Research On Transmission Architecture And Data Consistency In DDS Communication Middleware Technology

Today,distributed has become an essential part of system construction.Although distributed computing based on distributed systems has a larger scale and higher availability,it is not as good as parallel computing in terms of real-time performance.In the context of soft and hardware co-design,distributed systems should not only have functional and consistency guarantees,but also meet real-time and reliability requirements.In the edge computing scenario,the resources of nodes are limited,the network environment is complex,and data may be lost during transmission.Communication by traditional transport links not only lacks flexibility,but also has difficulty in meeting the requirement of low latency.Also,data exchange with publish/subscribe mechanism is also prone to data conflicts.The paper adopts a hierarchical generic DDS communication middleware architecture with availability,flexibility,and scalability,and focus on the implementation of the real-time in the transport architecture and the data consistency.From the perspective of real-time,DDS communication middleware needs to establish a low-latency transport links which allow messages to be distributed from one node to each node in the distributed system in a short time.Bypassing the network stack from the kernel of OS allows the adjustment of the implementation of the transport protocol to adapt different transmission scenarios and reduce the latency with best effort.The paper implements a customable network stack protocol through packet capture and supports both TCP and UDP links.From the perspective of high consistency,the DDS communication middleware should ensure that all nodes in the distributed system have the same data and can maintain data through a certain mechanism in case of conflicts.The paper implements a consensus algorithm based on Raft algorithm,using the characteristics of DDS publish/subscribe.In addition,the paper also realizes a specific application through the DDS communication middleware.The communication layer has a millisecond-level latency,and can be highly customized,with real-time and flexibility.The high consistency and availability are also verified with the consensus algorithm,which keeps the data in each node of the system consistent and has certain error recovery capability.The DDS communication middleware can correctly subscribe and publish topics for communication and data exchanging between nodes and meets the requirements of scalability.The implementation of the DDS communication middleware achieves its purpose.