High-speed lossless Ethernet forwarding network system based on pi-type structure

Abstract

The invention discloses a high-speed lossless Ethernet forwarding network system based on a pi-type structure, which belongs to the technical field of communication and comprises an input priority queue module, a PFC (power factor correction) frame generation module, a polling arbitration module, a frame analysis module and an output scheduling module. The input priority queue module supports multiple priorities and can define the priorities through the application terminal, the polling arbitration module provides extremely refined priority arbitration distribution, the output scheduling module can load the network and guarantee the stability of the network, the network can be monitored on the premise that the network is not interfered and lossless, and the service life of the network is prolonged. The high configurability realizes the diversity of accessible networks, and is a forwarding system which can perfectly adapt to a large-scale network data center environment and can support efficient flow management, flexible network configuration and strong security assurance.

Description

Technical Field

[0001] The invention belongs to the technical field of communications, and in particular relates to a network system based on high-speed lossless Ethernet forwarding of a π-type structure. Background Art

[0002] Under the technological wave triggered by ChatGPT (Chat Generative Pre-trained Transformer), major Internet companies have set off a wave of research on large language models and increased the deployment of cloud computing, intelligent computing centers and lossless data centers, which has greatly promoted the vigorous development of cloud computing, intelligent computing centers and lossless data centers. In order to effectively support the widespread application and in-depth practice of these cutting-edge technologies, further improve the data throughput requirements of the entire network, and build a more efficient, robust and secure network environment to meet the growing data processing needs, this puts higher requirements on the flexibility, security and reliability of hardware equipment. The network must be able to efficiently and accurately handle massive data traffic and ensure the speed and quality of data transmission.

[0003] However, traditional access devices are usually designed for tree-type network architecture, which works well in small or medium-sized networks and can ensure stable data transmission and smooth network operation. However, in the complex environment of large-scale network data centers, the level-by-level flow control method is difficult to ensure true flow balance and efficient scheduling, resulting in low network efficiency, and network flexibility, data security and reliability cannot be fully guaranteed.

[0004] For cloud computing, intelligent computing centers and lossless data centers, these challenges are undoubtedly unacceptable. The amount of data that needs to be processed in these fields is huge and diverse, which places extremely high demands on network performance, stability and security. Access devices that can perfectly adapt to large-scale network data center environments and advanced technology systems that can support efficient traffic management, flexible network configuration and strong security are needed.

[0005] Ethernet switches are based on Ethernet technology. They forward data from one interface to another through MAC addresses (Media Access Control Addresses). They identify the destination address of data packets and accurately send data packets to target devices to achieve efficient data transmission within the LAN. The switch port may fail, causing the device to be unable to connect and communicate normally, affecting the stability and reliability of the network, causing network congestion, and affecting the normal communication of other devices. At the same time, when the network load is large, the uncertainty of data transmission increases for traditional Ethernet switches, making it difficult to meet application scenarios with high real-time requirements such as industrial control.

[0006] Traditional Ethernet switches usually adopt the store-and-forward method. The problems and defects of the existing technology are: the existing easy-to-implement Ethernet access system has a single means of monitoring and observing the network, and is unable to accurately monitor and observe the network. The simple mirroring function is highly complex for debugging the entire network system, and the observation and adaptation capabilities for specified protocol types are insufficient; and the existing easy-to-implement Ethernet access system has a single way of processing priority traffic in the internal logic, and is unable to fine-tune different types of traffic models. It has poor adaptability to different network environments, poor priority differentiation and scheduling capabilities, and insufficient robustness. Summary of the invention

[0007] In order to solve the above problems existing in the prior art, the present invention provides a network system based on high-speed lossless Ethernet forwarding of a π-type structure.

