Method and device for tcp flow backup and smoothing in tcp hot standby architecture

Abstract

The invention discloses a method and device for backing up and smoothing TCP streams in a TCP hot standby architecture, relating to the field of data communication, including that when sending and receiving, the first active component close to the source first caches the stream, and then backs up the stream to the first standby component after success And update the length of the sending and receiving stream and then transmit it to the second active component; the second active component buffers the stream after receiving it, and after it is successfully backed up to the second standby component, it updates the length of the sending and receiving stream and transmits it to the next active component or remote End device; after receiving the confirmation of the next component based on the length of the sending and receiving flow or the confirmation of the remote device based on the TCP sequence number, the cache flow of the active and standby components can be deleted; after the standby is upgraded to the master, the protocol component is implemented based on the length of the sending and receiving flow Smoothing of buffered streams between and TCP components. The method and device for TCP stream backup and smoothing in the TCP hot backup framework of the present invention improve the recovery probability of the TCP-based MPLS / routing protocol after the backup is upgraded to the master, and improve the usability of the device.

Description

technical field [0001] The invention relates to the field of data communication, in particular to a method and device for backing up and smoothing TCP streams in a TCP hot-backup architecture. Background technique [0002] NSR (None Stop Routing, non-stop routing) refers to the configuration of two main control disks in the network communication equipment, one is the main control disk (hereinafter referred to as the active disk), and the other is the standby main control disk (hereinafter referred to as standby disk), in the event of a physical / software failure of the active disk or an active-standby switchover command, the neighbor does not perceive the failure and does not need to provide GR assistance, and the standby disk takes over the work of the active disk to restore the protocol neighbor status and routing labels and other business data, Technology that does not cause protocol and business interruption. [0003] Based on the MPLS / routing protocol component of the T...

AI Technical Summary

Problems solved by technology

These implementation methods can only partially meet the scenario where there is no TCP flow sent and received between protocol components during a fault, and large-scale router protocol components may send and receive TCP flows all the time, and failures may occur in any link of sending and receiving TCP flows. Therefore, the above The implementation process of TCP hot backup is difficult to solve the following problems:

[0008] (1) Reliable backup of the TCP stream cannot be guaranteed before backup to the master, and since the backup buffer cannot be infinitely large, the deletion time of the backup TCP stream has a greater impact;

[0009] (2) After backup, although both the protocol component and the TCP component have backup TCP streams, in the sending direction, the protocol component cannot know which TCP streams in its sending buffer have been sent to the TCP component, and which TCP streams have not been sent To the TCP component; in the receiving direction, the protocol component has no way of knowing which TCP streams have been received from the TCP component and which TCP streams have not been received

[0010] Therefore, it is difficult for the aforementioned implementation to guarantee the reliable transmission of TCP streams sent and received between the protocol components and TCP, so that the TCP streams sent and received between the protocol components and the protocol components of the remote device will not lose or repeat some bytes, so that it is difficult to cause NSR failure

Images