A dual decoupling method for FC switch input / output ports

Abstract

A dual decoupling method for input and output ports of an FC switch is proposed. This method employs a dual decoupling strategy based on both input and output ports, and a technical solution that independently configures the decoupling time on the output EGR port. A shorter decoupling time is set for the EGR port, which handles a large amount of information. If a GNT response is not received from the ING port after the time expires, the receiving port is decoupled from the current broadcast or multicast. This results in reduced information processing pressure on the receiving port, released information transmission pressure on the sending port, facilitates efficient information transmission, balances the overall traffic of the switch, and improves the information processing capacity of the switch. It also solves the technical problems that input port decoupling cannot balance switch traffic or differentiate between different situations.

Description

Technical Field

[0001] This invention belongs to the field of FC switch technology, specifically relating to a data transmission congestion handling technology. Background Technology

[0002] In FC switch operation, multicast and broadcast frames are frequently processed frame types. Compared to unicast, multicast and broadcast involve multiple destination ports. The different states of the destination ports result in uncertain processing times for broadcast and multicast frames. When destination ports experience congestion or other abnormal conditions, the switch's processing mechanism directly determines its performance and operational status. Current FC switches employ the following strategies to handle multicast and broadcast destination port congestion:

[0003] Based on the timeout policy of the input port, when a multicast or broadcast frame enters the switch's input port (ing port), the timer starts. If a receive response (gnt) is not received from the egr port after a certain time threshold T, the multicast or broadcast frame ends. In this policy, if the switch's ing port cannot receive the receive response (gnt) from the egr port in time, the input port will remain blocked for time T.

[0004] If a timeout occurs, a retransmission request is sent. When a multicast or broadcast frame enters the switching port, the timer starts. If a response (gnt) is not received from the egr port after a certain time threshold T, a retransmission request (req) is sent to the egr port until the egr port responds. In this strategy, the blocked egr port continues to receive transmission requests, increasing the burden on the egr port to process data.

[0005] No timeout is performed; the process waits indefinitely. In this strategy, if the egr port is blocked, the ing port will remain occupied and unable to process new data.

[0006] None of the above three processing methods can achieve balanced processing based on information priority within the FC switch. When the egr port is congested, this port will prioritize processing messages with higher priority. If the egr port cannot be decoupled from the multicast or broadcast with lower priority within a certain period of time, the ing port will remain in a waiting state, causing the ing port to be unable to process new frames and become blocked.

[0007] Patent application number 201410821815.6 buffers broadcast and multicast data, starts a timer, and deletes the buffered data if no retransmission request is received within the timeout period, assuming that all UEs have correctly received the data. Patent application number 201910451955.1 sends data to the device in broadcast or multicast form, and sends data to the device that has not responded within the timeout threshold in unicast form. Both solutions are timeout decoupling schemes based on input ports, which cannot effectively balance switch traffic or finely distinguish and handle different port situations separately. Summary of the Invention

[0008] To address the technical issues of input port decoupling failing to balance switch traffic and differentiate between different scenarios, a dual decoupling strategy based on input and output ports was adopted. This involved independently configuring decoupling time on each EGR port. Ports handling high data volumes were assigned shorter decoupling times; if no response was received by the specified time, the decoupling time would be used to resolve the issue. gnt When the receiving port is connected to the ing port, the receiving port is decoupled from the current broadcast or multicast, resulting in the technical effects of reducing the information processing pressure on the receiving port, releasing the information transmission pressure on the sending port, facilitating efficient information transmission, balancing the overall traffic of the switch, and improving the information processing capability of the switch.

[0009] Set the timeout decoupling time Ting for the ing port. The timer starts from when the ing port initiates a transmission req request. After time Ting, the data is sent for the first time to the multicast or broadcast destination port egr that received the reply gnt within time Ting. The req request is then re-initiated to the egr port that did not receive the reply gnt after time Ting. At the same time, the timer is cleared and the timeout period is calculated.

[0010] Furthermore, different timeout decoupling times are set for different ing ports, and the timeout decoupling time is adjusted according to the traffic of the ing port.

[0011] Set the timeout decoupling time TegrN for the egr port. The timer starts from when the ing port initiates a transmission req request. If the egr port N does not receive a reply gnt after the time TegrN expires, the egrN port will be decoupled, that is, the egrN port will be removed from the receiving port group. The next time the ing port timeout decoupling time Ting starts, no more req requests will be sent to the decoupled port.

