BGP-Mux

The following are some notes regarding integration of Georgia Tech's BGP-Mux into ProtoGENI:

There is a short (2 page) paper about BGP-Mux in CoNext 2007.

Subeject: Integrating with Nick's BGP mux
From: Robert P Ricci ricci at cs.utah.edu
Date: Mon Aug 4 19:59:56 MDT 2008

I talked with Nick Feamster for a while about what it would mean to get
his BGP mux integrated into a GENI prototype.

Recall: it's a sort of a proxy intended to allow multiple experimenters
to share one BGP feed, which they can use to collect data about routing
the core of the Internet, or publish routes (ie. get traffic for a
specific IP prefix routed to them.)

Some ways to integrate:
    An experimenter might ask to have an IP prefix allocated to them
    We might run his daemon in the "root" context on our ProtoGENI nodes
        in I2 colos
    We might try to get I2 to agree to 'peer' with these muxes (no idea
        whether or not this is realistic)
    We might try to get I2 to actually carry traffic to/from these
        prefixes on their IP network (we will already have a link to
        this network, but again, no idea if they'd go for this or not)
    If we do have places where I2 (or others) agree to carry traffic,
        the experimenter may ask to get access to one of these, at a
        specific site (this would probably look like allowing a
        node/vserver to use a specific range of IPs, or a VLAN virtual
        interface with a specific VLAN tag #)

Nick has a prefix big enough to support about 8 experimenters (you can't
sub-divide them too small, because that's considered anti-social because
it creates a lot of routes, so you get filtered out.) This seems like a
reasonable number for a start.