Login | Register For Free | Help
Search for: (Advanced)

Mailing List Archive: NANOG: users

small site multi-homing (related to: Small guys with BGP issues)

 

 

NANOG users RSS feed   Index | Next | Previous | View Threaded


mike-nanog at tiedyenetworks

Nov 3, 2009, 8:11 AM

Post #1 of 10 (144 views)
Permalink
small site multi-homing (related to: Small guys with BGP issues)

Small-site multi-homing is one of the great inequities of the
Internet and one that can, and should, be solved. I envision an Internet
of the future where anyone with any mixture of any type of network
connections can achieve, automatically, provider independence and
inbound/outbound load sharing across disparate links. Gone is the built
in hostage situation of having to either use your provider assigned IP's
(>%99 of internet connected sites today), or the quantum leap of being
an AS with PI space (and the associated technical baggage to configure
and manage that beast). End users should have the power to dictate
their own routing policies and not suffer thru 'damping', 'urpf', or
other policies imposed on how or when their packets come and go. So if
you want to use 2 dsl lines and a CDMA modem, or a satellite and a
fiber, or 27 dial up modems and a T1, you should be able to do that and
the network should work with you to deliver your packets no matter where
'you' connect or how.

What it's gonna take is new routing paradigms and new thinking about
the role of providers and users and a lowering of the barriers between
these two for more cooperation in the overall structure of the network.
Just like classfull addressing giving way to cidr, I belive hierarchal
routing will give way to truely dynamic routing where all participants
have equal capabilities over their own domain with no one (or group) of
'providers' having any more or less influence on global reachability for
any 'users' who choose to go their own way, and I expect that to be an
easy (or even default) choice in the future.

You may say I'm a dreamer, but I'm not the only one. I hope some day
you'll join us, and the world will live as one.



>> What is the issue here, that your DSL provider won't speak BGP with you
>> no matter how many times you've asked, so you're complaining to NANOG
>> about it because you don't have the ability or authority to change
>> providers? Please correct me if I'm reading this wrong, but the emails
>> so far haven't been very clear and this isn't making a lot of sense.
>>


braaen at zcorum

Nov 3, 2009, 9:28 AM

Post #2 of 10 (138 views)
Permalink
Re: small site multi-homing (related to: Small guys with BGP issues) [In reply to]

While the idea of seamless routing sounds great, so does world peace... I
don't think I will see either in my lifetime. There are some technical
hurdles you will have to solve first.

1st how do I solve security (preventing spoofing and other evil deeds done by
rouge networks).
2nd how can my system scale and achieve stability.
3rd how will my routes work and converge (unstable routes don't work really
well).
4th My system will need to work and scale on a much larger environment than a
lab.
5th How do I test and verify your system.
6th Politics/Layer 8 (think peering wars)
7th How do I propose for routers be able to store (2^128 + 2^32) * x routes in
their routing table, and possibly utilize current hardware (the whole world
isn't going to do a flag day forklift upgrade)
8th How am I going to get anyone to invest money and R&D into my system.

If you have any good idea's we'd love to hear them. I am open to such a
system, but do not think it can realistically happen anytime soon.

--

----------------------

Brian Raaen
Network Engineer
braaen[at]zcorum.com


On Tuesday 03 November 2009, Mike wrote:
>
> Small-site multi-homing is one of the great inequities of the
> Internet and one that can, and should, be solved. I envision an Internet
> of the future where anyone with any mixture of any type of network
> connections can achieve, automatically, provider independence and
> inbound/outbound load sharing across disparate links. Gone is the built
> in hostage situation of having to either use your provider assigned IP's
> (>%99 of internet connected sites today), or the quantum leap of being
> an AS with PI space (and the associated technical baggage to configure
> and manage that beast). End users should have the power to dictate
> their own routing policies and not suffer thru 'damping', 'urpf', or
> other policies imposed on how or when their packets come and go. So if
> you want to use 2 dsl lines and a CDMA modem, or a satellite and a
> fiber, or 27 dial up modems and a T1, you should be able to do that and
> the network should work with you to deliver your packets no matter where
> 'you' connect or how.
>
> What it's gonna take is new routing paradigms and new thinking about
> the role of providers and users and a lowering of the barriers between
> these two for more cooperation in the overall structure of the network.
> Just like classfull addressing giving way to cidr, I belive hierarchal
> routing will give way to truely dynamic routing where all participants
> have equal capabilities over their own domain with no one (or group) of
> 'providers' having any more or less influence on global reachability for
> any 'users' who choose to go their own way, and I expect that to be an
> easy (or even default) choice in the future.
>
> You may say I'm a dreamer, but I'm not the only one. I hope some day
> you'll join us, and the world will live as one.
>
>
>
> >> What is the issue here, that your DSL provider won't speak BGP with you
> >> no matter how many times you've asked, so you're complaining to NANOG
> >> about it because you don't have the ability or authority to change
> >> providers? Please correct me if I'm reading this wrong, but the emails
> >> so far haven't been very clear and this isn't making a lot of sense.
> >>
>
>
>


