Design And Implementation Of Supervisory System For Distributed Search Service

Services provided by searching engine clusters are playing an important role in modern large-scale websites.Generally these are large clusters and they consist of different parts with different roles working together interdependently.How to manage these services and data efficiently and conveniently is a hard problem that the searching service administrators are facing nowadays.To solve this problem,the author designs and implements a system for the service management and supervisory based on study about the daily operations and upgrade work flow with searching service.The author investigates and analyses the requirements,designs the modeling of main workflow of the system and writes the code to implement it.After lots of tests,the system is deployed in the production environment.The distributed system is constructed by two parts: the first is a system managing task publishing and system control named by Admiral which is the control and schedule center,the second is for manage task execution and system supervisory named by Marine which are deployed on every server computer.Users can interactive with Admiral directly via Web or command line,to query the status of service been managed or to publish management tasks.Admiral and Marines are all implemented by event-driven method and are based on the coordination service provided by Apache ZooKeeper.They work together as a system for service management and supervisory.The system automates and parallels the workflow and operations in the service management job,cutting down the time and energy devotion of service managers.The rigorous supervisory provided by the system greatly improves the credibility of management tasks.The system can be used as a DevOps platform for server machines management and service update,configuration management and system status monitoring.It has been proven that the system worked smoothly and well performed with 58.com searching service.