Real-time data distribution method and device based on shared memory

Abstract

The invention discloses a real-time data distribution method and device based on a shared memory, and the method comprises the steps of enabling a publishing process and a receiving process to load adata distribution assembly, enabling the data distribution assembly to apply for a plurality of shared memory channels, and enabling each shared memory channel to be used for the data distribution ofa data theme; receiving an interested data theme subscribed by the process, applying for a shared memory channel from a data distribution component, and fixing the data theme of the applied shared memory channel as the subscribed data theme; enabling the publishing process to publish the data to all shared memory channels corresponding to the data theme to which the data belongs; enabling the receiving process to regularly inquire whether data exists in the applied shared memory channel or not, and if yes, receiving the data in the shared memory channel; otherwise, continuing timing queries. The copy frequency in the data distribution process is reduced, and the universality is high.

Description

technical field [0001] The present application belongs to the technical field of computer communication, and in particular relates to a shared memory-based real-time data distribution method and device. Background technique [0002] Due to the limited CPU single-core performance improvement, the current high-performance CPU design is developing towards multi-core and multi-thread. In order to use CPU resources more efficiently and avoid the service provided by the application from stopping due to the crash of a single process, the application usually chooses the multi-process Developed application architecture as the preferred architecture for complex applications. Different from single-process multi-threaded applications: the inter-process data space isolation of multi-process applications cannot directly share data like multi-threaded architectures. Therefore, multi-process architecture applications need efficient message passing and data distribution mechanisms. The com...

AI Technical Summary

Problems solved by technology

[0004] However, the existing technology has many shortcomings in performance, real-time and versatility.

The implementation performance based on TCP / IP network communication and custom protocols is limited by the implementation of the network protocol stack of the operating system, and the performance of network communication in a single machine is low; the IP and port information of the receiving module and the sending module cannot be shared in real time, and forwarding needs to be designed module, the data is copied multiple times between the transceiver module and the forwarding module, the communication efficiency is low, the delay is large, and it cannot meet the use occasions with high real-time requirements (such as robot motion control, automatic driving control); the custom communication protocol lacks specifications And standards, usually highly coupled with the application framework, poor versatility

[0005] Data distribution based on a message bus (such as DDS middleware) is generally used for data transmission and reception between multiple machines, and it is also based on the TCP / IP protocol. There are also real-time and performance problems, and this type of middleware is complex to implement and depends on libraries. Many, it is a heavyweight message bus, which cannot be easily applied to embedded systems with high real-time requirements

[0006] Based on the method of data distribution provided by the message queue provided by the operating system, the message queue has a limit on the data length of a single message. Modifying the length limit requires configuration of the operating system kernel parameters, and it is difficult to form a general framework; more importantly, the message queue only supports one message. One-to-one, many-to-one data distribution cannot realize one-to-many, many-to-many data sending and receiving; the message queue implemented by the operating system has four memory copies in user space and kernel space, and the communication efficiency is low

Images