Optimization Design And Implementation Of Data Interaction And Storage In Cloud Platform

The application of cloud computing has become an inevitable trend,and the role of platform has changed to provide data interaction services to end users.From a vertical perspective,a service cluster is a layered system architecture that handles the corresponding service requests and data processing through different service layers.From a horizontal perspective,the synergistic effect of cluster nodes ensures that the pressure of single application is reduced and the abnormal situation of service handling is handled timely.In actual application scenarios,the user's request is unpredictable.When in a normal access state,the service architecture can provide more stable data support,but when a large number of service requests or abnormal requests occur,the system architecture needs to be improved in many aspects such as service architecture,build cost,performance improvement,etc.,to improve the various components in the system.Part of the service component has a more formative design and performance solutions,but is not suitable for the existing platform architecture.While improving the throughput and response time of the system by comparing the corresponding optimization schemes,we also consider whether some mature middleware can act as an intermediary coordinator in the cluster.Finally,combined with the characteristics of server request service,some optimization schemes and system structure improvement processing ideas are proposed.Starting from the load application scenario,this thesis introduces the processing strategies of different application loads.For the processing of proxy requests by service nodes,the application layer is divided into two processing cores: flow limiting processing and request distribution processing.Aiming at different processing priorities for different processing layers,corresponding core service processing algorithms and burst request token strategies are designed to ensure that high-priority requests can be processed quickly;Fine-grained division of requests forwarded by the current-limiting layer and design of corresponding load distribution strategies ensure that the service nodes allocated to client requests have normal business processing capabilities.In order to reduce the complexity of the system module and ensure the efficiency of the cache layer,Zookeeper is used as the coordinator to build the cache cluster and improve the system deployment strategy.Data of different structure types are stored separately,structured data is stored in relational database,unstructured data is stored in unrelational database,and index of unstructured database is transferred.In the process of storing data,database can be better combined with its own processing characteristics to improve its data read and write performance.The system not only optimizes the service module,but also applies the performance monitoring scheme to the existing design module in order to avoid the unnecessary trouble caused by the system-level monitoring component.Optimization and monitoring complement each other in the system.When a service node exceeds a set threshold,a timely warning is issued and real-time performance data can be viewed through the front end.After monitoring and finding the system's bottleneck,the system structure is optimized to provide better service.In the case of high concurrent requests,the system not only improves the troubleshooting processing efficiency,but also ensures the availability and stability of the system.