RE: An informal survey... round II

[tariq biziou]

> > > Why should we announce tiny recycled blocks? If there is a /16 in the
> > > swamp in which half the space is free but its all /24's, why wouldn't
> > > wouldn't we allocate all the free /24's to a single entity and
> > > instruct the entity to announce it as a "holey" /16? The existing /24
> > > holders will override (punch holes in) the /16 for their /24's.
> >
> > And when they withdraw the more specific or you glop them together in
> > your fib in the name of agregation a 3rd party gets all their traffic?
> > I'm sure that will work really well.
>
> Only the next hop. The game to play is "I want to dump traffic to a
> neighboring AS who has more a chance of getting it to its relevant
> destination." Partial routes (eg, filtering on say a /24 boundary
> with a default route) already sort of gives you that.

this idea has been flawed from the beginning:

1. prefix length indicates neither network capacity nor volume of traffic

2. most people prefer not to pay for traffic that is neither theirs nor a
customer's

-- 
--tariq

Re: An informal survey... round II

[John Curran]

At 2:14 PM -0400 8/30/07, Deepak Jain wrote:
>John --
>
>Great panic starting question.

Sorry, not my intent.  I'm just trying to get a handle on the
state of the art of what's available today, and whether it has
some really excellent scaling properties in case we see a much
more granular block reuse.  It's by no means inevitable, just
a probable outcome of any routing-indifferent IP address space
"market" (whether legitimate or not).

>I'd guess that by 2010, we'll be worried more about IPv6 growth than IPv4 
>growth but the archives will bite me in the you-know-where in 3 years when I'm 
>wrong (in either direction).

I'd love to believe that, but there's real world issues with IPv6
readiness as well as challenges communicating the business
need to make the necessary preparations.   When we're all
still debating in technical fora the need to move to IPv6 versus
viability of a heavily NAT'ed IPv4 endgame,  you can't expect the
folks in boardrooms to commit to any large investment decisions.

/John

RE: An informal survey... round II

[michael.dillon]

> > If there is any company whose IPv6 plans we should be 
> interested in, 
> > it is Verizon.
> 
> AKA UUNET? They've been doing IPv6 for _years_. I got my 
> first IPv6 tunnel from UUNET Netherlands way back when.

But when will all of Verizon, not just the UUNET parts, offer IPv6
transit and peering on roughly the same terms as IPv4 transit/peering?
Tunnel brokers don't count. UUNET's AS701 was the best connected AS on
the Internet according to various analyses. Verizon now has more than
just UUNET in its portfolio so they are even more core to the IPv4
Internet than UUNET was. When Verizon offers native IPv6 access
basically everywhere, this will have a major, major impact on the size
and reach of the IPv6 Internet. It will also have a knock-on effect as
Verizon peers (and pseudo peers who buy transit) connect their IPv6
networks to the IPv6 Internet. End user customers (businesses) will sit
up and take notice of the IPv6 Internet. I also expect that people who
report latency and hop-count analyses will also have some positive
things to say about the IPv6 Internet after this happens.

Let's not forget that we all depend on each other in this business, and
although we collectively operate a very good IPv4 Internet, the IPv6
Internet is still just taking baby steps. We need a short period of
exponential growth in the IPv6 Internet to make it ready for prime time.

--Michael Dillon

RE: An informal survey... round II

[michael.dillon]

> > Considering
> > Verizon's highly-connected position at the core of the IPv4 
> Internet, 
> > I would think that all it takes to cause a snowball effect, is for 
> > Verizon to start offering IPv6 transit and peering on the 
> same terms 
> > as IPv4. If there is any company whose IPv6 plans we should be 
> > interested in, it is Verizon.

> I believe that your sales peoples should know how to make 
> this happen for you, if not please ping the Customer Service 
> folk who would also be able to make some progress on this 
> issue for you.

Issue? It was more of an observation than an issue. Based on the CAIDA
Internet connectdness maps which show AS701 as the most connected AS in
the core of the Internet, combined with other Verizon networks, it seems
to me that Verizon's progress is the most interesting to watch. When
Verizon has full availability of IPv6 access (all ASes, all PoPs) then I
see that as a tipping point.

