| Service computation has made fundamental changes on software system structure, SOA and SaaS for typical representative. The existence of the software system, release means, operation and maintenance methods all have a great change, which promotes the development of the software system structure and its theory of technology. More and more enterprises have started to use a variety of SaaS application from different software companies and the scale of SaaS deployment increases gradually. There are more and more deals which involve large enterprises having thousands of even tens of thousands of users. In this new mode, the requirements of online service application for reliability and operation efficiency are higher. On one hand, SaaS application systems use multi-tenants mode broadly, while meeting the performance indicators of tenants, such as continuous service, service throughput, adapting to the requirements of the service ability fluctuation, as a provider of service, should also improve the overall resource utilization rate and reduce cost through the optimization of the deployment, which requires the system to use resources more efficiency and maximize the resource sharing between different users based on meeting the needs of users. On the other hand, because of the large quantity of service users, the requirements of online service system for real time reliability are also higher.Among the current popular SaaS application, the system used by end users is only a virtual system, which means that the end user use it but not possess it. The system is combined of several components, so the key factor of improve the performance of SaaS application and efficiency of resource is the ability of optimization deployment of the set of components corresponding to it. Different from the traditional optimization and reliability security, the deployment can be more efficient through the dynamic migration of components to meet the requirements of real time and online service.This paper presents an automatic service migration to solve the problems above. Making migration decision according to the business forecasts and load balance, allocating and scheduling the computing resource uniformly according to the change of server load, which will solve the problem of load unbalance, reduce the response time of services and improve quality and availability of services.The main contributions are as follows:1. Present a framework for real-time monitoring of the host environment, calculates the real-time load of the server based on monitoring parameters to decide whether to start the migration and reports the real-time status of the running service.2. Propose an approach of service migration based on business forecasts and real-time load testing, using Markov chain to predict the business fluctuations of the SaaS tenant application in the near future, combined with the real-time load testing of the service to decide whether to migrate the service and compute the optimized path for the migration.3. Propose an automatic service migration based on load balance. From these three aspects, load calculation, migration strategy, service migration, researches how to optimize the allocation of resource through service migration under the restriction of the limited resources. Calculate the load using the weight vector method; make a perfect migration strategy for the selection of source service instance, migration moment and destination node; propose a framework of migration and realize it based on load balance.4. Put forward a kind of service state protection method, which ensures the consistency of service state before and after the migration and guarantee the availability of the service during the migration process.Combined with actual demand of specific application, we implemented load computing, the Markov chain-based business forecast method and the load balancing-based service migration strategy in this paper. It is shown from the experiments that the method proposed in this paper can in real time sense the change of the host environment, accurately forecast the change of the amount of service business, guarantee the usability of service in the process of service migration, greatly enhance the resource utilization rate, and ensure the quality and high usability of service. In the end, conclusion and prospect in terms of research work are provided in this paper. |
PDF Full Text Request