Optimize The Data-intensive Oriented Application Of Web Services Composition

In recent years, with more and more network resources posted on the Internet, thedistributed computing has become the mainstream model of the industry on researchesand applications. In a distributed computing model, the computing tasks areaccomplished in collaboration by multiple computers distributed on the network. In orderto achieve cross-platform and cross-system and extend the interoperability of applicationsprovided by different providers, the service oriented computing based on the distributedcomputing platform has generated.Meanwhile, with the rapid development of informationization, the scale of data growsdramatically. The data-intensive computing has become the common research point inindustry and computer academia. Because the development of Service-Orientedcomputing and Web services technology has gained great success in data encapsulationand integration, the industry and the academia are both working on Web services andservice-oriented architecture to fulfill the development, execution and management ofdata-intensive applications.However, with the increasing of large-scale data, low execution efficiency hasgradually become an obstacle to the development of data-intensive application serviceoriented computing. Studying how to combine the parallelism with Web servicescomposition process will make difference to effectively improve the execution efficiencyof data-intensive application process.First of all, this paper aims at data-intensive applications to enable the user to get acomposite service with the shortest response time and maximum data processing at thelowest cost. We propose the concept of the Benefit Ratio of Composite Service (BROCS)and a model, based on which we derive the algorithm of DOP (Degree of Parallelism)calculation. The decision of the degree of parallelism ensures that the user has anacceptable response time, costs and the maximum amount of data processed.Afterwards, on terms of synchronous mode, it may cause a waste of time and resourcesto the composite service. We research how to optimize Web services composition basedon this problem. For data-intensive oriented applications of Web services, we divide them into I/O-intensive services and CPU-intensive ones based on the I/O operations and CPUcomputing operations. Accordingly, we put forward the concept of I/O operation rate anda parallel strategy of Web services. The basic principle is to make the I/O operations andCPU computing operations overlap to each other as possible. On one hand, the CPU willbe busy when I/O is waiting. On the other hand, it will spend the least on the I/Oscheduling of the CPU to avoid wasting CPU resources and I/O scheduling.Finally, we carry out an experiment with the data of the Sina to verify the validity ofthe BROCS model.