Improvement And Implementation Of Middleware Based On ICE

In the third wave of the computer revolution,middleware is one of the core technologies.Grid computing and ubiquitous computing are the main forward directions of middleware.Grid computing is also called “distributed computing”.Middleware's application integration capabilities are enhanced;ubiquitous computing is designed to extend middleware to a broader application environment.ZEROC's ICE middleware is a new object-oriented RPC framework that supports multi-language development and comes with a variety of services to handle complex application environments.ICE middleware is widely used and its performance is high.However,there are still some problems in use.Through in-depth study of ICE,the paper made the following two improvements to ICE middleware:In the micro-service framework built by ICE built-in multiple services,IceGrid is the core service of the whole framework.It provides services such as location request,application deployment,load balancing,etc.for the client,providing service management and service discovery for the server.Wait.When the IceGrid service is unable to provide services for some reason,the entire network communication will be interrupted.Although IceGrid itself is highly available,it uses the master-slave mode,and does not implement automatic master-slave switching,which is completely manual,which seriously affects the efficiency of the IceGrid service.Zookeeper is the popular distributed collaboration software.Its source code is open and provides consistent services for distributed applications.This paper first proposes a high-availability improvement architecture based on Zookeeper's IceGrid service.When the primary node cannot provide access,Zookeeper automatically discovers and uses its own consistency protocol to implement the master-slave switch of the IceGrid service.ICE middleware implements several load balancing algorithms to access services,but the built-in load balancing algorithm only uses the load information of the node server as the metric of service selection.It does not consider the performance difference between servers,lacks comprehensiveness and efficiency.Also greatly discounted.Aiming at the one-sidedness of the original load balancing algorithm,based on the performance of the overall service cluster,the original load balancing algorithm is improved.The improved algorithm calculates the load of a single server by calculating the percentage of performance of a single server as a percentage of the overall cluster performance,and combines the system information such as CPU usage.The node with the least load is selected as the node for processing the task.