[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Draft on ISP broad-band deployment scenarios
Hello Pekka,
Please see my response between &&&.
At 10:32 AM 9/24/2004 +0300, Pekka Savola wrote:
Inline...
On Thu, 23 Sep 2004, Salman Asadullah wrote:
> >There are native deployments today, but AFAIK they have
been
> >relatively small scale (thousands of users).
> >
> >The crux is whether the ISP wants to (try to) offer service to
*all*
> >of its customers, requiring no hardware upgrades or purchases,
with no
> >assumptions about what the user has deployed -- or just those
which
> >are OK the assumptions, upgrades/purchases, etc.
> >
> >I'm concerned that if we focus too much on "native
only", only
> >(relatively) few users will likely be able to get the
service.
> >
> >It's true that it may be difficult for an ISP to generate
revenue out
> >of e.g., tunneling based service, offered to anyone, but if some
ISP's
> >are still interested in this kind of approach, would it be
worth
> >considering?
>
> &&&
> This is true but we have observed many of the BB ISPs are interested
in
> deploying native IPv6, driven by new services such as VoD using
Multicast
> and etc. Some of the examples for these BB ISPs, who are deploying
IPv6
> are NTT in Japan, Spacenet in Germany, Dolphin in Switzerland, Nerim
in
> France, XS4ALL in The Netherlands and etc.
Sure. But do these ISPs use a link-layer which supports multicast,
e.g., bridged-mode DSL (I think it does..)?
That is, if you do something like L2TPoE (which I think many are
doing), doesn't that require that multicast transmission be
duplicated (on a lower layer) in any case, causing equal amount of
traffic as a tunneling based distribution?
&&&
Well, some might go for L2TPoE but not necessarily. We discussed that
option for the sake of completeness and in case some SPs really, really
like that model but .... if they do go for L2TPoE then yes, they will not
gain much new. On the other hand we talked about the alternatives where
the IPv6 traffic is routed while the IPv4 is still L2TPoE. That is for
example the case of one of the biggest BB ISP who are providing IPv6
commercial services, where IPv4 is L2TP but IPv6 is routed. They did that
for the very reason as they wanted to deploy mcast over it (what you
referring to).
&&&
> Our intent in this draft is to the make the worldwide BB ISP
community
> aware of possible options available to deploy IPv6 natively
today.
>
> We agree that for some ISPs the native IPv6 solution would not be
feasible
> due to several reasons as you mentioned previously, with the
information
> provided in our draft, some ISPs might realize that the extra
investment
> for a native deployment might make sense for their long term
> goals. Besides, for some countries this might be a viable option,
think of
> Japan, Korea and some other countries where the ISPs are getting tax
cuts
> from the government, if they deploy IPv6.
> &&&
Extra investment is always possible, and many will certainly do it.
The question is just how the ISPs would transition from
"v4-only" to
"v4+v6 natively". My argument is that it would be easier for
the ISPs
to start with tunneling and native v6 where possible, than to require
through-out upgrade to native v6.
That would reduce the number of up-front investment requirement, but
would not prevent from making the investment when it's clear that
there will be demand for the service.
That would also make it more lucrative to the customers whcih have
boxes that don't support v6. They wouldn't have to upgrade
immediately to get that service..
&&&
Ahh ....
We totally agree on this and finally understand where you are coming from
:-)
In order to achieve the above and make this document complete and
beneficial for all, we should include the tunneling options as well for
remaining environment, FTTH, DSL and Wireless. Tunnelling options
for Cable is already discussed (off course we have noted all the comments
and feedback which we will address in the next revision of the
document).
So, we think we would need couple of more things:
1. A general summary section up front stating the intent of the document
and available deployment options (tunneling and native).
2. Add a general tunneling section for FTTH, DSL and Wireless sections as
another option. We believe this section would be general since we can
simply talk about the different tunnelling solutions available and
reasons of choosing one or the other and point the reader to the
respective tunnel draft/rfc for details.
This will make the document complete with both options available in one
place for all the BB ISPs. Depending on what stage the BB ISPs are,
they can chose whatever they want.
Does this sounds like a good plan ?
&&&
> > > >This is one approach, as
noted at the start of the message; another is
> > > >providing tunneling support. Both could be feasible,
depending on
> > > >deployment, but the latter is more generic at
least.
