Rethinking cloud platforms: Network-aware flexible resource allocation in IaaS clouds

Most of the current platforms for cloud infrastructure management are designed to deal mainly with computing and storage resources. However, when deploying highly distributed applications with strict network requirements, such as low delay or bandwidth guarantees, the support for specification and configuration of such requirements still lacks. Moreover, resource allocation strategies and algorithms are usually hard-coded into the cloud platform's core, making it very difficult to improve or adapt these strategies to better lit individual application and environment needs. In this paper, we introduce a new approach to cloud platform design, emphasizing three main aspects: (i) robust networking for coupling cloud computing with modern network paradigms, (ii) specification of complex virtual infrastructures, including network topology and application requirements, and (iii) programmability via an API to ensure customization at the core of the platform's resource allocation and optimization strategies. We still present a proof of concept prototype that we have implemented and deployed over a modern network testbed and also evaluated on an emulated network using Linux virtualization containers, Open vSwitch, and mininet. Using our prototype, we have conducted a case study featuring an information-centric networking (ICN) application. Initial results show the feasibility of our approach and the application deployment statistics.