Valdis.Kletnieks at vt

Nov 3, 2009, 9:42 AM

Post #3 of 10 (138 views)
Permalink
Re: small site multi-homing (related to: Small guys with BGP issues) [In reply to]

On Tue, 03 Nov 2009 08:11:15 PST, Mike said:
>
> Small-site multi-homing is one of the great inequities of the
> Internet and one that can, and should, be solved. I envision an Internet
> of the future where anyone with any mixture of any type of network
> connections can achieve, automatically, provider independence and
> inbound/outbound load sharing across disparate links.

400 million Joe Sixpacks and their counterparts around the globe, all wanting
to run BGPto multihome the /29 in their basement.

Be careful what you ask for, you may get it.


cluestore at gmail

Nov 3, 2009, 10:01 AM

Post #4 of 10 (138 views)
Permalink
Re: small site multi-homing (related to: Small guys with BGP issues) [In reply to]

Well you and the rest of these so called "dreamers" can help with the
purchase of my new routers that don't exist yet to support you wanting to
multi-home a /29 and have the rest of the Internet world hold all of these
said /29's in their tables. Most folks who get a /29's don't care how they
get to and from the internet, they just want to always be able to get there.
TE at that granular of a level is not needed. So in other words, you and the
rest of the world of these dreamers can keep dreaming, because I doubt any
sensible ISP would accept and pass along anyone announcing /29's .... and
then there's V6, which I won't even get started on. Most ISP's are having a
hard time holding 300k ipv4 routes as of today, and you want to de-aggregate
even farther??

Clue

On Tue, Nov 3, 2009 at 10:11 AM, Mike <mike-nanog[at]tiedyenetworks.com> wrote:

>
> Small-site multi-homing is one of the great inequities of the Internet
> and one that can, and should, be solved. I envision an Internet of the
> future where anyone with any mixture of any type of network connections can
> achieve, automatically, provider independence and inbound/outbound load
> sharing across disparate links. Gone is the built in hostage situation of
> having to either use your provider assigned IP's (>%99 of internet connected
> sites today), or the quantum leap of being an AS with PI space (and the
> associated technical baggage to configure and manage that beast). End users
> should have the power to dictate their own routing policies and not suffer
> thru 'damping', 'urpf', or other policies imposed on how or when their
> packets come and go. So if you want to use 2 dsl lines and a CDMA modem, or
> a satellite and a fiber, or 27 dial up modems and a T1, you should be able
> to do that and the network should work with you to deliver your packets no
> matter where 'you' connect or how.
>
> What it's gonna take is new routing paradigms and new thinking about the
> role of providers and users and a lowering of the barriers between these two
> for more cooperation in the overall structure of the network. Just like
> classfull addressing giving way to cidr, I belive hierarchal routing will
> give way to truely dynamic routing where all participants have equal
> capabilities over their own domain with no one (or group) of 'providers'
> having any more or less influence on global reachability for any 'users' who
> choose to go their own way, and I expect that to be an easy (or even
> default) choice in the future.
>
> You may say I'm a dreamer, but I'm not the only one. I hope some day
> you'll join us, and the world will live as one.
>
>
>
> What is the issue here, that your DSL provider won't speak BGP with you
>>> no matter how many times you've asked, so you're complaining to NANOG
>>> about it because you don't have the ability or authority to change
>>> providers? Please correct me if I'm reading this wrong, but the emails
>>> so far haven't been very clear and this isn't making a lot of sense.
>>>
>>>
>
>


joelja at bogus

Nov 3, 2009, 10:05 AM

Post #5 of 10 (138 views)
Permalink
Re: small site multi-homing (related to: Small guys with BGP issues) [In reply to]

Valdis.Kletnieks[at]vt.edu wrote:
> On Tue, 03 Nov 2009 08:11:15 PST, Mike said:
>> Small-site multi-homing is one of the great inequities of the
>> Internet and one that can, and should, be solved. I envision an Internet
>> of the future where anyone with any mixture of any type of network
>> connections can achieve, automatically, provider independence and
>> inbound/outbound load sharing across disparate links.