[0008] The technical problem to be solved by the present invention is achieved through the following technical solutions:

[0009] In a first aspect, the present invention provides a network system based on high-speed lossless Ethernet forwarding of a π-type structure, the system comprising: an input priority queue module, a PFC frame generation module, a polling arbitration module, a frame parsing module and an output scheduling module;

[0010] The input priority queue module is used to perform priority parsing on the received Ethernet data frame and transmit the Ethernet data frame to the corresponding input queue according to the parsed priority;

[0011] The PFC frame generation module is used to generate a PFC frame to control the start and stop of the Ethernet data frame transmission;

[0012] The polling arbitration module is used to perform unfair polling on the Ethernet data frames stored in the input queue according to the polling arbitration algorithm, and send the polling result to the frame parsing module;

[0013] The frame parsing module is used to mark the protocol frames in the polling result according to the protocol frame configuration sent by the application terminal, and output the protocol frames and non-protocol frames in the polling result to the output scheduling module;

[0014] The output scheduling module is used to forward the protocol frame and the non-protocol frame according to the load of the output scheduling module.

[0015] Optionally, the system further comprises: a first device channel, a second device channel, a monitoring device channel and an application terminal channel;

[0016] The first device channel is used to access a network data center to transmit the Ethernet data frame to the input priority queue module;

[0017] The second device channel is used to transmit the protocol frame and the non-protocol frame to achieve forwarding of the protocol frame and the non-protocol frame;

[0018] The monitoring device channel is used to transmit the protocol frame and the non-protocol frame to the monitoring device for monitoring;

[0019] The application terminal channel is used to transmit configuration information input by the user in the application terminal; the configuration information is used to configure the input priority queue module, the PFC frame generation module, the polling arbitration module, the frame parsing module and the output scheduling module.

[0020] Optionally, the first device channel and the second device channel each include: a plurality of physical ports;

[0021] The first device channel is used to evenly distribute the Ethernet data frames received by the multiple physical ports when the load of the first device channel is greater than a load threshold;

[0022] The second device channel is used to evenly distribute the protocol frames and the non-protocol frames received by the multiple physical ports when the load of the second device channel is greater than the load threshold.

[0023] Optionally, the first device channel is used to transmit the protocol frame and the non-protocol frame when the second device channel fails;

[0024] The second device channel is used to transmit the Ethernet data frame when the first device channel fails.

[0025] Optionally, the input priority queue module includes: a COS priority extraction module, a TOS priority extraction module, a frame moving module, and a PFC identification and filtering module;

[0026] The COS priority extraction module is used to mark the COS priority of the Ethernet data frame;

[0027] The TOS priority extraction module is used to mark the TOS priority of the Ethernet data frame;

[0028] The frame moving module is used to perform priority determination according to the COS priority and TOS priority of the Ethernet data frame, and transmit the Ethernet data frame to the corresponding queue according to the priority according to the determination result;

[0029] The PFC identification and filtering module is used to extract the pause time of the priority contained in the PFC frame when identifying the PFC frame in the Ethernet data frame, and turn off the transmission function of the input queue corresponding to the priority until the pause time of the priority ends or a new PFC frame with a pause time of zero for the priority is identified.

[0030] Optionally, the polling arbitration algorithm includes a strict polling arbitration algorithm, a weighted polling arbitration algorithm and a differential weighted polling arbitration algorithm; the polling arbitration algorithm selects a corresponding algorithm for unfair polling according to network congestion conditions; the network congestion conditions are determined according to the input priority queue module and the output scheduling module.

[0031] Optionally, the protocol frame includes frame identification information and corresponding Ethernet data frame; the output scheduling module is also used to send the pause sending time when the input queue cache of different priorities is full to the polling arbitration module according to the frame identification information to determine the network congestion situation.

[0032] Optionally, the output scheduling module is further configured to send an alarm message to the application terminal when the first device channel, the second device channel and the monitoring device channel all fail.

[0033] The technical solution provided by the embodiments of the present invention may have the following beneficial effects:

[0034] In the above technical scheme, the present invention can ensure large-scale high-speed access forwarding of network data centers and provide highly reliable link-level monitoring. The device channel can achieve load balancing, improve the bandwidth of the link, and perform backup at the same time to avoid large-scale network interruptions caused by single point failures. The monitoring device channel can greatly ensure the reliability and definability of network monitoring. The input priority queue module supports multiple priorities and can define priorities through application terminals. The polling arbitration module provides extremely refined arbitration allocation. The output scheduling module can load the network and ensure network stability. The present invention can monitor the network without interfering with the lossless network. The high configurability realizes the diversity of accessible networks. It is a forwarding system that can perfectly adapt to the large-scale network data center environment and can support efficient traffic management, flexible network configuration and strong security.

