Master-slave deployment method for SPTN network controller

Abstract

The invention discloses a master-slave deployment method for SPTN network controllers, and the method comprises the steps: enabling the controllers to discover other controllers through static pre-configuration information, dynamically adjusting the addition and deletion of the controllers, and determining a unique master controller and a unique slave controller based on a priority and election algorithm of selected time. The health state between the master controller and other controllers is determined through heartbeat messages, and when the master controller or the slave controller breaks down, the remaining controllers re-determine new master controllers or slave controllers to replace the new master controllers or slave controllers without being limited by the number of the remaining controllers. The master controller receives a data writing request, the data writing request is copied to the slave controller in a full-amount and incremental combined mode, the slave controller copies the data writing request to all other controllers in the same mode, and meanwhile the copied data size is reduced through serialization and compression. After all the controllers are restarted, the new master controller recovers to the state before restarting according to the local persistent data; the controllers are split and combined, the election problem that a plurality of main controllers exist at the same time is solved by using election time, arbitrary division of fragments is supported, and controller data can finally achieve consistency.

Description

technical field [0001] The invention relates to the technical field of master-slave deployment of SPTN (Software Defined Packet Transport Network) network controllers, in particular to a method for master-slave deployment of SPTN network controllers. Background technique [0002] The SDN controller is the core of the SPTN network and manages all the resources of the entire network. In order to reduce the risk of a single point failure of the controller to the stable operation of the network, it is usually necessary to deploy multiple nodes to form a distributed system, using master-slave or load Equilibrium and other technologies manage network resources and improve network reliability. Distributed system design is the key technology for multi-node deployment of SDN controllers. The functions that need to be implemented include controller discovery, election, replication, switching, restart, splitting and merging, etc. [0003] The CAP theorem is a basic theorem of distribu...

AI Technical Summary

Problems solved by technology

This type of algorithm satisfies the C and P principles in the CAP theorem of distributed systems, and has strong consistency and partition tolerance, but sacrifices certain availability

This type of algorithm requires that the system can run normally when the number of healthy nodes of all the nodes that make up the system is greater than 1 / 2. In some usage scenarios, the requirements cannot be met. For example, when the system is composed of two nodes and 1:1 master-slave deployment, After the master node fails, the slave node cannot continue to work normally, and the node protection effect cannot be achieved

In addition, when submitting data to a system based on this type of algorithm, the leader node must receive responses from more than 1 / 2 of the follower nodes to consider the data submission successful. When the number of system nodes is large, the efficiency of data submission decreases

In addition, this type of algorithm has the randomness of leader election, which is not suitable for scenarios where leader election is controllable, and cannot add or delete multiple nodes to the system at the same time. It is not suitable for SPTN network controller deployment.

Images