The BBC launched a multicasting trial last year along with ITV
(http://support.bbc.co.uk/multicast/streams.html), and more recently some of
the commercial radio companies have joined it, but the list of participating
ISPs doesn't seem to have grown much:

http://support.bbc.co.uk/multicast/isps.html

and the big ISPs are conspicuous by their absence. So I was wondering what
people on here think is the prospect for multicasting over the next 2-3
years, and whether people think that any of the big ISPs are going to
convert their networks to support multicasting. Also, what is currently
stopping the bigger ISPs from converting their networks to support
multicasting?

I've also got a couple of questions about BT's 21CN: will it be an IPv6
network and therefore natively support multicasting? And if so, once an area
has been converted to 21CN would broadband connections in that area then
natively support multicasting, or would it still require the individual ISPs
to convert their equipment as well?


--
Steve - www.digitalradiotech.co.uk - Digital Radio News & Info

Find the cheapest Freeview & DAB prices:
http://www.digitalradiotech.co.uk/fr..._receivers.php
http://www.digitalradiotech.co.uk/dab/dab_radios.php

DAB sounds worse than FM <dab.is@dead> wrote:
AIUI.
For a non LLU ISP, using BT Centrals from the customers to the NOC,
there is no native support of multicasting. For ISPs with their own
bandwidth providers to the NOC, this may change.

A gigabyte costs about a pound to provide in peak time (as a
proportion of a BT central. (622 mbits, over a million a year to rent)

The ISP can possibly have multicast transmission to their NOC
but that's as far as it goes.

The connections to the end users are over completely isolated virtual
circuits, that only contend for bandwidth.

There is no mechanism in BTs current network for multicast traffic to
'break in' at the end of the circuit near the customer, it all needs
sent individually, even if there are thousands of copys of the same
packet going to the same exchange.

How 21CN changes this I don't know.
In principle, it could change.

The only 'near term' - 2-3 years change in the market that I'm aware of
is BT saying they are going to drop the price of centrals a bit.

Ian Stirling wrote:

Okay, thanks for the above explanation.



I think I'm right in saying that 21CN will be an IPv6 network, and IPv6 does
natively support multicasting:

http://en.wikipedia.org/wiki/Ipv6#Multicast

"Multicast is part of the base protocol suite in IPv6. This is in opposition
to IPv4, where multicast is optional.

Most environments do not currently have their network infrastructures
configured to route multicast; that is - the link-scoped aspect of multicast
will work but the site-scope, organization-scope and global-scope multicast
will not be routed."

So I think the crucial question is: when 21CN has been rolled out to a
certain area, will ADSL users in that area then be able to use multicasting,
or would it still require the individual ISPs to change their networks as
well?



Okay.


--
Steve - www.digitalradiotech.co.uk - Digital Radio News & Info

Find the cheapest Freeview & DAB prices:
http://www.digitalradiotech.co.uk/fr..._receivers.php
http://www.digitalradiotech.co.uk/dab/dab_radios.php

DAB sounds worse than FM <dab.is@dead> wrote:
Maybe.
Yes, it will be an IPv6 network, I don't think however it's been
revealed exactly how this is implemented.

If it's implemented as some sort of PPPoIP, or other virtual circuit,
from the customers ADSL linecard, to the ISP, then nothing - apart from
perhaps the cost - changes.

If each ADSL linecard has its own IP, connected by a pure IP network, there
are several implications.

How does charging work?

Might a gigabyte to my neighbour (with the linecard next to me) be
significatly cheaper than a gigabyte to London?

This would make peer-peer realtime video and radio of some sorts very
possible, and much, much kinder to the network than currently.

On 2007-01-22, Ian Stirling <root@mauve.demon.co.uk> wrote:
That's pretty funny. If you fetch this

http://www.soe.ucsc.edu/projects/pof/WSJ.pdf

you'll notice on page 2 that the wholesale price for 155 Mbps from
London to New York is US$3,400 per month. Doing the math this
means that a London<->New York 622 Mbps circuit would rent for
a bit under $165,000 per year. I guess it commands a 10x premium
when you call it a "central".

Or perhaps that is the problem? R&D for ATM switches stopped in about
1995, so anyone who is still using them for something is paying a
huge premium for hardware compared to packet equipment. Who ever
decided that the way to support DSL was to build an ATM backhaul
network made a huge mistake (I'll bet they got promoted afterward,
too).

Dennis Ferguson

On Tue, 23 Jan 2007 11:21:37 -0600, Dennis Ferguson
<dcferguson@pacbell.net> wrote:

no, it commands a premium when it pays for connection to 5,500
telephone exchanges rather than a single point to point link over the
Atlantic where there is plenty of surplus capacity - much of it
acquired from companies that went bust.