Hey there's always LISP, they even have code...

http://www.ietf.org/dyn/wg/charter/lisp-charter.html

The largest inequity of all is that cost delta to you when advertise one
more prefix (minor) vs the collective cost to the whole internet of
carrying it.

The fact is that a combination of technical conventions, business
considerations, and social pressures retard the growth in the routing
table to a rate which while not all that desirable from some
perspectives is manageable. It continues to be the case that the
barrier to entry is relatively low as the existance proof of new
entrants routinely shows.

There is in fact nothing other than a little money, time, and a business
need between you and multihoming. The fact that it may not be as cheap
or convenient as some might like is not the product of discrimination...

> 400 million Joe Sixpacks and their counterparts around the globe, all wanting
> to run BGPto multihome the /29 in their basement.
>
> Be careful what you ask for, you may get it.


davei at otd

Nov 3, 2009, 10:21 AM

Post #6 of 10 (136 views)
Permalink
Re: small site multi-homing (related to: Small guys with BGP issues) [In reply to]

Clue Store wrote:
> Well you and the rest of these so called "dreamers" can help with the
> purchase of my new routers that don't exist yet to support you wanting to
> multi-home a /29 and have the rest of the Internet world hold all of these
> said /29's in their tables. Most folks who get a /29's don't care how they
> get to and from the internet, they just want to always be able to get there.
> TE at that granular of a level is not needed. So in other words, you and the
> rest of the world of these dreamers can keep dreaming, because I doubt any
> sensible ISP would accept and pass along anyone announcing /29's .... and
> then there's V6, which I won't even get started on. Most ISP's are having a
> hard time holding 300k ipv4 routes as of today, and you want to de-aggregate
> even farther??
>

It's clear that you have some impatience with deaggregation, and with
cause. However, there are a few flaws in your position. The first is
that you contradicted yourself. If most folks who get a /29 don't care
how they get to and from the Internet, then there won't be a flood of
new /29s. It is the minority who do care how they get to and from the
Internet who will be adding routes. Currently, they are doing so by
getting more address space than they need assigned, so as to have a
block large enough to be heard. If 500 companies are currently
announcing /24s to be heard, but could be moved to /29s, then you still
have 500 route announcements. You just have a lot less waste.

The second is that you said "BGP." Mike didn't say BGP. He said he was
dreaming of the future. That future coudl easily include a lightweight
multihoming protocol, something that informs interested parties of
presence on multiple networks, or allows for extremely fast
reconvergence, so that a second route need only join the routing table
when needed. And he's right; if I want to change my name to Joe, grab a
sixpack, build a rack in my kitchen, and pay two providers for service,
it isn't unreasonable to want an infrastructure that supports my
configuration.

We shouldn't dismiss a dreamer's dream because it is hard, or we can't
do it right now with what we have. The desire to do what is not
currently possible is the source of innovation, and we shouldn't shoot
down innovation because it sounds hard and we don't like it.

-Dave


dave.nanog at alfordmedia

Nov 3, 2009, 10:33 AM

Post #7 of 10 (136 views)
Permalink
Re: small site multi-homing (related to: Small guys with BGP issues) [In reply to]

> If 500 companies are currently
> announcing /24s to be heard, but could be moved to /29s, then you still
> have 500 route announcements. You just have a lot less waste.

That's my situation here. I've got a /24 with fewer than 10 public IPs
active, because I need those 10 hosts to be reachable even after Bubba and
his backhoe finish tearing up the road in front of my office.
--
Dave Pooser, ACSA
Manager of Information Services
Alford Media http://www.alfordmedia.com


cluestore at gmail

Nov 3, 2009, 11:09 AM

Post #8 of 10 (136 views)
Permalink
Re: small site multi-homing (related to: Small guys with BGP issues) [In reply to]

I think you're missing my point and did not read my post completely.

First off, BGP was never mentioned in my post.

By the time these 'dreamers' want to announce a /29 to multiple providers
and have everyone accept them with this new light weight protocol you speak
about, there will hopefully be no /29's (as in v4 host sub-nets) as I dream
that IPv4 will be a forgotten protocol by the time BGP is replaced by this
magical protocol that does not exist in any form as today.