[0035] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 It is a schematic diagram of a network system based on high-speed lossless Ethernet forwarding of a π-type structure provided by an embodiment of the present invention;

[0037] Figure 2It is a structural schematic diagram of a system channel provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0038] The present invention is further described in detail below with reference to specific embodiments, but the embodiments of the present invention are not limited thereto.

[0039] Figure 1 Schematic diagram of a network system based on high-speed lossless Ethernet forwarding of a π-type structure provided by an embodiment of the present invention, such as Figure 1 As shown, the system includes: an input priority queue module, a PFC frame generation module, a polling arbitration module, a frame parsing module and an output scheduling module;

[0040] The input priority queue module is used to perform priority parsing on the received Ethernet data frames and transmit the Ethernet data frames to the corresponding input queues according to the parsed priorities.

[0041] Optionally, the input priority queue module includes: a COS priority extraction module, a TOS priority extraction module, a frame moving module, and a PFC identification and filtering module;

[0042] COS priority extraction module, used to mark the COS priority of Ethernet data frames;

[0043] TOS priority extraction module, used to mark the TOS priority of Ethernet data frames;

[0044] The frame moving module is used to determine the priority of the Ethernet data frame according to the COS priority and TOS priority, and transmit the Ethernet data frame to the corresponding queue according to the priority according to the result of the determination;

[0045] The PFC identification and filtering module is used to extract the priority pause time contained in the PFC frame when a PFC frame in the Ethernet data frame is identified, and shut down the transmission function of the input queue corresponding to the priority until the priority pause time ends or a new PFC frame with a priority pause time of zero is identified.

[0046] It can be understood that the COS (Class Of Service) priority extraction module marks the COS priorities of all Ethernet data frames according to the seven priorities represented by the three-bit field named "Priority Code Point PCP" in the TCL field in the Layer 2 802.1Q tag header; the TOS (Type Of Service) priority extraction module marks the COS priorities of all Ethernet data frames according to the seven priorities represented by the first three bits "IP Precedence" in the eight-bit TOS field in the Layer 3 IP packet header.

[0047] The frame moving module adopts a mixed priority mode for priority resolution. Specifically, it can make equal priority determination or unequal priority determination between TOS priority and COS priority. For example, if the TOS priority is higher than COS "n" at this time, then P(TOS) = P(COS) + n; for example, when n = 1, the Ethernet data frame with TOS priority of 0 and the Ethernet data frame with COS priority of 1 are regarded as having the same priority; the Ethernet data frame with higher TOS and COS priority is taken as the result, and all Ethernet frames are moved to the input queue in the order of priority according to the determined priority.

[0048] After identifying a PFC (Priority-based Flow Control) frame, the PFC identification and filtering module can extract the pause time of a certain priority in the PFC frame and shut down the sending function of the input queue of the corresponding priority until the pause time of the priority ends or a new PFC frame with a pause time of zero priority is identified; if the pause time has not ended but a pause frame of the same priority is received, the time is reset to the pause time recorded in this frame instead of accumulating the pause time.

[0049] The PFC frame generation module is used to generate PFC frames to control the start and stop of Ethernet data frame transmission.

[0050] It can be understood that the PFC frame generation module is used to actively generate a PFC frame according to the cache congestion of the input queue to notify the upstream device to stop sending frames or to generate a PFC frame with the sending pause time field filled with 0 to start sending frames; the PFC frame can be paused for different priority flows, but a PFC frame can only select one priority for use.

[0051] The polling arbitration module is used to perform unfair polling on the Ethernet data frames stored in the input queue according to the polling arbitration algorithm, and send the polling result to the frame parsing module.

[0052] Optionally, the polling arbitration algorithm includes a strict polling arbitration algorithm, a weighted polling arbitration algorithm and a differential weighted polling arbitration algorithm; the polling arbitration algorithm selects a corresponding algorithm for unfair polling according to network congestion conditions; the network congestion conditions are determined according to the input priority queue module and the output scheduling module.

