本文關鍵詞：面向超市電子貨架標簽的高并發(fā)服務器系統(tǒng)　出處：《華南理工大學》2015年碩士論文　論文類型：學位論文

  更多相關文章： 電子貨架標簽 高并發(fā) Reactor模式 線程池 數據庫連接池

【摘要】：電子貨架標簽用于標識商品的名稱、價格等信息,由于其具有綠色環(huán)保、低功耗、壽命長、數據更新便捷、顯示內容豐富、數據錯誤率低等優(yōu)點,越來越多的超市開始使用電子貨架標簽替代傳統(tǒng)的紙質價格標簽,這樣不僅提高了超市的管理效率,而且節(jié)省了超市的人力成本。由于大型連鎖超市的電子貨架標簽數量龐大,數據通訊頻繁,普通的服務器系統(tǒng)往往無法滿足電子貨架標簽高并發(fā)的上下行通信,所以,后臺的服務器往往成為電子貨架標簽系統(tǒng)的性能瓶頸。本文針對電子貨架標簽集群與服務器之間的高并發(fā)數據通信問題進行研究,分析了電子貨架標簽集群以及所使用的Zig Bee協(xié)議的通信特點。在對比分析了不同服務器框架的優(yōu)缺點之后,設計并實現了基于Java.NIO技術的Reactor模式高并發(fā)服務器。該服務器系統(tǒng)解決了電子貨架標簽集群和服務器之間高并發(fā)的訪問難題,有效地減小了服務器的平均響應時間,增大了最佳并發(fā)數,提高了每秒查詢率TPS。在設計服務器的同時,本文還設計了基于My SQL的高性能后臺數據庫,以及人性化的后臺管理監(jiān)控軟件。在Reactor模式服務器中,分別設計了應對高并發(fā)負載請求的線程池和數據庫連接池。在線程池的設計方案中,不僅實現了多任務的并行處理,還提出了基于M/M/1/Q/∞/Dynamic排隊模型的任務緩沖隊列算法,實現了在線程池容量有限的情況下,最大限度地提升線程池的并發(fā)能力,在資源消耗和線程池并發(fā)能力之間做了很好的平衡。在數據庫連接池的設計方案中,不僅減少了不必要的數據庫連接創(chuàng)建與銷毀,還提出了自適應的連接池容量調整策略和MRU數據緩沖策略,實現了數據庫連接的動態(tài)分配和自動回收,提高了對熱數據的訪問速度。并通過定時刷新緩沖區(qū)的方式,保證了緩沖區(qū)數據的真實性。
[Abstract]:Electronic shelf label is used to identify the name, price and other information of goods, because of its green environment, low power consumption, long life, convenient data update, rich display content, low data error rate and so on. More and more supermarkets begin to use electronic shelf label instead of traditional paper price label, which not only improves the efficiency of supermarket management. Because of the large number of electronic shelf labels and frequent data communication, the common server system often can not meet the high concurrency of electronic shelf labels up and down communication. Therefore, the backstage server often becomes the bottleneck of the performance of the electronic shelf label system. This paper focuses on the problem of high concurrent data communication between the electronic shelf label cluster and the server. This paper analyzes the communication characteristics of the electronic shelf label cluster and the Zig Bee protocol used. After comparing and analyzing the advantages and disadvantages of different server frameworks. The high concurrency server in Reactor mode based on Java.NIO technology is designed and implemented. This server system solves the problem of high concurrency access between the electronic shelf label cluster and the server. It can effectively reduce the average response time of the server, increase the optimal concurrency number, and increase the query rate per second. At the same time, the server is designed. This paper also designs the high performance background database based on my SQL, and the humanized background management monitoring software in the Reactor mode server. Thread pool and database connection pool are designed to deal with high concurrent load request respectively. In the design of thread pool, not only multi-task parallel processing is realized. A task buffering queue algorithm based on the M / M / 1 / Q / 鈭，

本文編號：1403943