I've snipped a few bits to make this easier to read. On 1/02/2012, at 11:49 AM, Alastair Johnson wrote:
What concerns me around the peering-is-an-issue flag being raised is that it is often raised a serious issue, possibly overshadowing more significant issues. There also seems to be a general lack of understanding about peering, and the current market situation.
I agree and would suggest that the more significant issue being overshadowed is the cost of transit.
Step 2: If the cost of this transit is sufficiently high then the next step is to look at what routes carry the most traffic and see if those routes are local and can be connected to by cheaper means. One way of doing that is with a direct contract with the network that has those routes[1] but the most common way by far is peering with other networks for exchange of *local* traffic.
To be clear, peering often involves contracts.
Yes I should have been clearer that I was differentiating between one party paying and settlement free.
Agreed, although from what I understand, peering is less about cost at this point than it is about performance. With the costs of international transit dropping dramatically while apparently domestic transit is not, it can be cheaper to push domestic-transit-destined traffic via international paths -- yet I am not aware of anyone actually doing this, despite it being an oft-repeated myth that domestic traffic can trombone 'via LA' or 'via Australia'.
Performance is only an issue because of the poor state of the NZ IXs, but with those being tidied up that should cease to be an issue over time. The tromboning you mention does not have to be actively configured, it is the default route for any content hosted in Auckland (so avoiding any national transit) that is pulled by another network where the two networks don't exchange traffic and neither do their NZ-based upstreams. This can only be prevented if all the resellers of SX bandwidth exchange with every other. Does anyone know if that is the case? What I know does happen from speaking to content providers is that some find it sufficiently cheaper to host their content in the US than in NZ, even though that content has originated in NZ.
Recognising this, our ISP/content provider from Step 1 sets out to peer locally to reduce their transit costs and find that in NZ that the market looks like this:
a. Two of the biggest ISPs don't peer, not even with an extremely tight definition of local, though one of them does offer [1] above.
At least one of those operators does peer as far as I can tell. They might not peer openly at the traditional IXPs, but they do peer, on a settlement-free basis.
From what I gather, and please correct me if I'm wrong here, if you do peer with them you have to pay their capital costs and so, while it might appear settlement-free it isn't really. Besides which you've had to pay far higher capital costs to get there than them since they basically put the peering points where it suits them the most.
Operators that don't peer, well, short of regulating it, what can you do? I suspect their position in the market might change over time post-UFB.
That would be tackling the wrong problem.
b. Those same two ISPs are major transit providers in a very small market of suppliers.
In this case I think you mean they're 'major ISPs' rather than 'major transit providers' -- supplying international transit (or not) is not really relevant to domestic peering. Holding a domestic customerbase captive is an issue and we've seen that happen in the past with some of the peering stand-offs that have occurred, and certainly those happen in the global market too.
I did mean 'major transit providers' but only in the context of the NZ market and I take your point that holding a customerbase captive is another factor at play. cheers Jay -- Jay Daley Chief Executive .nz Registry Services (New Zealand Domain Name Registry Limited) desk: +64 4 931 6977 mobile: +64 21 678840