Application Deployment at Distributed Clouds

Meer over het boek

A current trend in Networking and Cloud Computing is to provide compute resources at smaller centers or within the network itself, such as nearby backbone routers, rather than solely at large data centers. These resources are often closer to customers, offering better connectivity. This thesis explores the benefits and challenges of deploying applications across multiple sites, focusing on optimization problems related to network service quality (response time and data rate) and resource costs. It examines application architectures with multiple components and their interdependencies, and presents an automated management system for deploying, scaling, and configuring applications at runtime. A Template graph describes the application components and scaling properties, leading to the Template Embedding Problem, which allocates resources and scales applications efficiently. Additionally, the integration of queuing systems into a Facility Location Problem enhances the accuracy of resource allocation optimized for response time. This convex problem employs linearization techniques to balance solving time and solution quality. While these problems are solved centrally, a distributed algorithm for the Facility Location Problem yields high-quality solutions. The thesis also details a system design, including architecture and interfaces, along with a proof of concept and a testbed emulating a Distributed Cloud Computing scenario, whic