RIPE 87

Archives

Jad El Cham - 2023-11-28 17:56:39
Hi everyone, I'm Jad El Cham from the RIPE NCC. This chat panel is meant for discussion ONLY. If you have questions for the speaker and you want the session chair to read it out, please write it in the Q&A window also stating your affiliation. Otherwise, you can ask questions using the microphone icon.

Please note that all chat transcripts will be archived and made available to the public at https://ripe87.ripe.net/.
The RIPE Code of Conduct: https://www.ripe.net/publications/docs/ripe-766.

Tim Chown - 2023-11-28 18:05:11
no R&E perspective, maybe it's all gone...

Tim Chown - 2023-11-28 18:23:26
so users would not want to be tied to a stb.

Rudolf van der Berg - 2023-11-28 18:23:45
neither would operators ;-)

Tim Chown - 2023-11-28 18:24:19
the ietf discussion on multicast always comes back to the lack of content.

Tim Chown - 2023-11-28 18:24:55
no mention of bier!

Tim Chown - 2023-11-28 18:26:10
Tom seems to be in every queue :)

Tim Chown - 2023-11-28 18:26:25
or is Tom multicast to all sessions?

Viliyana Ivanova - 2023-11-28 18:26:30
yes, STB limits you and the one provided by the operator is usually the cheapest one unfortunately

Rob Evans - 2023-11-28 18:26:39
I also see an R&E representative in the queue... :)

Tim Chown - 2023-11-28 18:27:06
one with content? :)

Rudolf van der Berg - 2023-11-28 18:27:10
Mbone was once the big promise ;-)

Tim Chown - 2023-11-28 18:27:26
should be all v6 / bier / ssm now... but...

Tim Chown - 2023-11-28 18:29:35
2000, a newbie!

Viliyana Ivanova - 2023-11-28 18:29:52
2001 :)

Tim Chown - 2023-11-28 18:30:05
97 or 98, i remember early vic and rat.

Rudolf van der Berg - 2023-11-28 18:31:10
We had Mbone at Twente university in 95

Tim Chown - 2023-11-28 18:31:35
nice! that's almost pre web!

Tim Chown - 2023-11-28 18:32:11
much of eumedcast is tunnelled though.

Rob Evans - 2023-11-28 18:32:42
It is now, it wasn't all tunnelled at the start.

Rudolf van der Berg - 2023-11-28 18:33:24
https://www.streamingmedia.com/Articles/Editorial/Featured-Articles/The-Return-of-Multicast-Why-it-Succeeds-in-a-Live-Linear-World-108621.aspx some people have hope

Tim Chown - 2023-11-28 18:33:26
yeah remove multicast,

Tim Chown - 2023-11-28 18:33:45
this is a long question being asked

Rudolf van der Berg - 2023-11-28 18:34:12
https://www.cl.cam.ac.uk/teaching/1213/R02/papers/ipj1.pdf mind you in 1999 there were issues already

Tim Chown - 2023-11-28 18:34:17
if tunnelling, AMT works, but no one is providing content.

Tim Chown - 2023-11-28 18:34:51
yeah that's the UCL multicast, they write vic and rat.

Tim Chown - 2023-11-28 18:35:01
wrote

Rudolf van der Berg - 2023-11-28 18:35:02
In 2001 some people saw a future for multicast https://citeseerx.ist.psu.edu/document?repid=rep1&type=pdf&doi=a5b0a7b129f08dc38b352f1b8dd435c6ddf4ad9b

Tim Chown - 2023-11-28 18:35:54
i remember running a 30-way ipv6 multicast vic/rat session in 2001/02 or so. worked surprisingly well, though all tunnels.

Tim Chown - 2023-11-28 18:36:03
m6bone

Rob Evans - 2023-11-28 18:36:27
Even last year one of the major CDNs was saying "We concluded our research project and having got very positive results & feedback from 50+ external parties we decided to move Multicast to the next stage." Haven't heard much since then.

Tim Chown - 2023-11-28 18:39:31
at least there's more people in this BoF than ever attend the ietf mboned wg

Toerless Eckert - 2023-11-28 18:39:32
Rob: M&A issue

Viliyana Ivanova - 2023-11-28 18:39:36
Multicast definitely helps reduce peaks during football championships for example, but we easily compensate that with cache servers, that are positioned based on the telecom network and needs. And for non-live content you still need to have the caches anyways. And this way you do not need to support two protocols. The overall advantage to save traffic in a fiber optic world is less than the complexity of the network once needs to support. Also multicast ABR is not that flowless to do and then when the user home setup is bad, then it just does not work

Tim Chown - 2023-11-28 18:39:43
and the ietf often bemoans the lack of operators present

Tim Chown - 2023-11-28 18:44:06
use ssm

Toerless Eckert - 2023-11-28 18:44:43
Villyana: Avoiding those cache server buildout for worst case peak was one of the business argument for multicast: outsource replication to ISP routers that are present anyhow

Rudolf van der Berg - 2023-11-28 18:47:08
It doesn't save that much in caches, because you have to then have routers that manage state for lots of streams, without knowing the demand

Viliyana Ivanova - 2023-11-28 18:47:15
Toerless: how will you handle then the non-live traffic?

