[195300] in North American Network Operators' Group
Re: BGP peering question
daemon@ATHENA.MIT.EDU (H I Baysal)
Fri Jul 14 10:47:25 2017
X-Original-To: nanog@nanog.org
In-Reply-To: <24462E15-C0C2-4CB5-B854-75B97511F64C@delong.com>
From: H I Baysal <hibaysal@gmail.com>
Date: Fri, 14 Jul 2017 16:46:12 +0200
To: nanog@nanog.org
Errors-To: nanog-bounces+nanog.discuss=bloom-picayune.mit.edu@nanog.org
Hi,
I'm not sure if this is mentioned already but here goes,
You need to understand the difference between peering and a direct
interconnect.
with an interconnect you have to think about is the traffic enough to
"dedicate" a port for that connection on your edge. ( cost of port vs cost
if you would send the traffic over an IX or transit)
peering it does not matter that much, as in someone mentioned to peer as
much as you can, this will give you more control over announcements per
peer and the announcement to the IX.
for example you could not advertise the prefixes that attract alot of
traffic through the IX RR but to individual members. basically it gives you
more control over your announcement to different isps/network over an IX.
however you do have to think about the load on your router on the edge, the
more sessions the more "power" it needs and more processing when things go
wrong or flap.
I hope i didn't give redundant information and it helps.
good luck!!!
Regards,
Halil
2017-07-13 21:27 GMT+02:00 Owen DeLong <owen@delong.com>:
> If you develop a well tuned process for creating BGP sessions and even a
> moderate
> system for monitoring not the individual sessions, but meaningful traffic
> events on
> your network, then, maintaining a large number of peers and a promiscuous
> peering
> policy is not such a daunting process.
>
> As a general rule, promiscuous peering improves efficiency and keeps your
> options for
> traffic delivery open. Restrictive peering generally has the opposite
> effect.
>
> Route servers are a lazy form of promiscuous peering, with an attendant
> fate sharing
> which can produce suboptimal results. YMMV.
>
> I=E2=80=99ve worked for several networks of various sizes and observed th=
e
> industry in general
> for many years. As a general rule, a restrictive peering policy is a grea=
t
> way to lose
> momentum in the market and convert a major ISP into a bit-player (e.g.
> SPRINT), whereas
> promiscuous peering can be a key component in moving a trivial ISP into a
> major player
> in the industry (e.g. HE).
>
> Again, YMMV.
>
> Owen
>
> > On Jul 13, 2017, at 11:04 , Baldur Norddahl <baldur.norddahl@gmail.com>
> wrote:
> >
> > Speaking as a small ISP with 10 to 20 Gbps peak traffic. We are heavy
> > inbound as a pure eyeball network.
> >
> > We use the route servers. We only maintain direct BGP sessions with a f=
ew
> > large peers. Think Google, Netflix, Akamai etc.
> >
> > The reason for this is simply administrative overhead. Every BGP sessio=
n
> > has to be configured and monitored. We know that it will not move a lar=
ge
> > percentage of our traffic. We simply do not have the ressources current=
ly
> > when the gain is so little.
> >
> > Anyone who wants to pass traffic efficiently to us can either use the
> route
> > server or they can peer with Hurricane Electric. The later option will
> get
> > the traffic to us almost as efficiently as peering directly with us. In
> > this sense we outsourced the peering to them.
> >
> > Regards
> >
> > Baldur
> >
> > Den 11. jul. 2017 18.42 skrev "craig washington" <
> > craigwashington01@hotmail.com>:
> >
> >> Hello,
> >>
> >>
> >> Newbie question, what criteria do you look for when you decide that yo=
u
> >> want to peer with someone or if you will accept peering with someone
> from
> >> an ISP point of view.
> >>
> >>
> >> Thanks.
> >>
> >>
> >>
> >>
>
>
--=20
Met vriendelijke groet / kind regards,
Halil Ibrahim Baysal
T: +31 (0)6 20 14 20 79
E: hibaysal@gmail.com