--Michael Dillon
 

Re: An informal survey... round II

[Adrian Chadd]

On Thu, Aug 30, 2007, Joel Jaeggli wrote:

> > Why should we announce tiny recycled blocks? If there is a /16 in the
> > swamp in which half the space is free but its all /24's, why wouldn't
> > wouldn't we allocate all the free /24's to a single entity and
> > instruct the entity to announce it as a "holey" /16? The existing /24
> > holders will override (punch holes in) the /16 for their /24's.
> 
> And when they withdraw the more specific or you glop them together in
> your fib in the name of agregation a 3rd party gets all their traffic?
> I'm sure that will work really well.

Only the next hop. The game to play is "I want to dump traffic to a
neighboring AS who has more a chance of getting it to its relevant
destination." Partial routes (eg, filtering on say a /24 boundary
with a default route) already sort of gives you that.

(Of course, that game is a pain in the ass to play at times..)

Remember, BGP on the global internet isn't TE'ing "end to end".
You're actually hoping the next hop you have in your routing table ==
closer to the destination than you. :)



Adrian

Re: An informal survey... round II

[Iljitsch van Beijnum]

On 30-aug-2007, at 18:35, <[EMAIL PROTECTED]>  
<[EMAIL PROTECTED]> wrote:


If there is any company whose IPv6 plans we should be interested  
in, it is Verizon.


AKA UUNET? They've been doing IPv6 for _years_. I got my first IPv6  
tunnel from UUNET Netherlands way back when.

Re: An informal survey... round II

[Deepak Jain]

John --

Great panic starting question.

I'd guess that by 2010, we'll be worried more about IPv6 growth than 
IPv4 growth but the archives will bite me in the you-know-where in 3 
years when I'm wrong (in either direction).


And then we'll talk about how fast FIBs get eaten with both IPv4 
(legacy) and IPv6 (new coolness) routes are in the same 
TCAMs/SRAMS/RLDRAM/whatever.


How many of us realistically believe that the core routers we install 
today will be in use (as such) in 3-4 years??? I can't think of a time 
in the last 12+ years that that has been a good bet.


Deepak


John Curran wrote:

So with a fairly predictable growth of 3500 routes per month, we're
going to have some issues with the current equipment out there
(despite this being a rather predictable situation...)

So what might happen in three years if we have a sudden inflection
in new routes per month due to use by major backbones of non-
hierarchically allocated address space for new customer additions?
I.E.  If at some time unknown around 2010, ISP's stop receiving
new allocations from their RIR, and instead use of many smaller
"recycled" IPv4 address blocks, we could be looking at a 10x to
20x increase in routes per month for the same customer growth.

Is the equipment being installed *today* and over the next two
years capable of sustaining 50K new routes per month, and if so,
for how long?

Thanks,
/John

At 4:47 AM + 8/30/07, [EMAIL PROTECTED] wrote:

On Wed, Aug 29, 2007 at 06:48:43PM -0400, Jon Lewis wrote:

On Mon, 27 Aug 2007, David Conrad wrote:


For a few more months.  What are upgrade cycles like again?  How common
are the MSFC2s?

I think we'll find out in a few months, when the "internet breaks" in a
whole bunch of places where the admins aren't aware of this issue or
operations have been downsized to the point that things are mostly on
auto-pilot.  I'm guessing there are a good number of Sup2's in use, and
that a good % of them think they're fine...as they have 512MB RAM and on
the software based routers, that's plenty for current full BGP routes.

private replies suggest (w/ lots of handwaving) that perhaps 20-35%
of the forwarding engines in use might fit this catagory.


