL), the credit value issued to the first sender is: the credit value most recently sent by the receiver to the sender, the absolute value of the difference between the ratio of the target one-way delay to the current one-way delay and the ECN packet ratio, and the product of the credit value issuance coefficient k and the maximum transmission unit (MTU).

[0049] The ECN message ratio is the ratio of messages with ECN marking received by the receiver within the most recent inherent round-trip delay of the sender to the total messages received within the inherent round-trip delay.

[0050] The device provided by the present invention can be implemented in software, hardware, or a combination of software and hardware. When implemented in a software module, the program code of the software module is loaded into the storage medium of the device, and the program code in the storage medium is read and executed by the processor.

[0051] The present invention utilizes the ECN mechanism supported by switch devices and the evaluation of the link RTT by the end-side device, allowing the receiver to judge the link quality based on these two factors. When the link quality is poor, the credit value is dynamically adjusted according to the RTT or packet loss rate to reduce the credit value issued by the receiver to the sender on such a link. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments of the present invention or the description of the prior art. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings of the embodiments of the present invention.

[0053] Figure 1 This is a schematic diagram of the overall architecture of the UET protocol;

[0054] Figure 2 A flow chart showing the process of controlling the credit value release of the CBRC algorithm in the UET draft;

[0055] Figure 3 A schematic flow chart of the steps of a credit-based congestion control method provided in one embodiment of the present invention;

[0056] Figure 4 A schematic structural diagram of a congestion control device based on credit value provided in one embodiment of the present invention;

[0057] Figure 5 A schematic diagram of the structure of an electronic device for implementing the credit-based congestion control method provided by the present invention is provided in accordance with an embodiment of the present invention. DETAILED DESCRIPTION

[0058] The terms used in the embodiments of the present invention are intended solely to describe specific embodiments and are not intended to limit the embodiments. The singular forms "a," "the," and "the" used in the embodiments of the present invention are intended to include the plural forms, unless the context clearly indicates otherwise. Although the embodiments of the present invention may be described using terms such as first, second, and third, such terms are intended solely to distinguish similar information, entities, or steps, and are not intended to describe a specific order or precedence. For example, without departing from the scope of the embodiments of the present invention, first information may be referred to as second information, and similarly, second information may be referred to as first information. For another example, in some scenarios, first information may refer to a single piece of information or to multiple pieces of information of the same type. Furthermore, the term "if" may be interpreted as "when," "when," or "in response to a determination." "And / or" in the present invention is merely a term used to describe an association between related objects, indicating that three possible relationships exist. For example, "A and / or B" can mean: A exists alone, A and B exist simultaneously, or B exists alone. A and B may be singular or plural. Furthermore, in the description of the present invention, unless otherwise specified, "plurality" refers to two or more than two. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can mean: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or plural.

[0059] The transmission protocols used in traditional high-performance computing / AI scenarios include TCP, DCTCP, and RDMA. However, the congestion control algorithms involved in these protocols can only guide the control loop at the receiving end based on the interactive information generated by the process of switch queues adjusting different protocol data flows. This may lead to problems such as flow conflicts (unreasonable load balancing), unreasonable bandwidth allocation, incast congestion, and high latency.

[0060] To achieve efficient incast management, UET (Ultra Ethernet Transport) supports the Credit Based Rate Control (CBRC) congestion control algorithm. This algorithm draws heavily on the principles of EQDS (edge-queued datagram service), shifting queue-related operations and adjustments from the data center network to the host side. The core concept of CBRC is to distribute credits through the UET host receiving end and set a reasonable allocation policy to ensure that bandwidth is fairly shared among different traffic flows without overloading local traffic. This end-to-end congestion control, achieved without the involvement of switches, effectively optimizes bandwidth allocation, reduces latency, and effectively balances traffic load.

[0061] The following first explains the basic operation process of the CBRC algorithm.

[0062] Figure 1The UET protocol is a general architecture diagram, and the CBRC belongs to the congestion management module in the UET general architecture. The congestion management module is located in the packet delivery sublayer (PDS). The application layer first sends a message (msg) to the semantics sublayer (SES) through libfabrics (a general low-level network API), and the SES splits the received complete message msg into packets and delivers them to the PDS. On the UET sender side, when the packet reaches the PDS and needs to be sent, the PDS packet sending module will inquire whether the congestion management module can send the packet. The CBRC judges according to the current credit value (byte number) of the sender. If the credit value is sufficient to send the current size of the packet, the congestion management module will reply that the packet can be sent to notify the PDS that the packet can be sent. If the credit value is not enough, the PDS is not allowed to send the packet, and at the same time, the PDS sends a smaller request to send (RTS) control packet to the UET receiver to request the receiver to issue a larger credit value to the sender. From this point of view, the key to the design of the CBRC is the process of issuing the credit value of the receiver. The ideal credit value issuing strategy should not only reduce the probability of incast congestion, but also reasonably allocate bandwidth to multiple senders. Incast congestion is a specific type of network congestion phenomenon, which usually occurs in data center networks or distributed computing environments. Its characteristic is that multiple senders (source nodes) send a large amount of data to the same receiver (destination node) at the same time, which causes the network interface or switch buffer of the receiver to be quickly filled, and causes serious network performance problems and packet loss.

