Multicast traffic % in enterprise network ?

Wed Aug 8 21:53:19 UTC 2018

On Wed, 8 Aug 2018 18:49:52 +0000
"Mankamana Mishra (mankamis) via NANOG" <nanog at nanog.org> wrote:

>   *   If there is any data which can provide what % of traffic is
> multicast traffic. And if multicast is removed, how much unicast
> traffic it would add up?

Good question about the volume (and frequency).  I will see if I can
measure here, but all I can say if it helps any, is that is still
fairly widely used in the enterprise for "imaging" systems.  That is,
updating a set of systems and apps simultaneously from a single source.

> *   Since this forum has people from deployment area, I would love
> to know if there is real deployment problems or its pain to deploy
> multicast.

In my experience, real world IP multicast experience and expertise is
almost non-existent.  This is the big problem potentially, because it is
fairly popular on some campuses for niche applications, but it tends to
just work once setup and rarely changes.  For many R&E institutions,
the setup was done a long time ago and people rarely touch it now.

However, I rarely see IP multicast-related problems anymore.  It
helps that interdomain multicast is dying.  Traditional PIM-SM
with one or more anycast RPs isn't too complicated, and I can't
remember the last time anything out of the ordinary came up.

A decade or more ago at another institution that did IP/TV over IP
multicast to dorms, we'd run into head-banging-against-the-wall problems
about once a year.  Very painful experiences usually due to buggy
router code or interesting switch/router behavior with IGMP-snooping
and/or joins/leaves.  Those days seem to be behind us.

> These questions is to work / discussion in IETF to see what is pain
> points for multicast, and how can we simplify it.

I can't think of too much that should change now.  If I want to simplify
and secure things a little more, I'd probably look at better limiting
IGMP and PIM joins.  Limiting as much of the possibility for anyone or
anything to instantiate state in the network devices.  SSM without the
all the potential source-specific state would be nice.

John