Anyone want to bet there will be people posting to nanog and cisco-nsp in
a few months asking why either the CPU load on their Sup2's has suddenly
shot up or why they keep noticing parts of the internet have gone
unreachable?...oblivious to this thread.

that would be a sucker bet


--
 Jon Lewis   |  I route

--bill

RE: An informal survey... round II

[michael.dillon]

> > People keep saying that there is no business case for IPv6 when the 
> > answer is staring them in the face. Growing revenue is the absolute 
> > fundamental core of any business case, and in telecom 
> companies that 
> > is generally directly tied into growing the network.
> 
> Can you point me to BT's IPv6 deployment plans?  (A serious 
> request.   
> If you need an NDA, I'm happy to sign one)

We have been doing IPv6 for many years now, running an IPv6 exchange and
a tunnel broker and various other things. Some info is on our public
website here http://www.ipv6.bt.com/index.html especially in the
presentations section. And the presenters tend to include an email
address on their slides so you don't need to bother going through me.
But then I don't know what info you are interested in. If you still need
something, email me offlist and I'll see what I can do.

In addition to our old, old sub-TLA allocation and our exchange
allocation, just yesterday we picked up another big IPv6 prefix from
RIPE which will end up being used for commercial services in future. As
you can guess, we are in process of shifting from test mode to
production. 

When we counted up the days left until IPv4 exhaustion, we saw an
ominous number so we've posted our IPv4 doomsday clock at
http://penrose.uk6x.com/  I notice that the people at
http://www.ipv6forum.com have adopted it for their homepage too.

Even though there is not a lot of publicity surrounding it, I believe
that there are very few IP transit network operators who haven't got
solid IPv6 deployment plans being trialed right now. Considering
Verizon's highly-connected position at the core of the IPv4 Internet, I
would think that all it takes to cause a snowball effect, is for Verizon
to start offering IPv6 transit and peering on the same terms as IPv4. If
there is any company whose IPv6 plans we should be interested in, it is
Verizon.

--Michael Dillon

Re: An informal survey... round II

[Joel Jaeggli]

William Herrin wrote:
> On 8/30/07, John Curran <[EMAIL PROTECTED]> wrote:
>> I.E.  If at some time unknown around 2010, ISP's stop receiving
>> new allocations from their RIR, and instead use of many smaller
>> "recycled" IPv4 address blocks, we could be looking at a 10x to
>> 20x increase in routes per month for the same customer growth.
> 
> John,
> 
> Why should we announce tiny recycled blocks? If there is a /16 in the
> swamp in which half the space is free but its all /24's, why wouldn't
> wouldn't we allocate all the free /24's to a single entity and
> instruct the entity to announce it as a "holey" /16? The existing /24
> holders will override (punch holes in) the /16 for their /24's.

And when they withdraw the more specific or you glop them together in
your fib in the name of agregation a 3rd party gets all their traffic?
I'm sure that will work really well.

> Regards,
> Bill Herrin
> 
> 

Re: An informal survey... round II

[David Conrad]

Michael,

On Aug 30, 2007, at 7:35 AM, <[EMAIL PROTECTED]>  
<[EMAIL PROTECTED]> wrote:
People keep saying that there is no business case for IPv6 when the  
answer is staring them in the face. Growing revenue is the absolute  
fundamental core of any business case, and in telecom companies  
that is generally directly tied into growing the network.


Can you point me to BT's IPv6 deployment plans?  (A serious request.   
If you need an NDA, I'm happy to sign one)


Thanks,
-drc

Re: An informal survey... round II

[Brandon Butterworth]

> Is there a possible revenue stream here for larger ISP's to begin 
> charging their customers for not aggregating

or to pay a clearing house to reaggregate them, mutual trades so you
can accumulate enough of a block, approach current (non)users of space
and buy out to build blocks they can sell

If we hit the limits a possible outcome is larger blocks will sell
for more per address than smaller as there'll be lots of people filtering
on arbitrary rules. Time to standardise how the pain will be applied
now?

brandon

Re: An informal survey... round II

[Kevin Oberman]