[0012] Furthermore, different timeout decoupling times (TegrN) are set for different egr ports.

[0013] Set the overall timeout decoupling time Ttotal for the ing port. The timer starts from when the ing port initiates a transmission req request. After time Ttotal, the data transmission process ends and no more req requests are sent to the receiving ports that are not decoupled within time Ttotal.

[0014] Furthermore, different overall timeout decoupling times Ttotal are set for different ing ports, and the overall timeout decoupling time Ttotal is adjusted according to the traffic of the ing port.

[0015] Mark an input port as an ingress port (ing port), mark the port that receives a multicast message as a destination port (egr port), i.e., egr1-egrN ports, set the timer for marking the ingress port (ing port) to Tnative, configure the timeout decoupling time of the ingress port to Ting, set the timer for marking the destination ports (egr1-egrN ports) to T1-TN, and configure the timeout decoupling time of the egr1-egrN ports to Tegr1-TegrN.

[0016] Furthermore, the values ​​of Ting and Tegr1-TegrN can be modified separately.

[0017] When the ing port receives a multicast or broadcast message, it establishes a receiving port group for the egr1-egrN ports, initiates a req request to the egr1-egrN ports, and starts timers Tnative and T1-TN.

[0018] When timer Tnative reaches time Ting, the ing port sends data to the egr port that received the response gnt within the current time Ting, removes the port from the receiving port group, re-initiates the req request to the egr port that did not receive the response gnt, and the timer Tnative is reset to zero and starts counting again.

[0019] When timer Tnative reaches time Ting again, the ing port sends data to the egr port that received the reply gnt within this time Ting, removes the port from the receiving port group, re-initiates the req request to the egr port that did not receive the reply gnt, and the timer Tnative is reset to zero and starts counting again.

[0020] The timing continues until the number of members in the receiving port group is 0. Once the multicast processing of the ing port is completed, the next multicast task begins.

[0021] When the timeout decoupling time of the ing port has elapsed for time Ttotal, the number of members in the receiving port group is cleared to zero, the multicast processing of the ing port is determined to be over, and the next multicast task begins.

[0022] If a timeout decoupling time Tegr1-TegrN is reached for the corresponding egr port in timers T1-TN, then that port is removed from the receiving port group. If the port has already received a response gnt before the timeout, that is, the port was no longer in the receiving port group before the timeout occurred, then it is not removed. Attached Figure Description

[0023] Figure 1 This is a flowchart of multicast or broadcast transmission. Figure 2 This is a timing diagram for decoupling multicast or broadcast timeouts. Implementation

[0024] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings.

[0025] The process of an FC switch sending multicast is as follows: Figure 1 As shown, the multicast group members of multicast 1 are port 4, port 5, and port 6. When port 1 receives multicast 1, it is marked as the ingress port (ing port), and the receiving ports 4, 5, and 6 are marked as the destination port (egr port).

[0026] When the ing port receives multicast 1, it sends a transmission request (req) to the egr port. If the egr port can receive the information, it sends a reply (gnt) to the ing port to notify that it can receive the multicast.

[0027] The broadcast transmission is similar, with all ports except the ing port being received. For certain ports among multicast and broadcast members (i.e., some or all of the egr ports among the multicast members), if congestion occurs and the gnt response cannot be received, the strategy of this application is used, such as... Figure 2 As shown, the horizontal axis represents the time axis, and the vertical axis represents the transmission direction of multicast or broadcast and the data transmission diagram.

[0028] Port 0 is used as the ingress port (ing port), and the destination ports (egr ports) of the multicast members are port 1, port 2, port 3, port 4, port 5, and port 6, respectively.

[0029] At time 1, ing port 0 receives a multicast or broadcast, determines the multicast or broadcast member port, and sends a req request to all multicast member egr ports;

[0030] At time 2, port 1 of egr responds with gnt;

[0031] At time 3, ing port 0 replies to egr port 1 with acc, indicating that ing port has received the reply from egr port;

[0032] At time 4, port 2 of egr responds with gnt;

[0033] At time 5, ing port 0 replies to egr port 2 with acc, indicating that ing port has received the reply from egr port;

[0034] At time 7, the first input port timeout decoupling Ting is triggered. The ing port 0 sends data to the egr ports 1 and 2 that have responded with gnt, and resends req to the egr ports 3, 4, 5 and 6 that have not responded with gnt, and then clears the timeout.

