| The Service Innovation Platform is a series of network service infra-structures provided for developers in many scientific institutions and uni-versities.Such as SDKs(Service Development Toolkits),SEE(Service Execution Engine),abundant resources of network service and cloud-based infrastruction platforms.The goal of this platform is to lower the technical barrier of service development and innovation,giving full play to the power and wisdom of researchers and groups.The Service Innovation Platform helps researchers and developers to contact new experiments and research of services by providing an open and new environment for design,development,deployment and test.Currently,this platform could offer virtual machine services,database services,and Docker container services.However,the services based on container face two problems:(1)The number of container service copies can only be set when creating a service,which cannot be automatically adjusted according to volume and load changes.(2)The container service only supports static load balancing so that real-time service discovery and dynamic load bal-ancing cannot be performed according to the change of back-end contain-ers' status and addresses.In order to solve the problems above,this paper firstly uses the Docker swarm cluster management and orchestration tools to manage virtual ma-chine clusters and create the application services.Secondly,we implement a full set of elastic scaling and dynaimc load balancing system which can adjust the scale of containers and clusters according to volume and load by designing modules including monitor module,elastic scaling module,load balancing module and using Traefik for service discovery.Finally,the paper carries out functional and performance testing respectively,where test re-sults show that the elastic load balancing system can effectively reduce ser-vice response time,improve user experience,increase overall resource uti-lization and reduce users' cost. |
PDF Full Text Request