APT: A Practical Tunneling Architecture for Routing Scalability

Abstract

The global routing system has seen a rapid increase in table size and routing changes in recent years, mostly driven by the growth of edge networks. This growth reflects two major limitations in the current architecture: (a) the conflict between provider-based addressing and edge networks' need for multihoming, and (b) flat routing's inability to provide isolation from edge dynamics. In order to address these limitations, we propose A Practical Tunneling Architecture (APT), a routing architecture that enables the Internet routing system to scale independently from edge growth. APT partitions the Internet address space in two, one for the transit core and one for edge networks, allowing edge addresses to be removed from the routing table in the transit core. Packets between edge networks are tunneled through the transit core. In order to automatically tunnel the packets, APT provides a mapping service between edge addresses and the addresses of their transit-core attachment points. We conducted an extensive performance evaluation of APT using trace data collected from routers at two major service providers. Our results show that APT can tunnel packets through the transit core by incurring extra delay on up to 0.8% of all packets at the cost of introducing only one or a few new or repurposed devices per AS.