| With the rapid development of computer technology,micro-services as a distributed service solution has been widely recognized by the industry.However,the micro-services deployed on the Internet often have a single function and cannot meet the increasingly complex user needs.Therefore,how to reuse existing fine-grained services and use service composition methods to build coarse-grained services that meet users’ complex business needs and performance needs has become the research focus of researchers and engineers.However,the current service composition method has the problems of high implementation complexity,high cost of service reuse,low flexibility,and difficult to ensure the quality of composition.Therefore,it is necessary to design a flexible lightweight composition method.In addition,in the system built by organizing micro-services using the service composition method,multiple micro-services need frequent network communication.Due to the high latency of network communication,frequent network communication between micro-services will reduce the quality of service(Qo S)of the system.Using long connections can effectively reduce the delay of network communication.However,the traditional static long connection management mechanism cannot effectively meet the requirements of frequent changes in service communication links.Therefore,it is necessary to design a communication channel management method to dynamically schedule the establishment and use of communication channels.In view of the above problems,this paper has done the following work:(1)By analyzing the implementation process of existing micro-services composition methods,this paper finds that the problem of data and control interdependence in the form of API(Application Programming Interface)is the fundamental reason for limiting the flexibility of service composition mechanism.To solve this problem,this paper designs the Standardized Data Control Separation API(SCONDAS API).Using the SCONDAS API,this paper proposes a micro-component architecture that combines service choreography and service orchestration models.This architecture realizes efficient,flexible and lightweight micro-services composition.The experimental results verify the feasibility of this method.(2)Using micro-component architecture can efficiently and flexibly combine micro-services,and also reuse services at low cost.However,using multiple communication channels in the micro-component architecture may increase the cost of data transmission.Moreover,the traditional static long connection management mechanism cannot effectively meet the requirements of frequent changes in micro-services communication links.To solve this problem,this paper proposes an adaptive dynamic connection pool management method based on prediction.This method predicts the number of communication connections in real time based on the queuing model,thus effectively realizing the dynamic management of long connections in the connection pool under the scenario of frequent link changes.The experimental results show that this method can effectively improve the communication performance between services compared with the currently widely used working connection pool management method.(3)The SCONDAS API has the separating feature between the data and control.The control information of the service node needs to be stored locally,while the business function execution data will be transmitted in the network.In order to improve the efficiency of micro-services component interaction and further reduce the time and resources consumed by the standardized data receiver to parse data,it is particularly important to evaluate the serialization protocol suitable for SCONDAS API among many protocols.Based on this problem,this paper proposes an evaluation algorithm,and uses this algorithm to serialize and deserialize a variety of open source libraries. |