Design&Implementation On The Scalable Mechanism Of Remote Service Invocation In The Cloud Environment

Cloud computing has become a trend of development of the Internet, Softwareas a Service (SaaS) is an important type in cloud computing. In the cloud computingenvironment, the cooperation process needs to access resources distributed acrossthe network, a good communication mechanism is the core of all distributedsystems. Remote service invocation plays an important role in the SaaS fast deliveryplatform which provides reliable communications infrastructure for the entireplatform and makes the SaaS application run in a normal way.This paper provides a scalable and effective remote service invocationmechanisms for the private cloud environment of small companies.It introducesSOA into cloud, and decouples the SaaS application into three levels:dataservice,biz service and UI service.The SaaS developers can make stateless serviceand can create stateless service and add nodes for the service if need.So the servicecan scale out at feedom, the performance is almost linear growth as the nodesadded.In the program selection part of this paper, it first analyze and compare severalpopular remote service invocation mechanisms to draw the new RPC-based remoteservice invocation mechanism far more efficiency than the RMI, the Web Service aswell as traditional RPC. Then three new RPC frameworks like Thrift, Avro andProtocol Buffers in the way of serialization, deserialization, and function werecompared, and ultimately proposed a remote service invocation mechanism basedon the Thrift framework (TSF).TSF introduces SOA concept and decouples a SaaSapplication as three layers: data services, business services and representationservices, is a scalable service invocation framework with a set of service definition,service releases, service invocation and service management.TSF consists of threeparts: core service components, service registry components, and RESTFul wrappercomponents. The core service component is responsible for defining the service togenerate the calling code, publishing services and the provision of Thrift basedservice interface.Registry service component Zookeeper is for storage andmanagement services. RESTful service wrapper component is based on Wink clientframework to achieve the role to show the web service RESTful from Js pages. The tests show that,when service is deployed in the case of multiple virtual machinenodes compared with deployed in the single virtual machine node, the performanceis almost linear growth in TSF.In order to meet the need of "Track Trace system" in the aspect of messaging,TSF introduces the publish/subscribe mechanism based on Kafka, its perfectintegration with the existing framework in the performance testing satisfy the"Track Trace system" in the daily use.