[0063] Figure 2 The CBRC algorithm in the UET draft controls the flowchart of the credit value issuing process. The credit value distribution process is as follows:

[0064] The receiver maintains two sets of queues, namely the active sender list (ASL) and the idle sender list (ISL). The ASL queue saves all the senders whose requested credit values have not been satisfied, and the ISL queue saves all the senders whose requested credit values have been satisfied. When the credit value of the sender in the ASL is satisfied, it will be transferred to the ISL; when the sender in the ISL sends a new credit value to the receiver, it will be transferred to the ASL.

[0065] In the UET draft, a credit value sending timer is set to control the issuance of credit values. This timer issues specific credit values ​​to senders in the ASL or ISL at regular intervals, thereby continuously updating the sender's available sending window. The credit value sending timer interval is set as follows:

[0066] Credit value sending timer interval = k*MTU / linkspeed (unit: seconds)

[0067] Where k is the credit value issuance factor, MTU is the maximum transmission unit of the link (usually 1500 bytes), and linkspeed is the line speed of the port on the receiving device.

[0068] Analysis of the above-mentioned timing interval setting method shows that the size of the k value determines the burst or congestion of messages (from the sender) on the receiver-side link. The smaller the k value, the shorter the interval for issuing credit values, and the less likely the receiver-side link is to experience bursts. However, due to the smaller k value, the credit value sending timer will time out shorter, which will frequently trigger the sending of control messages carrying credit values, resulting in additional bandwidth usage. The larger the k value, the more likely it is to cause bursts, but it is less likely to cause additional bandwidth usage. The UET protocol draft stipulates that k is a fixed value and recommends a smaller value (such as k = 4).

[0069] After the k value is determined and the credit value sending timer is started, the timer will be triggered every k*MTU / linkspeed seconds. After the triggering timer first checks whether the ASL is empty.

[0070] If the ASL is not empty, it means that the credit value requested by the sender has not been met. At this time, the receiver will first send credit value to the sender at the head of the ASL. If the sender's request is met after sending the credit value, the sender will be removed from the ASL and placed at the end of the ISL. If it is still not met, it will be placed at the end of the ASL. After completing the above operations, the timer will exit the current process and be reset to the timer interval, waiting for the next trigger.

[0071] If the ASL is empty, it means that the credits requested by all senders at the current moment have been met. At this time, if the ISL is not empty, the receiver will send a credit value of k*MTU to the sender at the head of the ISL. If the credit value sent by the receiver does not exceed the credit value limit (the credit value requested by the sender plus the maximum send window max_cwnd), the sender will be placed back to the end of the ISL queue. If the credit value exceeds the credit value limit, the sender will be removed from the ISL (but will not be placed in the ASL, which is equivalent to being in a free state).

[0072] The CBRC credit value release process in the above-mentioned UET draft relies solely on the receiver to release credit values ​​based on its own receiving capabilities to drive the sender to send packets. The receiver cannot perceive link congestion or overload (overscription). The credit value release timer uses a fixed credit value release coefficient, which easily leads to blind credit value release, further exacerbating the deterioration of link quality.

[0073] To address the defects of the original CBRC rate control algorithm in issuing credit values, the present invention proposes a congestion control solution based on credit values. This solution utilizes the ECN mechanism supported by switching devices and uses the end-side device to evaluate the link RTT. This allows the receiver to judge the link quality based on these two factors. When the link quality is poor, the credit value is dynamically adjusted according to the RTT or packet loss rate to reduce the credit value issued by the receiver to the sender on such a link. This allows for more reasonable control of credit value issuance, avoids link congestion, and optimizes link quality.

[0074] The specific implementation process of the present invention is described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that the step processes shown in the accompanying drawings and embodiments can be executed in a computer system such as a set of computer-executable instructions, and one or more of the exemplified method steps can be singularly or aggregated together and run in the form of a software module or a virtual device in the computer system. Moreover, although the exemplified step processes have time and sequential logic, the present invention is not limited to these steps. In certain cases, these steps can be executed in an order different from that here. Under the guidance of the purpose of the present invention, simple changes or combinations of the execution order of these steps should also fall within the scope of the disclosure of the embodiments of the present invention.