> > >
> > > +++++++++++++++
> > > Right, so cable operators currently do provide a GWR which
is managed by
> > > them. In order
> > > to deploy IPv6 the operator can upgrade or replace this
GWR with a v6
> > > capable device.
> > > +++++++++++++++
> >
> >Right. But my practical concern is that if it costs the ISP XXX
XXX $
> >to replace the GWR's with ipv6-supporting versions, but XXXX $
(XXX
> >XXX >> XXXX) to provide tunneling support, I'd guess most
ISPs would
> >be interested in going for tunneling first, and just looking
at
> >replacing GWR's based on their natural EOL schedule.
>
> &&&
> You are correct, the first step in some cases will be tunneling (for
> example cable).
>
> But in other cases BB ISPs are taking the next step and inclined
more
> towards the native deployments
>
> Also the extra cost might be justified by the:
>
> - Services provided and charges ($$ earned) over IPv6 like VoD,
video
> streaming, IPT and etc.
> - Tax cuts received by the governments for IPv6 deployments
> - and etc.
> &&&
See above. I think all services which are available over native v6
can be offered, at least initially, through tunneled v6. If it's
still possible to get the service profits, and reduce the amount of
initial investment, why wouldn't the ISPs do
it?
&&&
Sure. Agreed upon.
&&&
>
> > > +++++++++++++++
> > > This is currently a normal practice for Cable Operator's
business
> > customers.
> > > There are only a handful of these customer routes which
are summarized and
> > > redistributed into the Cable Operator's IGP. This is not
really meant for
> > > residential
> > > customers which account for majority of the customer
base.
> > > +++++++++++++++
> >
> >But what I don't understand is why the business customers would
1)
> >need dynamic routing in the first place (what is the actual
reason for
> >using a routing protocol?),
>
> &&&
> The cable business customers in todays world get a /27 or something
> similar. The issue is how does the CMTS keep track of which
business
> customer owns that particular /27 and how does it get advertised to
the
> rest of the world. One way to do this is to use static routing but
the
> problem is that if the cable modem goes down, the CMTS will end up
black
> holing that traffic and vise versa, thus requiring an IGP. There are
> several reasons of using RIPv2, out of which a simple one is that
there are
> hardly any CM/GWR which support OSPF.
But the point is, unless those business customers have backup
connectivity, over which they also use the same IGP, the packets
will
be discarded in any case, so it doesn't seem to matter whether
static
routing does it whether they are discarded due to lack of a route.
The only difference (unless the business is multihomed, and I think
you said only few were) is whether the CMTS sends back an ICMP
unreachable message to the hosts trying to connect to the business,
before it discards the actual packet. That doesn't seem worth the
cost & complexity.
Further, putting an IGP there will allow business customers to
hijack
any addresses w/ appropriate prefix advertisements, practically
requiring duplicating the ACLs at the CMTS side to allow only the
valid prefix. But with the same trouble you could do a static
route.
> Today's cable deployments uses this, so we had to discuss it in the
draft
> to make it close to reality. The BB ISP will use RIPng as they were
using
> RIPv2. We felt this explanation is out of the scope of this document
so we
> did not include.
> &&&
It's OK to say they use RIPv2 and thus RIPng seems applicable, but
it
seems also justified to point the discussion above, that routing
protocol is not needed unless you're multihomed or you care a lot
about sending ICMP unreachables when the link goes down. It's also
worth noting the filtering requirement for the routing protocol, to
prevent hijacking.
&&&
Sure. We will incorporate both options in the next revision.
Another point to note that in last mile BB environment dual/multi-homed
scenarios are hard to find. Cable ISPs use dynamic routing since
they find it much more convenient. If they have to configure couple
of hundred static routes at each CMTS, in large deployments, it adds up
and becomes hard to manage. Again, this is how it is done out
there.
&&&
This has been a great discussion which clarified lot of points and
hopefully will make this document to be very beneficial in years to
come.
We would like to thank you and all the individuals on the list for their
interest and valuable feedback (we have received several unicast replies
with good points and valuable feedback).
We will integrate all the feedback in the next revision.
Regards,
Salman Asadullah
Technical Leader
Cisco Systems, Inc.
--
Pekka
Savola
"You each name yourselves king, yet the
Netcore
Oy
kingdom bleeds."
Systems. Networks. Security. -- George R.R. Martin: A Clash of
Kings