> Date: Thu, 30 Aug 2007 10:18:41 -0400
> From: Andrew D Kirch <[EMAIL PROTECTED]>
> Sender: [EMAIL PROTECTED]
> 
> 
> John Curran wrote:
> > At 9:12 AM -0400 8/30/07, William Herrin wrote:
> >   
> >> On 8/30/07, John Curran <[EMAIL PROTECTED]> wrote:
> >> 
> >>> I.E.  If at some time unknown around 2010, ISP's stop receiving
> >>> new allocations from their RIR, and instead use of many smaller
> >>> "recycled" IPv4 address blocks, we could be looking at a 10x to
> >>> 20x increase in routes per month for the same customer growth.
> >>>   
> >> John,
> >>
> >> Why should we announce tiny recycled blocks? If there is a /16 in the
> >> swamp in which half the space is free but its all /24's, why wouldn't
> >> wouldn't we allocate all the free /24's to a single entity and
> >> instruct the entity to announce it as a "holey" /16? The existing /24
> >> holders will override (punch holes in) the /16 for their /24's.
> >> 
> >
> > Consider large ISP's that can no longer obtain from the large blocks
> > (e.g. /12 to /16) but instead must beg/barter/borrow blocks from others
> > which are several orders  of magnitude smaller (e.g. /16 through /24)
> > every week to continue growing...  such obtained blocks would be
> > announced into the routing system very rapidly as we try to keep
> > IPv4 running post depletion of the free address pool.  When this
> > inflection point is reached, how much headroom do we have given
> > equipment being deployed today?
> >
> > /John
> Is there a possible revenue stream here for larger ISP's to begin 
> charging their customers for not aggregating, and creating a penalty fee 
> for each borken route?  We're running out of IPv4 space (and I don't 
> think this can be solved with IPv4).  We're running out of routes for 
> the Cisco Sup2 engine (among others), but unless someone makes money on 
> it, this won't be solved.

When customers start having problems due to router RIB and/or FIB
overflow, they will seek other providers. So someone will make money and
someone else will lose it. Economics works well when things break
(hardware). 

Of course, what happens when there is simply no IPv4 space is another
issue as there will be no new space for anyone. But some providers will
run out of their allocations long before others. If those who run out
have resources, the market in IPv4 addresses (which does not exist since
addresses are just numbers and not property) will get very active and
will provide its own incentives which may or may not be stabilizing or
beneficial. 
-- 
R. Kevin Oberman, Network Engineer
Energy Sciences Network (ESnet)
Ernest O. Lawrence Berkeley National Laboratory (Berkeley Lab)
E-mail: [EMAIL PROTECTED]   Phone: +1 510 486-8634
Key fingerprint:059B 2DDF 031C 9BA3 14A4  EADA 927D EBB3 987B 3751


pgptZQyOdTftm.pgp
Description: PGP signature

RE: An informal survey... round II

[michael.dillon]

> Consider large ISP's that can no longer obtain from the large 
> blocks (e.g. /12 to /16) but instead must beg/barter/borrow 
> blocks from others which are several orders  of magnitude 
> smaller (e.g. /16 through /24) every week to continue 
> growing...  such obtained blocks would be announced into the 
> routing system very rapidly as we try to keep
> IPv4 running post depletion of the free address pool.  When 
> this inflection point is reached, how much headroom do we 
> have given equipment being deployed today?

You have described a pretty desperate state of affairs. Given that the route 
announcements are part of the public record, this is tantamount to holding a 
press conference and telling everyone that your business is in a pretty 
desperate state and you are scrambling just to keep your head above water in 
the hopes that it will stave off disaster long enough for you to get a proper 
solution deployed.

I wonder how many companies will let things get this bad when IPv6 is right 
here, right now, and mitigates against this kind of disastrous state of 
affairs? I wonder how many investment analysts will note that companies without 
a solid IPv6 deployment plan in 2008 and 2009, are likely to start hemorrhaging 
in 2010 as customers scramble to find a stable supplier before the inevitable 
day of doom?

People keep saying that there is no business case for IPv6 when the answer is 
staring them in the face. Growing revenue is the absolute fundamental core of 
any business case, and in telecom companies that is generally directly tied 
into growing the network. If your fancy new IPTV home banking package deal 
depends on connecting new customers to your network, lack of IP addresses will 
stop your provisioning process dead in its tracks. And that stops growth dead 
in its tracks. And that shows that all the money that management invested in 
the fancy new IPTV home banking package was actually wasted (or shall we say 
misdirected) investment like all that fancy décor on the Titanic.