[0075] Figure 3 A schematic flow chart of the steps of a credit-based congestion control method provided by one embodiment of the present invention. Figure 3 The example method steps are applied to the UET protocol receiver device. UET involves the entire communication stack. The present invention only optimizes the packet transmission sublayer (PDS) of the UET protocol on the network card side. Before executing this method, a UET protocol connection must be established between multiple end-side devices that support the UET protocol. In order to enable the receiver to accurately measure the one-way delay (OTT) between the i-th sender and the receiver, cur (i) The sender and the receiver also need to use a precise network time protocol to synchronize time between terminals, such as the Precision Time Protocol (PTP).

[0076] The credit value issuance method provided by the present invention dynamically adjusts the credit value issued to the sender in the credit value issuance timer based on the one-way delay detected between the sender and the receiver and the presence of ECN information carried by the message. The method includes:

[0077] S301. The receiver PDS receives the message sent by the first sender and measures the current one-way delay OTT from the first sender to the receiver. cur (i);

[0078] After receiving the message delivered by the semantic sublayer SES, the sender PDS sends the message to the receiver's PDS. When sending the message, the sender PDS will carry a sending timestamp in the message.

[0079] Here, i represents any one of the multiple senders that establish a connection relationship with the receiver. Therefore, the first sender in the present invention refers to any one of the multiple senders that establish a connection relationship with the receiver.

[0080] S302. The receiver is based on the current one-way delay OTT cur (i) judging the network congestion status based on the ECN carried in the message, and dynamically adjusting the credit value issued to the first sender according to the network congestion status.

[0081] Currently, most switching devices support the ECN (Explicit Congestion Notification) marking feature. ECN works as follows: When an interacting device on a link detects link congestion, the switching device adds an ECN marker to the IP header of the forwarded message. When the receiver receives the ECN-marked message, it knows that a certain link segment on the link is at risk of congestion.

[0082] Round-trip time (RTT) is also a key signal for detecting link congestion. When a link's RTT suddenly increases from a stable state, it indicates that a certain segment of the link is likely experiencing packet congestion or even packet loss. However, RTT measurement is often limited to the sender, because the sender uses the same clock and can detect the send timestamp of each packet and the corresponding ACK receive timestamp. Simply subtract the send timestamp from the receive timestamp, and then subtract the ACK delay consumed by the receiver to process the packet, to obtain the RTT value.

[0083] In one embodiment of the present invention, the receiving PDS is based on the detected current one-way delay OTT cur (i) and the case where the message carries the explicit congestion notification ECN, the network congestion status is divided into the following three cases:

[0084] [Link status is good]

[0085] The conditions for determining a good link status are: target one-way delay OTT obj (i) Compared with the current one-way delay OTT cur (i) The ratio is greater than 1 and no ECN-marked message is received;

[0086] Among them, the target one-way delay OTT obj The calculation method for (i) is:

[0087]

[0088] RTT in formula (1) fix (i) is the inherent round-trip delay between the i-th sender and receiver, which is usually measured by sending probe messages after the network topology is determined. α is the one-way delay OTT gain coefficient, a system-configurable parameter with a recommended value range of [1.0, 1.2]. The target one-way delay is actually equivalent to the fixed one-way delay adjusted by the gain coefficient. When the ratio of the target one-way delay to the current one-way delay is greater than 1, it means that the currently measured one-way delay is less than the target one-way delay, indicating that the current link status is in good condition and the link is unobstructed.

[0089] When the link status is good, the receiver increases the credit value issued to the sender in a certain proportion as follows:

[0090] S311. When the first sender is in the active sender queue ASL, the credit value issued by the receiver to the first sender is the minimum of the first value and the second value; the first value is the credit value most recently requested by the sender from the receiver; the second value is the credit value most recently sent by the receiver to the sender, and the target one-way delay OTT is obj (i) Compared with the current one-way delay OTT cur The sum of the ratio of (i) and the maximum transmission unit MTU. The formula is as follows:

[0091]

[0092] Among them, sender(i).last_pull on the right side of the equal sign is the credit value that the receiver sent to the sender the most recently, and sender(i).cur_pull on the left side of the equal sign is the credit value that the receiver wants to send to the sender this time. MTU is the maximum transmission unit (usually set to 1500 bytes). OTT cur (i) is the current one-way delay from the i-th sender to the receiver measured in real time. obj(i) is the target one-way delay. sender(i).pull_target is the credit value most recently requested by the sender to the receiver. min is a function that takes the smallest value. The reason for taking the smallest value is that each sender's desired credit value may vary significantly. In this case, the credit value can only be issued according to the receiver's own receiving capacity.

