[187365] in North American Network Operators' Group
Re: IX ARP Timeout
daemon@ATHENA.MIT.EDU (Owen DeLong)
Thu Jan 28 04:22:04 2016
X-Original-To: nanog@nanog.org
From: Owen DeLong <owen@delong.com>
In-Reply-To: <662200281.11447.1453950132791.JavaMail.mhammett@ThunderFuck>
Date: Thu, 28 Jan 2016 01:20:22 -0800
To: Mike Hammett <nanog@ics-il.net>
Cc: NANOG <nanog@nanog.org>
Errors-To: nanog-bounces@nanog.org
Unless your IX has an unusual amount of churn, a four hour timeout =
really shouldn=E2=80=99t be a problem.
Stale records really shouldn=E2=80=99t be a problem as they should get =
overwritten with gratuitous ARPs when needed.
OTOH, having the ARP be somewhat sticky can not only reduce broadcast =
traffic, but also preserve visibility of what was when trying to =
troubleshoot.
I=E2=80=99m trying to think of a downside to a 400 second ARP timeout =
for an XP and I guess I=E2=80=99m short of creativity at the moment =
because I=E2=80=99m coming up blank.
Owen
> On Jan 27, 2016, at 19:02 , Mike Hammett <nanog@ics-il.net> wrote:
>=20
> So I'm looking at the policies, recommended configurations, etc. of =
other IXes. We try to model a lot of ourselves on what the Europeans do =
(even if we come up short in some areas). I was reading through the =
AMS-IX guide.=20
>=20
> =
https://ams-ix.net/technical/specifications-descriptions/config-guide#3.1=20=
>=20
> They recommend a four hour ARP timeout. Thoughts? Seems a bit =
excessive, but I don't have over 700 networks on my IX. That said, I =
don't have over 700 members on my IX generating a ton of ARP traffic, so =
I'm probably fine recommending a smaller value.=20
>=20
> I understand it's a balance between stale records and ARP volume. Just =
trying to gauge what the community thinks.=20
>=20
>=20
>=20
>=20
> -----=20
> Mike Hammett=20
> Intelligent Computing Solutions=20
> http://www.ics-il.com=20
>=20
> Midwest-IX=20
> http://www.midwest-ix.com=20
