Research And Implementation Of Distributed Marketing System Based On Microservice

Diverse marketing systems are emerging with continuous innovation of corporate marketing methods and Internet marketing concepts.According to specific rules of various channels of marketing they collect all kinds of user information to promote sales and products.However,with the characteristics of centralized sending volume,large statistical time span,and large amount of basic data,marketing systems are challenged by performance and stability.To solve these problems,this paper adopts microservice to improve performance and stability by the parallel processing in an industrial case of marketing system for an airline company.And the whole system is also described in depth,which includes the process of user data transmission,data cleaning and warehousing,marketing transmission processing,push and send,data collection and statistics,etc.In the distributed architecture of microservice,there is a bottleneck problem of gateway routing service nodes.This paper focuses on the load balancing algorithm who dispatch other service in distributed processing.The purpose of load balancing is to maximize the resource utilization and to make dispatching more reasonable in distributed sever.The main load balancing algorithms include the dynamic load balancing,the static load balancing,and the adaptive load balancing.The specific algorithms has the least connection algorithm,round robin algorithm,weighted round robin algorithm,etc.However,these algorithms have the problem that they cannot be dynamically adjusted based on the usage of different server resources.According to the CPU and memory utilization,this paper proposes a dynamic load balancing algorithm which combines the task weight of the system,so as to improve the overall performance of the system.Compared with the traditional round robin algorithm,the transmission efficiency of 100,000 messages is improved by 2 times.Compared with common CPU and memory based algorithms,the efficiency of 100,000 messages is improved by 1 times approximately.Based on MongoDB's partition distributed storage,for the nodes heterogeneity in multiple machine environment and nodes failure in single machine,this paperoptimizes MongoDB's partition cluster,and adds the consistency hash algorithm with virtual nodes to ensure the balance and uniform storage.By introducing algorithm,the system can complete the normal sending task within 24 hours in case of single node failure.