[0093] In the present invention, the MTU increment added to the most recently sent credit value (sender(i).last_pull) in the UET protocol's CBRC-based credit calculation formula is called the credit value MTU increment. Comparing the credit value published by the sender in the ASL in the original UET protocol draft shows that the credit value MTU increment in the original draft was 1 MTU. In the present invention, the credit value MTU increment is dynamically adjusted using the ratio of the target one-way delay to the current one-way delay. Because this ratio is typically greater than 1 when the link is in good condition, the credit value increment is greater than 1 MTU.

[0094] S312. When the first sender is in the idle sender queue ISL, the credit value issued by the receiver to the first sender is: the credit value of the receiver's most recent send to the sender, the target one-way delay OTT obj (i) Compared with the current one-way delay OTT cur The sum of the ratio of (i) and the product of the credit value publishing coefficient k and the maximum transmission unit MTU. The formula is as follows:

[0095]

[0096] k is the credit value issuance coefficient specified in the original UET protocol draft. It determines the maximum amount of data that can be sent by the sender after the receiver issues credit to the sender. The smaller the k value, the less data a single credit value message (issued by the receiver) can trigger (sent by the sender), which reduces the relative pressure on the receiver's link. However, because the k value also determines the timer trigger frequency, a smaller k value will cause the timer to trigger more frequently, meaning that the receiver needs to send credit value messages more frequently than a larger k value. Therefore, the selection of the k value needs to take into account the above considerations. The protocol recommends k = 4.

[0097] Furthermore, in one embodiment of the present invention, when the link status is good, the timing interval of the credit value release timer is dynamically adjusted by synchronously adjusting the ratio of the target one-way delay to the current one-way delay, that is, the timing interval of the credit value release timer specified in the original UET protocol draft is modified to the currently calculated credit value release number, namely sender(i).cur_pull.

[0098] In one embodiment of the present invention, the current one-way delay OTT from the i-th sender to the receiver is measured in real time. cur The measurement method for (i) is as follows: the receiver reads the message sending timestamp from the most recently received message sent by the i-th sender, and subtracts the message sending timestamp from the clock synchronized with the local precise time protocol to obtain the current one-way delay OTT. cur (i).

[0099] Packets accumulate on the link

[0100] The judgment condition for packet accumulation in the link is: target one-way delay OTT obj (i) Compared with the current one-way delay OTT cur The ratio of (i) is less than or equal to 1 and no ECN-marked messages are received.

[0101] When the ratio of the target one-way delay to the current one-way delay is less than or equal to 1, the currently measured one-way delay is greater than or equal to the target one-way delay. The lack of ECN-marked packets indicates that the link's switching equipment has not yet detected severe congestion. In this case, packets may be backlogged in the forwarding device's queue. In this case, the present invention steadily reduces the credit value issued to the sender at a certain ratio, but the reduction cannot be too rapid or excessive, otherwise the sender's sending rate will be drastically reduced.

[0102] In one embodiment of the present invention, when packets accumulate in a link, the receiver reduces the credit value issued to the sender by a certain ratio in the following manner:

[0103] S321. When the first sender is in the active sender queue ASL, the credit value issued by the receiver to the first sender is the minimum of the first value and the second value; the first value is the credit value most recently requested by the sender from the receiver; the second value is the credit value most recently sent by the receiver to the sender, and the target one-way delay OTT is obj (i) Compared with the current one-way delay OTT cur The sum of the ratio of (i) and the product of the maximum transmission unit MTU. That is, the formula used to publish the credit value is still the same as formula (2).

[0104] S322. When the first sender is in the idle sender queue ISL, the credit value issued by the receiver to the first sender is: the credit value of the receiver's most recent send to the sender, the target one-way delay OTT obj (i) Compared with the current one-way delay OTT cur The sum of the ratio of (i) and the product of the credit value publishing coefficient k and the maximum transmission unit MTU. That is, the formula used to publish the credit value is still the same as formula (3).

[0105] Although packets accumulate on the link, the receiver still uses the same credit calculation method as when the link is in good condition. However, because the ratio of the target one-way delay to the current one-way delay is less than 1, the MTU increment of the credit issued must be less than one MTU. This reduces the credit issued to the sender by a certain proportion, thereby reducing the degree of packet accumulation on the link.

[0106] Furthermore, in one embodiment of the present invention, when packets accumulate in a link, the timing interval of the credit value issuance timer is dynamically adjusted by synchronously adjusting the ratio of the target one-way delay to the current one-way delay. That is, the credit value issuance coefficient k specified in the original UET protocol draft is replaced by the product of the ratio of the target one-way delay to the current one-way delay and k.