In general, telecoms companies (read ISPs) are trying to move up the value 
chain, but any product which depends on network connectivity probably has a 
direct dependence on growing the network. This means it is directy dependent on 
a steady supply of fresh IP addresses. We stopped manufacturing IPv4 addresses 
a long time ago. ARIN and ICANN have issued end-of-life announcements for IPv4 
addresses. Does it make any business sense at all to bet the farm on products 
which only work using IPv4 addresses?

The fact is that IPv6 and IPv4 can interwork quite well so it is not a great 
technical feat to make IPv6 products that work. It certainly costs more up 
front to develop such products, but I doubt that anyone has done a real cost 
analysis comparing this to marketing costs, and other softer product 
development costs. Companies can introduce expensive products and still make 
money at it. Cutting prices to the bone is not the only way to make money. 
Engineers like to exclaim that IPv6 costs too much, but engineers rarely ever 
have the data to quantify exactly what that cost is and why it is too much. 

In fact, IPv6 doesn't cost too much. Lots of companies are using it today. 
Government agencies are installing it because the Federal government's GAO 
thinks that IPv6 is money well spent, i.e. it does NOT cost too much. IPv6 is 
appearing with more frequency on RFPs, not just in questions about future 
plans, but as a real requirement for service today. Many customers will be 
satisfied with Hexago boxes connected to your IPv4 network. More will be 
satisfied with Softwires across an IPv4 network. They will be exstatic if you 
offer 6PE over MPLS (or dual stack) and geographically diverse IPv6 peering. 
IPv6 is real and moaning about how hard it is will not make it go away, 
especially when it gets on your CEO's radar.

--Michael Dillon

P.S. rant directed at the IPv6 naysayers, not the many people on this list who 
are deploying IPv6 today or at least planning their deployments for next year.

Re: An informal survey... round II

[Andrew D Kirch]

John Curran wrote:

At 9:12 AM -0400 8/30/07, William Herrin wrote:
  

On 8/30/07, John Curran <[EMAIL PROTECTED]> wrote:


I.E.  If at some time unknown around 2010, ISP's stop receiving
new allocations from their RIR, and instead use of many smaller
"recycled" IPv4 address blocks, we could be looking at a 10x to
20x increase in routes per month for the same customer growth.
  

John,

Why should we announce tiny recycled blocks? If there is a /16 in the
swamp in which half the space is free but its all /24's, why wouldn't
wouldn't we allocate all the free /24's to a single entity and
instruct the entity to announce it as a "holey" /16? The existing /24
holders will override (punch holes in) the /16 for their /24's.



Consider large ISP's that can no longer obtain from the large blocks
(e.g. /12 to /16) but instead must beg/barter/borrow blocks from others
which are several orders  of magnitude smaller (e.g. /16 through /24)
every week to continue growing...  such obtained blocks would be
announced into the routing system very rapidly as we try to keep
IPv4 running post depletion of the free address pool.  When this
inflection point is reached, how much headroom do we have given
equipment being deployed today?

/John
Is there a possible revenue stream here for larger ISP's to begin 
charging their customers for not aggregating, and creating a penalty fee 
for each borken route?  We're running out of IPv4 space (and I don't 
think this can be solved with IPv4).  We're running out of routes for 
the Cisco Sup2 engine (among others), but unless someone makes money on 
it, this won't be solved.

Re: An informal survey... round II

[John Curran]

At 9:12 AM -0400 8/30/07, William Herrin wrote:
>On 8/30/07, John Curran <[EMAIL PROTECTED]> wrote:
>> I.E.  If at some time unknown around 2010, ISP's stop receiving
>> new allocations from their RIR, and instead use of many smaller
>> "recycled" IPv4 address blocks, we could be looking at a 10x to
>> 20x increase in routes per month for the same customer growth.
>
>John,
>
>Why should we announce tiny recycled blocks? If there is a /16 in the
>swamp in which half the space is free but its all /24's, why wouldn't
>wouldn't we allocate all the free /24's to a single entity and
>instruct the entity to announce it as a "holey" /16? The existing /24
>holders will override (punch holes in) the /16 for their /24's.

