Research And Development Of Key Technologies For Flexible And Scalable Application Layer Routing

With the rapid development of the Internet industry,various Internet services have appeared in large numbers.The Internet used today has been generated after traffic planning by Internet operators.Under normal circumstances,network traffic will be routed to the shortest path calculated by the IGP for transmission,with sufficient bandwidth on the shortest path.In order to ensure that the Internet can provide services stably,the Internet topology is more stable than the host topology,and the routing table is not updated frequently.In normal state,network does not require traffic engineering to avoid congestion.However,in some extreme cases,such as failure of some important nodes,or traffic exceeding the planned bandwidth,network congestion will appear.At this time,the path provided by the routing table of the network layer may not provide normal services,resulting in communication failure.In the case of such communication failure,it may be necessary to use some special methods,instead of directly using the path provided by the network layer routing table,but to direct the network traffic to the currently available path,so as to ensure that when an error occurs,network applications are still able to communicate properly.The existing application layer routing technology provides a feasible solution to solve such problems,but has different degrees of bottlenecks in terms of scalability and reliability.This paper finds the shortcomings of the existing application layer connection proxy technology by studying the existing application layer routing related technology and the previous version of the Application Layer Connection Agent(ALCA)technology:when the number of deployed nodes is large,there will be node clustering.The situation,thus making it necessary to select mutually dispersed targets,but eventually will be brought together,thus invalidating some of the functions in the design;the existing application layer connection agent technology collects and distributes all the information of the deployed nodes.The anchor function is bundled with the common forwarding node role function,so that the newly deployed node cannot obtain complete information in the first time and the internal maintenance overhead of the system is too high;using the database as the data storage medium of the runtime,slow IO will block the processing speed of network messages.Based on the research of the prior art,the existing application layer routing technology is optimized in terms of scalability,and a new solution to the defects of ALCA is proposed.The concept of"RTT circle layer"is proposed.According to the distribution of transmission delay,the target node is selected purposefully to avoid the situation of target aggregation;the function roles as the"anchor point"are separately extracted and work independently,and the multi-anchor points of the message synchronization are designed to run at the same time;ALCA performs structural function design and coding,called ALCA2.0.This paper first conducts background research,introduces several existing application layer routing related technologies,and the bottlenecks they face,and discusses the value of these existing technologies for designing ALCA2.0.It then describes the shortcomings of the existing ALCA technology and how ALCA 2.0 will be improved.Next,introduce the scene-oriented and analysis requirements of ALCA2.0,so as to design the characteristics of ALCA2.0,and design a scenario to introduce the principle and workflow of ALCA2.0.7Then,the specific design and coding work of the ALCA2.0 system will be introduced.The detailed process of all the processes in the system will be described.Finally,in the wide area network composed of multiple cloud hosts in the world,ALCA2.0 is deployed and tested to verify the function and performance of the system.