[0107] [Severe link congestion]

[0108] The criteria for determining severe link congestion are: target one-way delay OTT obj (i) Compared with the current one-way delay OTT cur The ratio of (i) is less than or equal to 1 and the ECN-marked message is received.

[0109] When the ratio of the target one-way delay to the current one-way delay is less than or equal to 1, the currently measured one-way delay is greater than or equal to the target one-way delay. The receipt of an ECN-marked message indicates that the switch in the link has detected severe congestion, potentially leading to packet loss. In this case, the present invention rapidly reduces the credit value issued to the sender by a certain percentage.

[0110] In one embodiment of the present invention, when a link is severely congested, the receiver reduces the credit value issued to the sender as follows:

[0111] S331. When the first sender is in the active sender queue ASL, the credit value issued by the receiver to the first sender is the minimum of the first value and the third value; the first value is the credit value most recently requested by the sender from the receiver; the third value is the credit value most recently sent by the receiver to the sender, and the target one-way delay OTT is obj (i) Compared with the current one-way delay OTT cur The sum of the products of the absolute value of the difference between the ratio of (i) and the ECN message ratio loss_rate(i) and the maximum transmission unit MTU. The formula is as follows:

[0112] sender(i).cur_pull=

[0113]

[0114] Wherein, the ECN packet proportion loss_rate(i) is the proportion of the packets with ECN mark received by the receiver in the last inherent round-trip time RTT fix (i) in the inherent round-trip time RTT (i) to the total received packets, and other parameters and variable meanings are the same as in formula (2).

[0115] By comparing the credit value MTU increment part of the original UET protocol draft, it can be seen that the MTU increment of the credit value MTU increment part in the original protocol draft is 1 MTU, and the ratio of the target one-way delay to the current one-way delay is subtracted by the absolute value of the ECN packet proportion loss_rate(i) to dynamically adjust the credit value MTU increment part. In the case of serious link congestion, the closer the ratio of the target one-way delay to the current one-way delay is to the ECN packet proportion, the more serious the link congestion is, and therefore the credit value MTU increment part of the credit value issued is less. The faster the speed of approaching, the faster the speed of credit value reduction. Through this way, the credit value MTU increment part issued to the sender can be quickly reduced.

[0116] S332. When the first sender is in the idle sender queue ISL, the credit value issued by the receiver to the first sender is: the credit value sent by the receiver to the sender last time, the ratio of the target one-way delay OTT obj (i) to the current one-way delay OTT cur (i), and the absolute value of the difference between the ratio and the ECN packet proportion loss_rate(i) is multiplied by the sum of the credit value issuing coefficient k and the maximum transmission unit MTU. The formula is as follows:

[0117]

[0118] Wherein, the parameters and variable meanings are explained in formula (2) and formula (4).

[0119] Further, in an embodiment of the present application, in the case of serious link congestion, the timing interval of the credit value issuing timer is dynamically adjusted by the ratio of the target one-way delay to the current one-way delay and the ECN packet proportion, that is, the timing interval of the credit value issuing timer in the original UET protocol draft is modified to the credit value issuing quantity sender(i).cur_pull calculated at present.

[0120] Figure 4 A structure schematic diagram of a credit value-based congestion control device provided in an embodiment of the present application is shown. The device 400 is applied to a super Ethernet transmission (UET) protocol receiver device, and the device 400 comprises a receiving and measuring module 410 and a credit value issuing module 420.

[0121] The receiving and measuring module 410 is used to receive the message sent by the first sender at the receiving side packet transmission sublayer PDS and measure the current one-way delay OTT from the first sender to the receiver. cur (i);

[0122] Credit value publishing module 420 is used by the receiver to calculate the credit value based on the current one-way delay OTT cur (i) judging the network congestion status based on the ECN carried in the message, and dynamically adjusting the credit value issued to the first sender according to the network congestion status.

[0123] Furthermore, the credit value issuing module 420 includes:

[0124] Congestion status judgment unit 421 is used to determine the congestion status based on the current one-way delay OTT cur (i) The network congestion status is judged by the ECN message; when the target one-way delay OTT obj (i) The link is considered to be in good condition when the ratio of the target one-way delay to the current one-way delay is greater than 1 and no ECN-marked message is received; the link is considered to be in a good condition when the ratio of the target one-way delay to the current one-way delay is less than or equal to 1 and no ECN-marked message is received; the link is considered to be severely congested when the ratio of the target one-way delay to the current one-way delay is less than or equal to 1 and an ECN-marked message is received;

[0125] The first adjustment unit 422 is configured to increase the credit value MTU increment in the credit value calculation method according to the ratio of the target one-way delay to the current one-way delay when the link status is good:

[0126] The second adjustment unit 423 is configured to reduce the credit value MTU increment in the credit value calculation method according to the ratio of the target one-way delay to the current one-way delay when packets accumulate in the link:

[0127] The third adjustment unit 424 is configured to reduce the credit value MTU increment in the credit value calculation method according to the ratio of the target one-way delay to the current one-way delay and the proportion of ECN packets when the link is severely congested;

[0128] The target one-way delay is the inherent round-trip delay RTT between the first sender and the receiver. fix (i) The product of the one-way delay OTT gain coefficient α and divided by 2.

