Unleashing network-layer evolution for Future Internet Architectures: Michel Machado
- 10:30 am on Monday, February 4, 2013
- 12:30 pm on Monday, February 4, 2013
- MCS 148
Abstract: As clean-slate designs of the Internet accumulate, IP’s inability to evolve its network layer is slowly becoming perceived as a straitjacket for innovation. However, two factors have been releasing this mounting pressure to date, namely: (1) the fact that most existing clean-slate designs are siloed and elevate a few network use cases above others, and (2) designers find it difficult to bring a new design into fruition and demonstrate its merits, given the lack of a suitable platform to do so. Since the Internet brings together a myriad of stakeholders whose goals are not always aligned, replacing our Internet with a new design that does not benefit everyone is sure to find resistance. Furthermore, the lack of a proper platform to develop these new designs sharply limits the designs that are tested and lend themselves to experimentation beyond the hands of their creators. This thesis presents the eXpressive Internet Architecture (XIA), itself a clean-slate design. We argue that XIA’s highest goal should be to support evolution to the extent that other clean-slate projects can use XIA as a platform to implement, test, and deploy their designs. At the same time, XIA-enabled networks will realize the benefit of enriching their infrastructure by aggregating these new designs together. We implement XIA in Linux as a native network stack, addressing issues pertinent to production deployment, and evaluate Linux XIA's performance against that of Linux's IP stack. Finally, due to the lack of a metric to quantify how evolvable XIA is, we bring Serval, a component of the Future Internet Architecture (FIA) Nebula project, to XIA, in order to showcase the implementation of an alien design and to evaluate its performance against its original IP-based implementation.