With routing, inbound and outbound are handled completely separately. We can dictate outbound routing by favouring certain paths or adjusting various factors with BGP filters. Just remember, what you see as a traceroute going to a server is not always the route that network chooses to return traffic on (one of the caveats of BGP). Inbound is a little more complex, we only have two tools; communities (where we explicitly tell an upstream what to do) and AS path length (making a specific path look longer or shorter based on AS hops). We have a strategy for managing DDOS with this new provider in tow. With regards to inbound traffic, our strategy in the first few days once we go live will be to advertise equally across all upstreams in the region (CPT) to see how traffic is shifted around and what users are seeing. We'll bring up a bunch of probes globally and run some tests inbound to CPT prefixes as well. Once we have a clear picture of how the rest of the world sees us behind this new upstream, we can start tweaking; so user feedback during this week or so soak period will be vital. We'll keep you posted along the way.
Sounds good. Will this alleviate the black hole issues for those of us unlucky enough to be on Octotel? I forgot about about it and an automated download killed my connection during an SSH session yesterday, so I've limited NNTP again for now.