On 24 Sep 2015, at 19:12, Mark Tinka wrote:
On 24/Sep/15 18:45, Wayne Simpson wrote:
On the latest 'HotLINX' newsletter it suggests over 400k routes
available, but you won't get anywhere near that. Most operators don't
even respond to peering requests, and those that do claim we should
be buying transit off of them rather than peering with them.
Sometimes it feels a bit like you are dealing with the old boys club
of ISP's!!
I suppose that from LINX's point of view, the routes available on the
route server + those from folk either not on the route server or
culling
what they send to the route server would equal the claimed number.
LINX's view would be shielded by the "we let members do their own
peering" number.
LINX will be quoting what’s on the collectors, which is the
unobtainable (unless you are AS5459) nirvana of peering with every LINX
member.
It’s a useful number only in so far as it gives some indication of how
much of the Internet is theoretically available there.
The route server is a good indication of what’s easy to get, but ..
ugh .. route servers :( IMO there’s no substitute for a direct BGP
session.
It must be remembered that peering is a commercial decision.
It’s also a decision which rides on real relationships, and getting
out and meeting other peering managers is something you need to do.
Attend LINX meetings, Global & European Peering Forums, etc. Get to know
people.
The ideal for settlement-free peering is that it’s entirely fair and
equitable, however no two networks are alike and that’s really very
very rarely the case - but in anything other than trivial or
inconsequential cases there does need to be some value for both parties.
Everyone wants to have a peering relationship with someone bigger,
that’s plain. What you should consider is why someone bigger would
want to have a peering relationship with you, and what the alternatives
look like for both of you. Often, for them, the alternatives are “buy
it - from me or someone else”.
Peering isn’t free - there’s work to be done agreeing it,
configuring it, securing it, monitoring it, capacity planning for it,
and so on. The routers and associated OSS need to maintain more state
and use more resources. Some operators will point to these as factors.
Some operators may be trying to position themselves in a way that
isn’t compatible with being widely peered. $bignet might not see much
value in peering with $smallnet who peers directly with his customers
and/or when it changes inter-region traffic flows.
You’ll likely find very different attitudes between domestic networks,
foreign networks, and international networks. That is too much of a
generalisation too, though. What is each operator trying to achieve?
does peering with you fit into that story?
Sometimes you might look like good or bad idea to them because of the
traffic the peering would take away from other interconnects. Some peers
have strict traffic ratio requirements, and additional peerings can
either help or harm those. The same, incidentally, goes for sales..
..and could affect pricing.
Once upon a time, long ago when I was younger and more naive, I was
quite chuffed to obtain peering with a large UK carrier at a long since
lost exchange in Manchester. I thought that was great, we moved (what
was then) lots of traffic with them. It turned out, however, that whilst
I had a 1Gb/s port theirs was just 100Mb - and we saturated it. Their
response was not to upgrade - but instead to leave the exchange. It
simply didn’t make economic sense for them to upgrade to carry traffic
to/from that location.
d.