[0129] Furthermore, the first adjusting unit 422 , the second adjusting unit 423 and the third adjusting unit 424 are further configured to adjust the timing interval of the credit value issuing timer in the same manner while dynamically adjusting the credit value MTU increment part.

[0130] Further, the first adjusting unit 422 and the second adjusting unit 423 dynamically adjust the credit value MTU increment part in the credit value calculation method according to the ratio of the target one-way latency to the current one-way latency as follows:

[0131] When the first sender is in the active sender queue ASL, the credit value issued to the first sender is the minimum value of the first value and the second value; the first value is the credit value requested by the sender to the receiver last time; the second value is the credit value sent by the receiver to the sender last time, the product of the ratio of the target one-way latency to the current one-way latency and the maximum transmission unit MTU; and the sum of the product.

[0132] When the first sender is in the idle sender queue ISL, the credit value issued to the first sender is the credit value sent by the receiver to the sender last time, the product of the ratio of the target one-way latency to the current one-way latency and the credit value issuing coefficient k and the maximum transmission unit MTU; and the sum of the product.

[0133] The third adjusting unit 424 reduces the credit value MTU increment part in the credit value calculation method according to the ratio of the target one-way latency to the current one-way latency and the ECN packet proportion as follows:

[0134] When the first sender is in the active sender queue ASL, the credit value issued to the first sender is the minimum value of the first value and the third value; the first value is the credit value requested by the sender to the receiver last time; the third value is the credit value sent by the receiver to the sender last time, the product of the absolute value of the difference between the ratio of the target one-way latency to the current one-way latency and the ECN packet proportion and the maximum transmission unit MTU; and the sum of the product.

[0135] When the first sender is in the idle sender queue ISL, the credit value issued to the first sender is the credit value sent by the receiver to the sender last time, the product of the absolute value of the difference between the ratio of the target one-way latency to the current one-way latency and the ECN packet proportion and the credit value issuing coefficient k and the maximum transmission unit MTU; and the sum of the product.

[0136] The ECN packet proportion is the proportion of the packets with ECN flag received by the receiver within the last inherent round trip time of the sender to the total packets received within the inherent round trip time.

[0137] Figure 5This is a schematic diagram of the structure of an electronic device for implementing the credit-based congestion control method provided by the present invention, provided in accordance with one embodiment of the present invention. The device 500 includes a processor 510, such as a central processing unit (CPU), a communication bus 520, a communication interface 540, and a memory 530. The processor 510 and the memory 530 can communicate with each other via the communication bus 520. The memory 530 stores a computer program that, when executed by the processor 510, implements the functions of one or more steps of the credit-based congestion control method provided by the present invention.

[0138] Memory refers to a device based on a storage medium used to store computer programs and / or data. It can be either volatile memory (VM, often referred to as RAM) or non-volatile memory (NVM). RAM refers to internal storage that exchanges data directly with the processor. It can read and write data quickly and at any time, serving as a temporary data storage medium for the operating system and other running programs. RAM can be synchronous dynamic random access memory (SDRAM) or dynamic random access memory (DRAM). Non-volatile memory refers to storage that uses persistent storage media, boasting large capacity and the ability to persist data. Examples include storage-class memory (SCM), solid-state drives (SSDs), NAND flash memory, and magnetic disks. SCM is the industry's general term for new storage media between RAM and flash memory. It is a hybrid storage technology that combines the characteristics of persistent storage and RAM, with access speeds slower than DRAM but faster than SSDs.

[0139] The processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic device, or discrete hardware component.