Consider large ISP's that can no longer obtain from the large blocks
(e.g. /12 to /16) but instead must beg/barter/borrow blocks from others
which are several orders  of magnitude smaller (e.g. /16 through /24)
every week to continue growing...  such obtained blocks would be
announced into the routing system very rapidly as we try to keep
IPv4 running post depletion of the free address pool.  When this
inflection point is reached, how much headroom do we have given
equipment being deployed today?

/John

Re: An informal survey... round II

[Jon Lewis]

On Thu, 30 Aug 2007, William Herrin wrote:


Why should we announce tiny recycled blocks? If there is a /16 in the
swamp in which half the space is free but its all /24's, why wouldn't
wouldn't we allocate all the free /24's to a single entity and
instruct the entity to announce it as a "holey" /16? The existing /24
holders will override (punch holes in) the /16 for their /24's.


Except when there are /24-holder outages, at which point their traffic 
gets hijacked by the /16 announcer.  Would you want to trust some random 
company to not take advantage of that situation in any way (collection of 
passwords, sampling your web traffic, putting up a fake "your org" web 
site, etc.)?  As a holey /16 announcer, would you want all the junk 
traffic that results from /24-holder outages?  What if one of them was 
running NS's for a popular DNSBL, and their outage basically caused a DDoS 
attack against your network?


--
 Jon Lewis   |  I route
 Senior Network Engineer |  therefore you are
 Atlantic Net|
_ http://www.lewis.org/~jlewis/pgp for PGP public key_

Re: An informal survey... round II

[William Herrin]

On 8/30/07, John Curran <[EMAIL PROTECTED]> wrote:
> I.E.  If at some time unknown around 2010, ISP's stop receiving
> new allocations from their RIR, and instead use of many smaller
> "recycled" IPv4 address blocks, we could be looking at a 10x to
> 20x increase in routes per month for the same customer growth.

John,

Why should we announce tiny recycled blocks? If there is a /16 in the
swamp in which half the space is free but its all /24's, why wouldn't
wouldn't we allocate all the free /24's to a single entity and
instruct the entity to announce it as a "holey" /16? The existing /24
holders will override (punch holes in) the /16 for their /24's.

Regards,
Bill Herrin


-- 
William D. Herrin  [EMAIL PROTECTED]  [EMAIL PROTECTED]
3005 Crane Dr.Web: 
Falls Church, VA 22042-3004

An informal survey... round II

[John Curran]

So with a fairly predictable growth of 3500 routes per month, we're
going to have some issues with the current equipment out there
(despite this being a rather predictable situation...)

So what might happen in three years if we have a sudden inflection
in new routes per month due to use by major backbones of non-
hierarchically allocated address space for new customer additions?
I.E.  If at some time unknown around 2010, ISP's stop receiving
new allocations from their RIR, and instead use of many smaller
"recycled" IPv4 address blocks, we could be looking at a 10x to
20x increase in routes per month for the same customer growth.

Is the equipment being installed *today* and over the next two
years capable of sustaining 50K new routes per month, and if so,
for how long?

Thanks,
/John

At 4:47 AM + 8/30/07, [EMAIL PROTECTED] wrote:
>On Wed, Aug 29, 2007 at 06:48:43PM -0400, Jon Lewis wrote:
>>
>> On Mon, 27 Aug 2007, David Conrad wrote:
>>
>> >For a few more months.  What are upgrade cycles like again?  How common
>> >are the MSFC2s?
>>
>> I think we'll find out in a few months, when the "internet breaks" in a
>> whole bunch of places where the admins aren't aware of this issue or
>> operations have been downsized to the point that things are mostly on
>> auto-pilot.  I'm guessing there are a good number of Sup2's in use, and
>> that a good % of them think they're fine...as they have 512MB RAM and on
>> the software based routers, that's plenty for current full BGP routes.
>
>   private replies suggest (w/ lots of handwaving) that perhaps 20-35%
>   of the forwarding engines in use might fit this catagory.
>
>> Anyone want to bet there will be people posting to nanog and cisco-nsp in
>> a few months asking why either the CPU load on their Sup2's has suddenly
>> shot up or why they keep noticing parts of the internet have gone
>> unreachable?...oblivious to this thread.
>
>   that would be a sucker bet
>
>> --
>>  Jon Lewis   |  I route
>
>--bill