[0053] The frame parsing module is used to mark the protocol frames in the polling results according to the protocol frame configuration sent by the application terminal, and output the protocol frames and non-protocol frames in the polling results to the output scheduling module; wherein the protocol frames include frame identification information and corresponding Ethernet data frames.

[0054] It can be understood that when the Ethernet data frames in the polling results sent from the arbitration polling module are received, the source MAC address, destination MAC address, frame type, frame length, priority and other information of these Ethernet data frames are used as frame identification information; the frame identification information is stored in the frame information FIFO of the frame parsing module; the Ethernet data frame is stored in the data frame FIFO (First Input First Output) of the frame parsing module; among them, the source MAC address, destination MAC address, frame type, priority and other information can be extracted at the front end of the Ethernet data frame, and the frame length information needs to be synchronized and counted at the end of the Ethernet data frame; after the complete frame identification information is recorded, the Ethernet data frame and the corresponding frame identification information are sent to the output scheduling module.

[0055] The output scheduling module is used to forward protocol frames and non-protocol frames according to the load of the output scheduling module.

[0056] It can be understood that the output scheduling module mainly performs direct forwarding and convergence forwarding on protocol frames and non-protocol frames based on the current perceived load of the output scheduling module. Direct forwarding refers to the forwarding of Ethernet data frames between multiple devices; convergence forwarding refers to the merging of data originally transmitted on multiple physical ports and forwarding through one port when a channel of the forwarding system fails.

[0057] Optionally, the output scheduling module is further used to send the pause sending time of Ethernet data frames of different priorities when the input queue buffer is full to the polling arbitration module according to the frame identification information, so as to determine the network congestion situation.

[0058] It can be understood that the network congestion situation is jointly determined by the input priority queue module and the output scheduling module. The judgment is made based on the empty and full conditions of different input queue FIFOs output by the input priority queue module, the time when different input queues in the output scheduling module suspend sending, and the average frame length of different input queues in the output scheduling module over a period of time. The judgment process is as follows:

[0059] When the default polling arbitration algorithm is the strict polling arbitration algorithm, the polling is carried out according to the priority of the Ethernet data frame; the Ethernet data frame with a higher priority is sent out first, and the Ethernet data frame with a lower priority can only be sent after the Ethernet data frame with a higher priority than itself is completely sent. In this way, long-term congestion will cause network congestion; if the input queue of the lower priority is full and the suspension time exceeds the maximum suspension time line of 1ms, the polling arbitration algorithm will be switched to the weighted polling arbitration algorithm after the Ethernet data frame currently being transmitted is completely output;

[0060] The weighted round-robin arbitration algorithm can solve the long-term congestion of the input queue when there is a large amount of business, but it will cause uneven bandwidth allocation; therefore, if the current polling arbitration algorithm is the weighted round-robin arbitration algorithm, and at this time within a certain time range, the average frame length of the high-priority transmission is much longer than the average frame length of the low-priority transmission, the polling arbitration algorithm will be switched to the differential weighted round-robin arbitration algorithm.

[0061] For example, the maximum pause timeline is configurable by the application terminal, and the default value is 1ms; the default value of the certain time range required to switch from the weighted polling arbitration algorithm to the differential weighted polling arbitration algorithm is 5ms, which is also configurable by the application terminal; the low priority input queue number can be 0, 1, 2; the high priority input queue number can be 5, 6, 7.

[0062] Optionally, refer to Figure 1 , the system further includes: a first device channel, a second device channel, a monitoring device channel and an application terminal channel;

[0063] A first device channel, for accessing a network data center to transmit Ethernet data frames to an input priority queue module;

[0064] The second device channel is used to transmit protocol frames and non-protocol frames to achieve forwarding of protocol frames and non-protocol frames;

[0065] Monitoring device channel, used to transmit protocol frames and non-protocol frames to the monitoring device for monitoring;

[0066] The application terminal channel is used to transmit the configuration information input by the user in the application terminal; the configuration information is used to configure the input priority queue module, the PFC frame generation module, the polling arbitration module, the frame parsing module and the output scheduling module.

