Implementation of publisher-subscriber technology to facilitate interaction between moving and stationary objects

Introduction. Publisher-Subscriber technology is an effective tool for organizing real-time interactions between moving and stationary objects, as well as facilitating asynchronous message exchange between system components.

The aim of the research was to improve the efficiency and reliability of communication between moving and stationary objects for use in robotic systems (RTS). The paper discusses the most popular implementations of Publisher-Subscriber technology, including SocketIO-based brokers, hbmqtt, Mosquitto, and RabbitMQ. The study evaluates the computational resources consumed by these implementations, specifically RAM usage, CPU utilization, and client connection times to the brokers.

Findings. Implementation results for the RabbitMQ broker: RAM consumption was 86 MB, and CPU utilization was 0%. For the Mosquitto broker, RAM consumption was 6 MB, and CPU utilization was 0%. The SocketIO-based broker consumed 32 MB of RAM, with 0% CPU utilization. The hbmqtt-based broker used 30 MB of RAM and 0% CPU. The following results demonstrate resource consumption under load. RabbitMQ: RAM consumption was 94 MB, with CPU utilization at 14.3%. The message sending rate was 4,000 messages/second. During the first 15 seconds, latency increased but then dropped to minimal levels, and the broker delivered all subsequent messages with minimal latency. Mosquitto: RAM consumption was 12 MB, with CPU utilization at 17%. The message sending rate was 18,000 messages/second. At 54,000 messages/second, latency exceeded 18 seconds. SocketIO: RAM consumption was 34 MB, with CPU utilization at 3.3%. The message sending rate was 36,000 messages/second. Latency gradually increased, with a 99.8% message loss. hbmqtt: RAM consumption was 60 MB, with CPU utilization at 99.7%. The message sending rate was 20,000 messages/second. At 60,000 messages/second, latency increased to 45 seconds.

Conclusion. It is quite clear from the data that the RabbitMQ broker demonstrates the best performance. Although, it has a limitation of 4,000 messages per second, this does not hinder message passing. Instead, it helps minimize delays and reduces system load. RabbitMQ also offers functionality that is either absent or difficult to implement in the other brokers evaluated.

Practical significance. The results can be applied to further implement Publisher-Subscriber technology for organizing interactions between moving and stationary objects, with the potential for scaling these interactions in RTS.