Scaling Service-Oriented Applications into Geo-distributed Clouds

With the significant prevalence of cloud computing, more and more data centers are built to host and deliver various online services. However, a key challenge faced by service providers is how to scale their applications into geo-distributed data centers to improve application performance as well as minimizing the operational cost. While most existing deployment methods ignore the service dependencies in an application, this paper proposes a general dynamic service deployment framework to bridge this gap, in which a deployment manager and a local scheduler are designed to optimize data center selection and auto-scale the service instances in each data center respectively. More specifically, we formulate the deployment problem across multiple data centers as a compact minimization model, which can be solved efficiently by a genetic algorithm. To evaluate the performance of our approach, extensive experiments are conducted based on a large-scale real-world latency dataset. The experimental results show that our approach substantially outperforms the other existing methods.