wg-multicast - Re: Turn it on everywhere?
Subject: All things related to multicast
List archive
- From: John Zwiebel <>
- To: Toerless Eckert <>
- Cc: ,
- Subject: Re: Turn it on everywhere?
- Date: Thu, 04 Oct 2001 11:10:29 -0700
- Organization: Procket Networks
Toerless Eckert wrote:
>
> > > Hey all,
> > > So what do you all think the repercussions will be in terms of network
> > > operation if we take this step?
I'd like to echo what Toerless said.
I've always pushed for turning it on everywhere.
The problems generated by trying to control the multicast path
and pretending that PIM isn't dependent on unicast routing outweigh
by far the problems you might encounter with a multicast session using
up all the BW on a given link.
Remember that, for at least right now, there isn't that much multicast
traffic being offered. Most folks need to get comfortable with how it
(PIM) works and the deficiencies of the protocol need to be found. That
can't be done if you think you can go "half-way". However, I'm greatly
encouraged by the new spec that Fenner et. al. have put out. There are
a couple of places where maybe things aren't quite right, but so
far, it has been my experience that a different implementation choice
hasn't prevented interoperability -- and those area don't happen that
often in "real life".
What I'm saying is that from a protocol perspective, it is now even
safer to "turn it on everywhere" since I believe we are very close to
a PIMv2 protocol that has nailed down all of the problems.
This shouldn't be interpreted as meaning you can run an 11.0 image
successfully on the same network as a 12.2 image, or that PIMv1 and
PIMv2 won't have problems (especially with asserts). Nor should you
believe that multicast over LANE is going to work well.
You now have to decide between multicast and <X> where <X> might
be MPLS or VPNs or "something else". It isn't that multicast
can't work with these "something else" items, its that they don't
work well "right now". Which do you believe will provide you the
best results "right now" and which do you want to leave others to
work out the "bleeding edge" for you.
If you ever want to have cheap video/audio content over IP, that allows
anyone with $10K (or much less) to be in a position to provide quality
content over the internet, I believe you need to choose multicast now.
If you want to have content provided by the same large entertainment
conglomerates that control it now, then don't turn it on. Multicast
will get deployed, of that I have no doubt. The question is when and
what will be available through that medium.
Are you taking a risk turning it on? Probably. What is the risk you
are taking by keeping it off?
I hope you all recognize that SSM deployment is going to greatly ease
your potential problems.
z
- Turn it on everywhere?, Tim Ward, 10/03/2001
- Re: Turn it on everywhere?, Tony Rimovsky, 10/03/2001
- Re: Turn it on everywhere?, Joel Jaeggli, 10/03/2001
- Re: Turn it on everywhere?, Guy Almes, 10/03/2001
- <Possible follow-up(s)>
- Re: Turn it on everywhere?, Alan Crosswell, 10/03/2001
- Re: Turn it on everywhere?, Toerless Eckert, 10/04/2001
- Re: Turn it on everywhere?, John Zwiebel, 10/04/2001
- Re: Turn it on everywhere?, Toerless Eckert, 10/04/2001
- Re: Turn it on everywhere?, John Zwiebel, 10/04/2001
- Re: Turn it on everywhere?, Toerless Eckert, 10/04/2001
- Re: Turn it on everywhere?, Alan Crosswell, 10/03/2001
- Re: Turn it on everywhere?, ken lindahl, 10/04/2001
- Re: Turn it on everywhere?, Brian Haberman, 10/04/2001
- RE: Turn it on everywhere?, Richard Mavrogeanes, 10/08/2001
Archive powered by MHonArc 2.6.16.