[0067] Understandably, Figure 2 is a schematic diagram of the structure of a system channel provided by an embodiment of the present invention, such as Figure 2 As shown, the first device channel, the second device channel and the monitoring device channel are all implemented as Ethernet ports, supporting 1g / 10g / 25g rate adaptation; the application terminal channel is implemented as a PCIE (Peripheral Component Interconnect Express, a high-speed serial computer expansion bus standard) channel. Except for the application terminal channel, each device channel is composed of multiple actual physical ports, and the multiple actual physical ports are functionally used for one device.

[0068] The application terminal channel is used to transmit instructions on the application terminal. The application terminal is used by the user to implement the priority configuration of the input priority queue module, configure the differential priority between the evaluation frame priority TOS priority and COS priority; configure the specific protocol type that controls the Ethernet data frame when the monitoring device channel is forwarding, and implement real-time effectiveness; configure the maximum pause timeline for changing the polling arbitration algorithm, and the time required to switch from the weighted polling arbitration algorithm to the differential weighted polling arbitration algorithm.

[0069] Optionally, the first device channel and the second device channel each include: a plurality of physical ports;

[0070] The first device channel is used to evenly distribute the Ethernet data frames received by the multiple physical ports when the load of the first device channel is greater than the load threshold;

[0071] The second device channel is used to evenly distribute the protocol frames and non-protocol frames received by the multiple physical ports when the load of the second device channel is greater than the load threshold.

[0072] It can be understood that the ports of the first device channel and the second device channel have port aggregation and load balancing functions. The aggregation function refers to aggregating multiple ports together to form an aggregation group under high load conditions to achieve load balancing and traffic sharing of each port under high load conditions, simulating the physical aggregation of ports in the form of logical aggregation; when the link state of one of the ports is lost, the system will automatically distribute the traffic of the link on the port to other ports aggregated with the port.

[0073] The load balancing function is to evenly distribute the traffic to the sub-port links of the aggregation port according to the characteristic value of the source MAC address + destination MAC address in the forwarded Ethernet data frame. By mapping the 11-bit hash value of the source MAC address + destination MAC 32-bit value as the characteristic value, 1 to 29 are distributed to one port and 29+1 to 210 are distributed to another port. The hash function selects ITQ (iterative quantized hashing) to achieve even distribution of traffic and load balancing.

[0074] It is worth mentioning that the first device channel and the second device channel are only used for distinction. The first device channel can be the second device channel, and the second device channel can also be the first device channel.

[0075] Optionally, the first device channel is used to transmit protocol frames and non-protocol frames when the second device channel fails;

[0076] The second device channel is used to transmit Ethernet data frames when a failure occurs in the first device channel.

[0077] Optionally, the output scheduling module is further used to send alarm information to the application terminal when the first device channel, the second device channel and the monitoring device channel all fail.

[0078] It is understandable that, under low load conditions, when multiple physical ports in the first device channel, the second device channel, and the monitoring device channel are connected, half of the ports are enabled by default as master ports to receive Ethernet data frames, and the other half of the ports are enabled as slave ports to transmit protocol frames and non-protocol frames according to the configuration of the application terminal device. Under high load conditions, if the port load transmitting Ethernet data frames exceeds 80% of the rated forwarding rate of the port, port classification is stopped, and load balancing is implemented on the port of the monitoring device channel, giving priority to ensuring link stability and reliability;

[0079] The active / standby port switching function means that if only some of the multiple physical ports in the first device channel, the second device channel and the monitoring device channel have established physical connections, then the Ethernet data frames, protocol frames and non-protocol frames are directly transmitted from the ports that have established physical connections, ensuring that the network can still operate normally in the event of a device failure. According to the Ethernet data frames and frame identification information from the frame parsing module, the active / standby port switching and port classification are completed through the fair polling mechanism of the physical port, which is achieved through fair polling of the actual physical output port. When the system detects that both physical port links are in normal state, the monitoring port forwarding module will subdivide the Ethernet frames into control Ethernet frames and data Ethernet frames according to the protocol message type configured by the terminal application. According to the scheduling instructions of the polling arbitration module, the control Ethernet frames and data Ethernet frames are output from two different physical ports respectively, ensuring that the control frames and data frames specified by the user can be transmitted separately. If the system detects that only one of the multiple physical port links is normal, it will no longer distinguish between control Ethernet frames and data Ethernet frames, but will forward all Ethernet data frames from this normal port. If the system detects that multiple physical port links are abnormal, it will discard these Ethernet data frames and send an alarm message to the terminal application to prompt the user to troubleshoot the problem.