Phil

On 2007-01-22, DAB sounds worse than FM <dab.is@dead> wrote:
I've done development work on multicast routing protocols and I can tell
you that multicast routing is a really tough, difficult to scale, problem.
I've seen this stuff usefully deployed in smaller networks, but in a
large network (let alone, say, an entire country) the routers are already
working hard and turning on multicast routing is like adding insult to
injury. That's not to say it won't be done; things can always be made to
work if you are willing to suffer long enough. I just wouldn't rank the
likelihood of multicast deployment on a large scale at a high probability.

Note that the problem of routing IPv6 multicast is identical to the problem
of routing IPv4 multicast; if you don't want to run one of them you are
unlikely to want to run the other. And I'm not sure about the need to
"convert equipment" to run multicast (with the exception of certain
backhaul edge networks of unfortunate design); routers have supported
the functionality for years, it is mostly just that no one wants to
turn it on.

Dennis Ferguson

On 2007-01-23, Phil Thompson <phil.thompson@spamcop.net> wrote:
Despite that, bandwidth is still inherently more expensive on the
transatlantic route because of the limitations on the use of WDM there
which don't exist on land routes.

In any case, I think I otherwise might have made your point in the part of
the post you deleted, where I said:

I don't think 5,500 is the problem, though how the 5,500 are connected
may be.

Dennis Ferguson

Ian Stirling wrote:

Right.



Okay, thanks again for the explanation.


--
Steve - www.digitalradiotech.co.uk - Digital Radio News & Info

Find the cheapest Freeview & DAB prices:
http://www.digitalradiotech.co.uk/fr..._receivers.php
http://www.digitalradiotech.co.uk/dab/dab_radios.php

Dennis Ferguson wrote:

If an ISP limited multicasting support to just the 30 or so channels on the
BBC's multicasting trial:

http://support.bbc.co.uk/multicast/streams.html

rather than supporting an unlimited number of multicast streams, would that
make the problem more tractable?



Do you think that as the demand for live streamed TV increases a time will
come when it becomes in the ISPs' own interest to support multicasting? The
alternative is having shitloads of unicast and P2P streams flying around,
which I'd have thought would provide a greater strain on bandwidth/routing
(although admittedly I don't actually understand the situation very well).



Right.



Fair enough.


--
Steve - www.digitalradiotech.co.uk - Digital Radio News & Info

Find the cheapest Freeview & DAB prices:
http://www.digitalradiotech.co.uk/fr..._receivers.php
http://www.digitalradiotech.co.uk/dab/dab_radios.php

On Tue, 23 Jan 2007 13:41:12 -0600, Dennis Ferguson
<dcferguson@pacbell.net> wrote:

it isn't expensive if you buy it for $1 from the company that went
bust after laying it, and the real driver of price is competition -
plenty of that across the atlantic.

speculation as to whether ATM kit is an issue. Having to lay in new
connections of modest bandwidth to 5500 exchanges is bound to result
in a high cost per unit total capacity - trenching a fibre a few miles
to feed a rural exchange with a few hundred lines must have a fairly
high $/Mbps outcome.

Phil

Dennis Ferguson <dcferguson@pacbell.net> wrote:
Whatever it's called, BT are the only suppliers of the glue between the
customer that's on an exchange you don't have an LLU presence in, and
your operations centre.
They charge what they charge, and there is no competition.

ATM is not unreasonable, if you assume there will only be one ISP, and
you don't envision multicast being a big thing.

An 'proper' IP network raises all sorts of hastles that you simply don't
get with this design.

"DAB sounds worse than FM" <dab.is@dead> wrote in message
news:7eHth.78861$z01.45020@newsfe3-gui.ntli.net...

And life is better still for minimising complexity if you limit where those
streams enter the network.

At that point you basically have IP multicast pretending to be a cable TV
distribution system - which is probably exactly what the broadcasters would
want. Ironically, Cisco et al are busy pushing this distribution network
design to cable companies all over the world so they can "go IP"......

The broadcasters would love this as it moves large scale plumbing out of
their part of the network - they send 1 copy of each "live" stream (or maybe
1 for each resolution / data rate) and let the ISPs deal with the complexity
of making copies of the packets.

Much better for the broadcaster to send 10 flows each to 100 ISPs than
100,000 point to point flows to 100,000 users.....

The issue for generic IP multicast such as PIM SM or DM is that the general
purpose protocols build route management structures on the fly in every IP
router that touches each flow, on a per flow basis, when triggered by data
flow.
So processor load ,memory and other resource use in each router go up as the
number of streams go up, even if the level of multicast traffic stays
constant.

A better structure for a "broadcast" style format is to use some of the
deliberately less flexible design templates and protocols such as source
specific multicast.

But - this gives you a "broadcast" style traffic model, where sources are
restricted, or makes some flows follow non shortest path routes around the
network - which starts to eat into the bandwidth savings.

It gets even worse when you have more than a simple native IP network - say
you use tunnelling protocols or MPLS.
the obvious people who might want to do this are ISPs who are also
broadcasters - Sky, NTL?

