Welcome to the home page for the network of Autonomous System 1050. We are an overlay network experimenting with emerging network technologies in addition to providing connectivity to partner organizations and projects.
One of our primary research topics include the complete automation of our network. All routers, servers and services are 100% automated and managed using a declarative approach via Ansible. This allows all components to be taken from initial OS installation to full service without any direct interaction. All virtual routers are currently Ubuntu based utilizing a customized version of the BIRD2 routing suite.
Secondly, the network is designed with scalability in mind. The backbone consists of an IPv6-only ZeroTier SD-WAN VPN. Backbone routers are minimally addressed and utilize IPv6 link-local addressing to communicate and route traffic. The routers utilize OSPFv3 to find two redundant route reflectors. Edge routers are able to directly communicate with each other and require no hair-pinning through any matrix of point-to-point VPN connections. This configuration allows for routers to be added or removed anywhere, anytime with very little effort while maintaining a very minimal intra-AS routing table.
Due to the nature of our network and the presence of a virtual/overlay backbone, edge routers are weighted to prefer routes learned via its local peerings. This prevents hair-pinning traffic and promotes the closest available real egress without affecting downstream route selection. In summary, traffic received on an edge router in almost all cases uses a local connection versus tunneling across the virtual backbone despite another edge router having a “better” path.
Downstream routers perform BGP best path selection to select the most appropriate edge router based on customized communities, local preference and other traffic engineering techniques that may be under research.
Our future goals include Vector Packet Processing (VPP) with Segment Routing IPv6 and FRR/BIRD2 for FIB programming. This would best be accomplished with physical hardware. Unfortunately as a hobbiest/research network we are unable to sustain the cost of most co-location options at any kind of scale. If you happen to have fairly minimal rack space (1-2U), BGP peering options and would be interested in supporting our efforts, we would love to hear from you!
We are always interested to hear from other like-minded network engineers, computer scientists, software developers, etc. If you have suggestions, interest in participating or have other general inquiries please contact us.