[0080] In one embodiment, the output scheduling module may further include a polling arbitration feedback module, which is mainly used to forward the Ethernet data frame and frame identification information from the frame parsing module to the monitoring port forwarding module and the device port forwarding module at the same time, and pass the frame identification information to the polling arbitration feedback module;

[0081] The polling arbitration feedback module transmits the suspension time of sending when the input queue buffer of different priorities is full and the average sending frame length of these Ethernet data frames within a period of time to the polling arbitration module according to the frame identification information, so that it can select the arbitration polling algorithm.

[0082] After receiving the frame identification information, if the current polling arbitration algorithm is strict priority arbitration, then according to the priority field in the frame identification information, by obtaining the empty and full signals of the low-priority input queue in the input queue, the pause time of the low-priority input queue will be counted; if the low-priority queue remains congested for 1ms and no Ethernet data frames of this priority are sent, the polling arbitration feedback module will provide the arbitration polling module with pause time congestion feedback information and clear the timing information of the priority timer at the same time. Afterwards, the timing will be restarted according to the state after sending the feedback information; if the low-priority input queue has Ethernet data frames sent during the timing, the polling arbitration feedback module will clear the current timing information and restart the timing.

[0083] After receiving the frame identification information, if the current polling arbitration algorithm is a weighted polling arbitration algorithm, the polling arbitration feedback module will start to count the average sending frame lengths of high priority and low priority according to the frame length and priority field in the frame identification information. This is achieved by recording the number of sent frames of each priority level and calculating the average sending frame length each time.

[0084] The polling arbitration feedback module will count the number of frames received within 5ms, and calculate the real-time average frame length information according to the average calculation formula each time; if within 5ms, the average frame length of the low priority is less than one twentieth of the average frame length of the high priority, the polling arbitration feedback module will provide the arbitration polling module with the congestion feedback information of the frame length, and at the same time clear the timing and counting information of the priority, and then re-time and count according to the state after sending the feedback information. On the contrary, if within 5ms, the average frame length of the low priority is greater than or equal to one twentieth of the average frame length of the high priority, the polling arbitration feedback module will only clear the timing and counting information and re-time and count, and will not send feedback information to the arbitration polling module.

[0085] The monitoring port forwarding module is based on the Ethernet data frame and frame identification information from the frame parsing module. This process is completed through the fair polling mechanism of the physical port and is achieved through fair polling of the actual physical output port.

[0086] When the system detects that both physical port links are in normal state, the monitoring port forwarding module will subdivide the Ethernet frame into control Ethernet frame and data Ethernet frame according to the protocol message type configured by the terminal application.

[0087] Subsequently, according to the scheduling instructions of the polling module, the control Ethernet frame and the data Ethernet frame will be output from two different physical ports respectively, ensuring that the control frame and data frame specified by the user can be transmitted separately.

[0088] In the present invention, unless otherwise clearly specified and limited, a first feature being "above" or "below" a second feature may include that the first and second features are in direct contact, or may include that the first and second features are not in direct contact but are in contact through another feature between them. Moreover, a first feature being "above", "above" and "above" a second feature includes that the first feature is directly above and obliquely above the second feature, or simply indicates that the first feature is higher in level than the second feature. A first feature being "below", "below" and "below" a second feature includes that the first feature is directly below and obliquely below the second feature, or simply indicates that the first feature is lower in level than the second feature.

[0089] The above contents are further detailed descriptions of the present invention in combination with specific preferred embodiments, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, several simple deductions or substitutions can be made without departing from the concept of the present invention, which should be regarded as falling within the protection scope of the present invention.