Agreed.

And dont forget "something" has to make copies of packets at some point -
even if multicast is deployed all the way to consumers, they nearly all live
on the end of separate DSL lines.

All the argument here is about is where the copying happens, and in whose
equipment.

So - in a source server (at the broadcaster)
in an ISP core network (multicast part way, then convert to point to point)
at the broadband aggragation router (where separate PVCs or IP tunnel per
user hit the ISP)
at the DSLAM (if the DSLAM understand IP multicast, and BT choose to turn it
on).
Regards

stephen_hope@xyzworld.com - replace xyz with ntl

"Dennis Ferguson" <dcferguson@pacbell.net> wrote in message
news:slrnercp6j.lb.dcferguson@akit-ferguson.com...
FWIW ATM switches are perfectly capable of supporting multicast in the
hardware (ie at the cell level) - that was one of the reasons broadcasters
used to use ATM a lot.

Now - turning it on and making IP multicast co-operate - different issue,
but probably doable if you used different PVCs for point to point and point
to multipoint.

But that probably destroys the scaling properties.

Anyhow - BT charges by the Mbps / hour to ISPs - so why enable a complicated
service where the main driver is reduce traffic flow?

They dont gain from it but would get penalised.....
Regards

stephen_hope@xyzworld.com - replace xyz with ntl

stephen wrote:

That certainly seems to be the biggest problem, but Ofcom has just given the
BBC the green light to start distributing live TV streams (but it objected
to some of the BBC's plans for its iPlayer, such as allowing people to
download full series of programmes
http://www.ofcom.org.uk/tv/ifi/marke..._ondemand.pdf),
so it's looking inevitable that the BBC is going to start distributing live
streams, so which technology out of multicasting and P2P would create the
lowest bandwidth requirements for the ISPs - they're going to have to take a
hit on bandwidth, so if multicasting requires lower bandwidth than P2P you'd
expect them to get their shit together and support multicasting.


--
Steve - www.digitalradiotech.co.uk - Digital Radio News & Info

Find the cheapest Freeview & DAB prices:
http://www.digitalradiotech.co.uk/fr..._receivers.php
http://www.digitalradiotech.co.uk/dab/dab_radios.php

stephen wrote:

Yes, that's exactly how the BBC multicasting trial works - they have
"guaranteed" streams to each participating ISP, which then forwards the
streams to any of its customers that have requested them.



That's understandable for some countries, because some countries only have a
small percentage of people that watch terrestrial TV (e.g Holland only has
something like 5% of the population that watch terrestrial TV and the rest
watch cable or satellite). Unfortunately that leaves us screwed...



Absolutely. And the BBC had said that they would only distribute live
streams via multicasting because of bandwidth constraints, but they're
looking at using P2P live streaming as well now.



That's what I thought. But do you think that limiting the number of streams
reduces the processor load etc enough for the ISPs to think it's feasible?
And if so, why do you think so few ISPs are participating in the BBC's
trial?



Right.



Okay.



Indeed, but the problem with them is that Sky hates the BBC, and NTL is
probably not too keen on helping the BBC either, because they'd prefer it if
people subscribe to cable instead. And BT would also probably prefer not to
help the BBC in case it affects their BT Vision thing as well.



Which is the most efficient technology for the ISPs in terms of bandwidth
out of multicasting and P2P live streaming?


--
Steve - www.digitalradiotech.co.uk - Digital Radio News & Info

Find the cheapest Freeview & DAB prices:
http://www.digitalradiotech.co.uk/fr..._receivers.php
http://www.digitalradiotech.co.uk/dab/dab_radios.php

"DAB sounds worse than FM" <dab.is@dead> wrote in message
newsDQth.81456$Qa6.23899@newsfe6-gui.ntli.net...
i think that it is going to be a balance.

a big ISP probably peers directly with the BBC
http://support.bbc.co.uk/support/peering/

so there isnt a per packet bill to someone for the traffic (but they will
have to build a "big enough" pipe, and then backhaul the traffic to the rest
of their network).

But - either the ISP "does" multicast across its network, or at some point
it converts the traffic to unicast streams - and basically has the stream
replication problem.

So the real Q is - whats costs most, more bandwidth / bigger pipes or clever
networking / servers?
Regards

stephen_hope@xyzworld.com - replace xyz with ntl

stephen wrote:

Considering the cost of bandwidth, I'd imagine that using clever
networking/servers would be far more cost efficient, but that's only a
guess.


--
Steve - www.digitalradiotech.co.uk - Digital Radio News & Info

Find the cheapest Freeview & DAB prices:
http://www.digitalradiotech.co.uk/fr..._receivers.php
http://www.digitalradiotech.co.uk/dab/dab_radios.php

DAB sounds worse than FM <dab.is@dead> wrote:
Clever networking - at the moment - means installing equipment in every
exchange that can do 'clever' stuff.
You can't do anything clever with BTs network kit.