[0140] It should be appreciated that embodiments of the present invention can be implemented or implemented by computer hardware, a combination of hardware and software, or by computer instructions stored in non-transitory (or non-persistent) memory. The methods can be implemented in a computer program using standard programming techniques, including a non-transitory storage medium configured with a computer program, wherein the storage medium configured in this manner causes the computer to operate in a specific and predefined manner. Each program can be implemented in a high-level procedural or object-oriented programming language to communicate with the computer system. However, if desired, the program can be implemented in assembly or machine language. In any case, the language can be a compiled or interpreted language. In addition, the program can be run on a programmed application-specific integrated circuit for this purpose. In addition, the operations of the processes described in the present invention can be performed in any suitable order, unless otherwise indicated by the present invention or otherwise clearly contradicted by the context. The processes described in the present invention (or variations and / or combinations thereof) can be executed under the control of one or more computer systems configured with executable instructions and can be implemented as code (e.g., executable instructions, one or more computer programs, or one or more applications) that is executed collectively on one or more processors, by hardware, or a combination thereof. The computer program includes multiple instructions that can be executed by one or more processors.

[0141] Furthermore, the methods can be implemented in any type of computing platform that is operably connected to a suitable computer, including but not limited to a personal computer, a minicomputer, a mainframe, a workstation, a network or distributed computing environment, a separate or integrated computer platform, or in communication with a charged particle tool or other imaging device, etc. Various aspects of the present invention can be implemented as machine-readable code stored on a non-transitory storage medium or device, whether removable or integrated into a computing platform, such as a hard disk, an optical read and / or write storage medium, RAM, ROM, etc., so that it can be read by a programmable computer, and when the storage medium or device is read by the computer, it can be used to configure and operate the computer to perform the processes described herein. In addition, the machine-readable code, or portions thereof, can be transmitted over a wired or wireless network. When such media includes instructions or programs that implement the steps described above in conjunction with a microprocessor or other data processor, the invention described herein includes these and other different types of non-transitory computer-readable storage media. When programmed according to the methods and techniques of the present invention, the present invention also includes the computer itself.

[0142] The foregoing is merely an embodiment of the present invention and is not intended to limit the present invention. It will be apparent to those skilled in the art that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.

Claims

1. A congestion control method based on credit value, characterized in that: The method is applied to a receiver device of the Ultra Ethernet Transmission (UET) protocol, and includes: The receiver's packet transport sublayer (PDS) receives the message sent by the first sender and measures the current one-way delay (OTT) from the first sender to the receiver. cur (i) The receiver is based on the current one-way delay OTT cur (i) determining the network congestion status based on the ECN information contained in the message, and dynamically adjusting the credit value issued to the first sender based on the network congestion status; When the link is in good condition or packets are accumulated in the link, When the first sender is in the active sender queue ASL, the credit value issued by the receiver to the first sender is the minimum of the first value and the second value; the first value is the credit value most recently requested by the sender from the receiver; the second value is the credit value most recently sent by the receiver to the sender, the sum of the ratio of the target one-way delay to the current one-way delay and the maximum transmission unit (MTU); When the first sender is in the idle sender queue (ISL), the credit value issued by the receiver to the first sender is: the credit value most recently sent by the receiver to the sender, the ratio of the target one-way delay to the current one-way delay, and the product of the credit value issuance coefficient k and the maximum transmission unit (MTU). When the link is judged to be severely congested, When the first sender is in the active sender queue (ASL), the credit value issued by the receiver to the first sender is the minimum of the first and third values. The first value is the credit value most recently requested by the sender from the receiver. The third value is the credit value most recently sent by the receiver to the sender, the sum of the absolute value of the difference between the ratio of the target one-way delay to the current one-way delay and the ECN message ratio, multiplied by the maximum transmission unit (MTU). When the first sender is in the idle sender queue (ISL), the credit value issued by the receiver to the first sender is: the credit value the receiver recently sent to the sender, the absolute value of the difference between the ratio of the target one-way delay to the current one-way delay and the ECN packet ratio, and the product of the credit value issuance coefficient k and the maximum transmission unit (MTU). The ECN message ratio is the ratio of messages with ECN marking received by the receiver within the most recent inherent round-trip delay of the sender to the total messages received within the inherent round-trip delay.

2. The method according to claim 1, characterized in that The method for dynamically adjusting the credit value issued to the first sender according to the network congestion condition is: When the target one-way delay is OTT obj (i) If the ratio of the target one-way delay to the current one-way delay is greater than 1 and no ECN-marked packets are received, the link is considered to be in good condition. The receiver increases the credit value MTU increment in the credit value calculation method by the ratio of the target one-way delay to the current one-way delay: When the ratio of the target one-way delay to the current one-way delay is less than or equal to 1 and no ECN-marked packets are received, packets are considered to be backlogged on the link. The receiver reduces the credit value MTU increment in the credit value calculation method by the ratio of the target one-way delay to the current one-way delay: If the ratio of the target one-way delay to the current one-way delay is less than or equal to 1 and an ECN-marked packet is received, the link is considered severely congested. The receiver reduces the credit value MTU increment in the credit value calculation method based on the ratio of the target one-way delay to the current one-way delay and the proportion of ECN packets. The target one-way delay is the inherent round-trip delay RTT between the first sender and the receiver. fix (i) The product of the one-way delay OTT gain coefficient α and divided by 2.

