High-concurrency Server System For Supermarket Electronic Shelf Labels

The electronic shelf label is used to identify the goods’ name and price, etc. Because it has green, low power consumption, long life, convenient data updating, rich content displayed, low data error rate and other advantages, more and more supermarkets began using electronic shelf labels to replace the traditional paper price tags, this will improve the management efficiency of the supermarket, as well as save labor costs. The large supermarket chains have large number of electronic shelf labels, and they need to communicate frequently, regular server system can not meet the high-concurrency uplink and downlink of electronic shelf labels, so the server backend systems often become performance bottlenecks.This paper studies the problem for highly concurrent data communication between the electronic shelf label clusters and server,and analyzes the characteristics of electronic shelf label clusters’ communication and Zig Bee networks used. After comparing and analyzing the advantages and disadvantages of different server framework, this paper designs and realizes of technology-based Java.NIO Reactor pattern high-concurrency server. The server system solves the high-concurrency access problem between electronic shelf label clusters and server, effectively reduces the average response time of the server, increases the number of concurrent best and improves the query rate per second,TPS. In the design of the server at the same time, this paper also designs a high-performance database backend based on My SQL, and designs humanized management and monitoring software backend.In Reactor pattern server, thread pool and database connection pool are designed to cope with high concurrent load request. In the thread pool design plan, not only achieving parallel processing of multiple tasks, but also proposing task buffer queue algorithm based on M/M/1/Q/∞/Dynamic queuing model. Under circumstance of limited thread pool size, the task queue buffer algorithm can achieve maximize concurrency of thread pool,make a good balance between resource consumption and concurrency of thread pool. In the database connection pool design plan, not only reducing unnecessary database connection’s creation and destruction, but also proposing adaptive capacity adjustment strategy of connection pool and MRU data buffer strategy, achieve a dynamic allocation and automatic recovery of database connections, increase the access speed of the thermal data. And ensure the authenticity of the data buffer by way of periodically refreshing buffer.