Towards Deep Network & Application Integration: Possibilities, Challenges, and Research Directions

The collaboration between networks and applications provides opportunities to both applications to improve their performances and network service providers to increase business offering. Although many systems are proposed to support such collaborations, they are point or incremental solutions. In this paper, we propose the exploration of a more integrated architecture with huge possibilities taking a network-application integration (NAI) approach. Specifically, we explore the NAI possibilities in two concrete aspects: application-aware networking and network-aware applications. We review recent progress in these two aspects, and identify the key challenges in systematically realizing such a deep integration. To address these challenges, we present the initial design of a generic NAI possibilities exposure and discovery framework, called PED, based on satisfiability modulo theories (SMT). The key components of PED include a unified, abstract representation of network information using mathematical programming constraints, a declarative language for applications to express their intents on discovering network information, and an efficient compiler to translate application intents to constraint programming problems and discover corresponding network information. Preliminary evaluation results demonstrate the potentials of the PED framework. At the end of the paper, we also discuss a series of key future research directions toward deep NAI.