3. The method according to claim 2, characterized in that While dynamically adjusting the credit value MTU increment, the timing interval of the credit value release timer is simultaneously modified to the currently calculated credit value release quantity.

4. The method according to claim 1, wherein The first sender and the receiver use a precise time synchronization protocol to perform clock synchronization.

5. A congestion control device based on credit value, characterized in that: The device is applied to a receiver device of the Ultra Ethernet Transmission (UET) protocol, and includes: The receiving and measuring module is used to receive the message sent by the first sender at the receiving side packet transmission sublayer PDS and measure the current one-way delay OTT from the first sender to the receiver. cur (i) Credit value publishing module, used by the receiver to calculate the OTT value based on the current one-way delay cur (i) determining the network congestion status based on the ECN information contained in the message, and dynamically adjusting the credit value issued to the first sender based on the network congestion status; The credit value issuing module includes a first adjustment unit, a second adjustment unit and a third adjustment unit: The first adjustment unit and the second adjustment unit are specifically configured to: When the first sender is in the active sender queue ASL, the credit value issued to the first sender is the minimum of the first value and the second value; the first value is the credit value the sender most recently requested from the receiver; the second value is the credit value the receiver most recently sent to the sender, the ratio of the target one-way delay to the current one-way delay, and the sum of the maximum transmission unit (MTU). When the first sender is in the idle sender queue (ISL), the credit value issued to the first sender is: the credit value most recently sent by the receiver to the sender, the ratio of the target one-way delay to the current one-way delay, the credit value issuance coefficient k, and the maximum transmission unit (MTU). The third adjustment unit is specifically configured to: When the first sender is in the active sender queue (ASL), the credit value issued to the first sender is the minimum of the first and third values. The first value is the credit value most recently requested by the sender from the receiver. The third value is the credit value most recently sent by the receiver to the sender, the sum of the absolute value of the difference between the ratio of the target one-way delay to the current one-way delay and the ECN message ratio, multiplied by the maximum transmission unit (MTU). When the first sender is in the idle sender queue (ISL), the credit value issued to the first sender is: the credit value most recently sent by the receiver to the sender, the absolute value of the difference between the ratio of the target one-way delay to the current one-way delay and the ECN packet ratio, and the product of the credit value issuance coefficient k and the maximum transmission unit (MTU). The ECN message ratio is the ratio of messages with ECN marking received by the receiver within the most recent inherent round-trip delay of the sender to the total messages received within the inherent round-trip delay.

6. The device according to claim 5, characterized in that The credit value issuing module includes: Congestion status judgment unit, used to determine the congestion status based on the current one-way delay OTT cur (i) The network congestion status is judged by the ECN message; when the target one-way delay OTT obj (i) When the ratio of the target one-way delay to the current one-way delay is greater than 1 and no ECN-marked packets are received, the link is considered to be in good condition. When the ratio of the target one-way delay to the current one-way delay is less than or equal to 1 and no ECN-marked packets are received, the link is considered to be in a state of packet accumulation. When the ratio of the target one-way delay to the current one-way delay is less than or equal to 1 and an ECN-marked packet is received, the link is considered to be severely congested. The first adjustment unit is configured to increase the credit value MTU increment in the credit value calculation method according to the ratio of the target one-way delay to the current one-way delay when the link status is good: The second adjustment unit is configured to reduce the credit value MTU increment in the credit value calculation method according to the ratio of the target one-way delay to the current one-way delay when packets accumulate in the link: The third adjustment unit is configured to reduce the credit value MTU increment in the credit value calculation method according to the ratio of the target one-way delay to the current one-way delay and the proportion of ECN packets when the link is severely congested; The target one-way delay is the inherent round-trip delay RTT between the first sender and the receiver. fix (i) The product of the one-way delay OTT gain coefficient α and divided by 2.

7. The device according to claim 6, characterized in that The first adjustment unit, the second adjustment unit and the third adjustment unit are further configured to simultaneously modify the timing interval of the credit value issuance timer to the currently calculated credit value issuance quantity while dynamically adjusting the credit value MTU increment portion.

8. An electronic device, characterized in that: It includes a processor, a communication interface, a storage medium and a communication bus, wherein the processor, the communication interface and the storage medium communicate with each other via the communication bus; Storage medium for storing computer programs; A processor, configured to implement the method according to any one of claims 1 to 4 when executing a computer program stored on a storage medium.

Images