[0035] At time 9, egr port 3 responded with gnt;

[0036] At time 10, ing port 0 replies to egr port 3 with acc, indicating that ing port has received the reply from egr port;

[0037] At time 11, the timeout decoupling of Tegr4 on egr port 4 is triggered, and egr port 4 is decoupled from multicast, that is, ing port no longer sends req requests to egr port 4;

[0038] At time 13, the second timeout decoupling of the ing port is triggered. The ing port sends data to the egr port 3, which has responded with gnt, and resends req to egr ports 5 and 6, which have not responded with gnt and are not decoupled, and then clears the timeout.

[0039] At time 15, the timeout decoupling of Tegr5 on egr port 5 is triggered, and egr port 5 is decoupled from multicast, that is, ing port no longer sends req requests to egr port 5;

[0040] At time 18, the overall timeout Ttotal of ing port 0 is triggered, and ing port ends the processing of this multicast.

[0041] If EGR port 4 is a switch cascade port and processes a large amount of information, it will prioritize processing data with higher priority, which may cause data with lower priority to fail to respond to GNT in a timely manner. A shorter decoupling time Tegr can be set for this port to decouple data in a timely manner during multicast or broadcast.

[0042] Setting a shorter wait time for the ing port reduces the number of low-priority message requests on the egr port, allowing higher-priority messages to be processed first.

[0043] If the timeout period TegrM is reached for the Tm corresponding to the multicast member egrM port, the port will be removed from the receiving port group. If the port has responded with gnt before the timeout, meaning that the ing port has already sent data to the egrM port, and the port is no longer in the sending task member library, then the deletion will be ignored.

[0044] The above are embodiments of the present invention and do not limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention are included within the protection scope of the present invention.

Claims

1. A method of double decoupling of an input / output port of an FC switch, characterized in that, Comprising: Marking the input port as ing port, marking the multiple output ports as egr port, i.e. egr1-egrN port; Marking the timer of ing port as Tnative, setting the timeout decoupling time Ting of ing port, marking the timer of egr1-egrN port as T1-TN, and configuring the timeout decoupling time of egr1-egrN port as Tegr1-TegrN respectively; When ing port receives multicast or broadcast message, establishing the receiving port group for egr1-egrN port, initiating req request to egr1-egrN port, starting timer Tnative and timer T1-TN, and using gnt to represent the reply message of egr port after receiving req request sent by ing port; Taking the first transmission req request initiated by ing port as the starting point of timing, after the time Ting of timer Tnative arrives, ing port sends data to the egr port which receives gnt within this time Ting, deletes the egr port from the receiving port group, reinitiates req request to the egr port which does not receive gnt, clears timer Tnative, and restarts timing; After the time Ting of timer Tnative arrives again, ing port sends data to the egr port which receives gnt within this time Ting, deletes the egr port from the receiving port group, reinitiates req request to the egr port which does not receive gnt, clears timer Tnative, and restarts timing; Timer Tnative repeats timing, and ing port repeats sending until there is no egr port in the receiving port group, and it is determined that the processing of ing port to multicast or broadcast message is finished, and the next multicast or broadcast task is started; Taking the first transmission req request initiated by ing port as the starting point of timing, when one of timer T1-TN passes its corresponding timeout decoupling time and its corresponding egr port still does not send gnt, the egr port is decoupled, deleted from the receiving port group, and req request is no longer sent to the decoupled port; Setting the overall timeout decoupling time Ttotal of ing port, taking the first transmission req request initiated by ing port as the starting point of timing, and after the time Ttotal passes, the sending process of multicast or broadcast message is finished, and req request is no longer sent to the receiving port group.

2. The method of double decoupling of FC switch input / output port according to claim 1, wherein, Further comprising: Different ing ports set different timeout decoupling time Ting, and the timeout decoupling time Ting is adjusted according to the traffic of ing port.

3. The method of Claim 1, wherein, Further comprising: Different ing ports set different overall timeout decoupling time Ttotal, and the overall timeout decoupling time Ttotal is adjusted according to the traffic of ing port.

4. The method of Claim 2, wherein, Further comprising: The values of Ting and Tegr1-TegrN are modified respectively.

Images