At last month’s APNIC 38 conference, Tony Hain made a presentation about an intriguing model for a ‘virtual ISP’ system being deployed nationwide in China. This is using modern layer-2 and layer-3 methods to construct an open-bus for companies to compete with service delivery, as a ‘virtual’ ISP on top of the base service, fibre to the home.
The service is interesting because it leverages the power of a large national deployment of the underpinning ‘utility’ layer, analogous to the ‘pits and poles’ discussions we see in power distribution networks, or many national broadband networks (for example, the Australian NBN as originally planned). The project, called ‘aBitCool‘, aims to achieve multi-gigabit speed and if there are no competing entities in your region you can even get dark fibre.
Due to regulatory requirements, the BGP routing is still very centralized, but the design of the system has been done to allow more open peering when the environment changes, and it looks very applicable for many economies looking to have competing ISPs on top of a utility national infrastructure.
But, having noted that the design pushes BGP to the ‘edge’, the overlay providers wind up being owners of their own routing and peering. This has upsides and downsides, and is certainly interesting. As Tony said, as well as being a routing protocol, BGP is a ‘policy engine’ and there are other ways to do policy at the edge.
What struck me is how strongly the OSI model of software development has been embedded in the design: their default routing engine is based on OpenWRT for the edge, and they have looked at ‘Chromecast’ as a proprietary Google only solution, and worked with Firefox to develop a neutral platform equivalent called ‘Matchstick‘.
Another interesting observation from Tony was that none of this kind of design pushes the edge of technology very hard: we’ve had capabilities to do this kind of large-scale competing provider model for some time.
This is a good fit for an IPv6 network; with a large deployment, lots of devices, end-to-end behaviours and an interest (noted in the NIR SIG) in IoT developments from Chinese companies, they expect to be coming back to APNIC to get their IPv6 blocks!
The views expressed by the authors of this blog are their own and do not necessarily reflect the views of APNIC. Please note a Code of Conduct applies to this blog.