On 28/Jan/16 03:36, Owen DeLong wrote:
I disagree with this last part.
I realize that the common wisdom among execs at so-called tier-1 providers is that refusing SFI protects their revenue stream, but I believe it’s not true.
In fact, I think that a willingness to peer with your customers and anyone else on the internet wherever you can do so for very little cost (for example, where it’s just one more peering session at an IXP, no additional port cost, circuit, XC, etc.) settlement free can only increase your business.
IMHO, a truly good tier-1 will charge for transit, set their metrics and prefs such that their paid ports are preferred over their non-revenue ports, and provides peer routes only on the SFIs.
This turns out to be mostly a win-win situation for everyone, including the tier-1 in the long run.
I tend to agree with Owen on this one. We, last year, transitioned from selective to open peering - despite our scope - in the region we serve (primarily Africa). It has only improved the quality of our service (a great deal of Africa still exchanges traffic in Europe), lowered costs, made customers happy and generated a lot of community goodwill. Obviously, we do not provide free transit across SFI ports, and we have practical implementations in place to ensure that we only handle customer traffic through customer-facing links, removing the potential of handling customer traffic through peering links (particularly with customers who are multi-homed to you and another SFI peer of yours). While I do not disagree that larger providers looking to protect their revenues is an economically-sound objective, I think the typical peering policies of old do not entirely hold up in 2016. Mark.