Claims

1. A network system based on high-speed lossless Ethernet forwarding of a π-type structure, characterized in that: The system comprises: an input priority queue module, a PFC frame generation module, a polling arbitration module, a frame parsing module and an output scheduling module; The input priority queue module is used to perform priority parsing on the received Ethernet data frame and transmit the Ethernet data frame to the corresponding input queue according to the parsed priority; The PFC frame generation module is used to generate a PFC frame to control the start and stop of the Ethernet data frame transmission; The polling arbitration module is used to perform unfair polling on the Ethernet data frames stored in the input queue according to the polling arbitration algorithm, and send the polling result to the frame parsing module; The frame parsing module is used to mark the protocol frames in the polling result according to the protocol frame configuration sent by the application terminal, and output the protocol frames and non-protocol frames in the polling result to the output scheduling module; The output scheduling module is used to forward the protocol frame and the non-protocol frame according to the load of the output scheduling module.

2. The network system based on high-speed lossless Ethernet forwarding of π-type structure according to claim 1 is characterized in that: The system further comprises: a first device channel, a second device channel, a monitoring device channel and an application terminal channel; The first device channel is used to access a network data center to transmit the Ethernet data frame to the input priority queue module; The second device channel is used to transmit the protocol frame and the non-protocol frame to achieve forwarding of the protocol frame and the non-protocol frame; The monitoring device channel is used to transmit the protocol frame and the non-protocol frame to the monitoring device for monitoring; The application terminal channel is used to transmit configuration information input by the user in the application terminal; the configuration information is used to configure the input priority queue module, the PFC frame generation module, the polling arbitration module, the frame parsing module and the output scheduling module.

3. The network system based on high-speed lossless Ethernet forwarding of π-type structure according to claim 2 is characterized in that: The first device channel and the second device channel each include: a plurality of physical ports; The first device channel is used to evenly distribute the Ethernet data frames received by the multiple physical ports when the load of the first device channel is greater than a load threshold; The second device channel is used to evenly distribute the protocol frames and the non-protocol frames received by the multiple physical ports when the load of the second device channel is greater than the load threshold.

4. The network system based on high-speed lossless Ethernet forwarding of π-type structure according to claim 3 is characterized in that: The first device channel is used to transmit the protocol frame and the non-protocol frame when the second device channel fails; The second device channel is used to transmit the Ethernet data frame when the first device channel fails.

5. The network system based on high-speed lossless Ethernet forwarding of π-type structure according to claim 1 is characterized in that: The input priority queue module includes: a COS priority extraction module, a TOS priority extraction module, a frame moving module, and a PFC identification and filtering module; The COS priority extraction module is used to mark the COS priority of the Ethernet data frame; The TOS priority extraction module is used to mark the TOS priority of the Ethernet data frame; The frame moving module is used to perform priority determination according to the COS priority and TOS priority of the Ethernet data frame, and transmit the Ethernet data frame to the corresponding queue according to the priority according to the determination result; The PFC identification and filtering module is used to extract the pause time of the priority contained in the PFC frame when identifying the PFC frame in the Ethernet data frame, and turn off the transmission function of the input queue corresponding to the priority until the pause time of the priority ends or a new PFC frame with a pause time of zero for the priority is identified.

6. The network system based on high-speed lossless Ethernet forwarding of π-type structure according to claim 1, characterized in that: The polling arbitration algorithm includes a strict polling arbitration algorithm, a weighted polling arbitration algorithm and a differential weighted polling arbitration algorithm; the polling arbitration algorithm selects a corresponding algorithm for unfair polling according to the network congestion situation; the network congestion situation is determined according to the input priority queue module and the output scheduling module.

7. The network system based on high-speed lossless Ethernet forwarding of π-type structure according to claim 6, characterized in that: The protocol frame includes frame identification information and corresponding Ethernet data frame; the output scheduling module is also used to send the pause sending time when the input queue buffer of different priorities is full to the polling arbitration module according to the frame identification information to determine the network congestion situation.

8. The network system based on high-speed lossless Ethernet forwarding of π-type structure according to claim 4 is characterized in that: The output scheduling module is further used to send alarm information to the application terminal when the first device channel, the second device channel and the monitoring device channel all fail.

Images