Toerless Eckert - 2023-11-28 18:48:44
Villyana: for the short tail, there are intelligent codec based mechanisms pioneered in 2002 by Digital Fountain.

Viliyana Ivanova - 2023-11-28 18:49:54
Toerless: but only for pause/restart? not for recordings

Rudolf van der Berg - 2023-11-28 18:49:59
Multicast was implemented in 3G, 4G and 5G standard documents, but not in handsets and networks.

Toerless Eckert - 2023-11-28 18:50:55
Rudolf: You could do fan-out demand based switchover from unicast/tunnel to in-network replication, you could do BIER for stateless replication, but the problem is that you now talk about a problem at higher layers (libraries, applications), not the network anymore.

Toerless Eckert - 2023-11-28 18:51:41
Rudolf: 3GPP mostly did their own proprietary multicast mapping to radio broadcast, but AFAIK very little if no IP multicast. So it's an island solution.

Khwaja Zubair Sediqi - 2023-11-28 18:51:47

I was wondering how was the user experience, the latency, video resolution, during peak hours watching live streaming events using multicast services. Any idea on that?

Toerless Eckert - 2023-11-28 18:53:06
Khwaja: most IPTV based TV services today are still using IP multicast for at least part of the transmission path. So ask any of the hundreds of millions of users world-wide of those TV services.

Jad El Cham - 2023-11-28 18:54:06
Hi Khawaja, please write your question in the Q&A panel (icon with question mark) or please wait until the end of the talk and ask your question using audio. Thank you!

Rudolf van der Berg - 2023-11-28 18:54:46
Toerless: You could do lots of things. They've been proposed and tried at every level of the network, but failed in the practice.

Tim Chown - 2023-11-28 18:55:20
ripe needs the ietf meetecho queue system.

Toerless Eckert - 2023-11-28 18:55:38
Rudolf: Mayority of technologies fail due to bad execution, and most often they've also tried at the wrong time and place.

Toerless Eckert - 2023-11-28 18:55:54
Tim: just raise your hand in meetecho

Rudolf van der Berg - 2023-11-28 18:56:50
The reality is that unicast with a cache deep in the network is easier. Latest caching boxes go up to a terabit/s which means that you need one per 1 million customers for stuff like champions league.

Tim Chown - 2023-11-28 18:57:20
oh, so dont use the Q&A box like they ask you to?

Tim Chown - 2023-11-28 18:57:49
anyway, dinnertime :)

Jad El Cham - 2023-11-28 18:57:50
You can either write the question in the Q&A or as for audio

Jad El Cham - 2023-11-28 18:58:03
ask*

Rob Evans - 2023-11-28 18:58:05
Generally yes, use the Q&A box, but I'm not sure the BOF chairs are reading them...

Tim Chown - 2023-11-28 18:58:15
no, its not very good chairing

Tim Chown - 2023-11-28 18:58:48
if you use ssm, addressing is not an issue

Viliyana Ivanova - 2023-11-28 18:59:17
👍

Tim Chown - 2023-11-28 18:59:23
OR IPv6!!

Tim Chown - 2023-11-28 18:59:31
or IPv6 SSM

Viliyana Ivanova - 2023-11-28 19:00:27
Some IXPs do that already - channel aggregation and delivery. And I think it is very nice to have

Tim Chown - 2023-11-28 19:01:00
the thing with using meetecho queue is that people in the room have to use it too, and they arent

Rudolf van der Berg - 2023-11-28 19:01:52
Toerless your slides on multicast from 1995 aren't available anymore https://www4.cs.fau.de/~eckert/

Jad El Cham - 2023-11-28 19:08:52
This session has now ended. The next sessions are Address Policy and DNS and they will start tomorrow at 09:00 UTC+1 . More info on the RIPE 87 meeting plan: https://ripe87.ripe.net/programme/meeting-plan/

Azael Fernandez - 2023-11-28 19:09:04
Thank you.

Rudolf van der Berg - 2023-11-28 19:09:10
thanks

Viliyana Ivanova - 2023-11-28 19:09:17
Thank you, enjoy the social!