If I accept a /29 for the minority and pass that prefix along to the next
provider, I have to accept it for the majority and pass them along to the
next provider. And these 500 company's you speak about, the other blocks
given back to <insert RIR or LIR here> would be hashed back out which WOULD
still increase prefixes in the global table as they want to advertise their
/29's. I agree that it would save v4 space right now for those who wouldn't
announce the remainder /29's, but you're thinking short term as we all know
that v4 space has out-welcomed it's stay (thank you NAT). Yes, it will run
paraellel for 3, 5, maybe 7 years until enough folks get a clue and make the
switch to v6, but in the end, v4 will go away.

Having all that said, I am not knocking the 'dreamers' out there one bit. I
encourage new ideas to help solve issues that we've discussed in this very
thread. But at this point, there's more dreaming than solutions and revenue.
And de-aggreation is one of the biggest problems with global routing today.
Add v6 and the possibility of /48's being permitted into the global table,
and most folks with a router from any vendor today couldn't support a full
global table.

I'll stop my rant at that, but again, im not knocking the dreamers. I'm just
having to deal with more problems that don't have valid solutions today.

Clue
On Tue, Nov 3, 2009 at 12:21 PM, Dave Israel <davei[at]otd.com> wrote:

>
> Clue Store wrote:
> > Well you and the rest of these so called "dreamers" can help with the
> > purchase of my new routers that don't exist yet to support you wanting to
> > multi-home a /29 and have the rest of the Internet world hold all of
> these
> > said /29's in their tables. Most folks who get a /29's don't care how
> they
> > get to and from the internet, they just want to always be able to get
> there.
> > TE at that granular of a level is not needed. So in other words, you and
> the
> > rest of the world of these dreamers can keep dreaming, because I doubt
> any
> > sensible ISP would accept and pass along anyone announcing /29's .... and
> > then there's V6, which I won't even get started on. Most ISP's are having
> a
> > hard time holding 300k ipv4 routes as of today, and you want to
> de-aggregate
> > even farther??
> >
>
> It's clear that you have some impatience with deaggregation, and with
> cause. However, there are a few flaws in your position. The first is
> that you contradicted yourself. If most folks who get a /29 don't care
> how they get to and from the Internet, then there won't be a flood of
> new /29s. It is the minority who do care how they get to and from the
> Internet who will be adding routes. Currently, they are doing so by
> getting more address space than they need assigned, so as to have a
> block large enough to be heard. If 500 companies are currently
> announcing /24s to be heard, but could be moved to /29s, then you still
> have 500 route announcements. You just have a lot less waste.
>
> The second is that you said "BGP." Mike didn't say BGP. He said he was
> dreaming of the future. That future coudl easily include a lightweight
> multihoming protocol, something that informs interested parties of
> presence on multiple networks, or allows for extremely fast
> reconvergence, so that a second route need only join the routing table
> when needed. And he's right; if I want to change my name to Joe, grab a
> sixpack, build a rack in my kitchen, and pay two providers for service,
> it isn't unreasonable to want an infrastructure that supports my
> configuration.
>
> We shouldn't dismiss a dreamer's dream because it is hard, or we can't
> do it right now with what we have. The desire to do what is not
> currently possible is the source of innovation, and we shouldn't shoot
> down innovation because it sounds hard and we don't like it.
>
> -Dave
>
>
>


davei at otd

Nov 3, 2009, 11:30 AM

Post #9 of 10 (136 views)
Permalink
Re: small site multi-homing (related to: Small guys with BGP issues) [In reply to]

Clue Store wrote:
> I think you're missing my point and did not read my post completely.
>
> First off, BGP was never mentioned in my post.

Oops, you are correct. Somebody else said "BGP." You spoke of the
existing table, and so I had BGP in my mind, and I muddled the two
together. Mea culpa.

> If I accept a /29 for the minority and pass that prefix along to the
> next provider, I have to accept it for the majority and pass them
> along to the next provider. And these 500 company's you speak about,
> the other blocks given back to <insert RIR or LIR here> would be
> hashed back out which WOULD still increase prefixes in the global
> table as they want to advertise their /29's. I agree that it would
> save v4 space right now for those who wouldn't announce the remainder
> /29's, but you're thinking short term as we all know that v4 space has
> out-welcomed it's stay (thank you NAT). Yes, it will run paraellel for
> 3, 5, maybe 7 years until enough folks get a clue and make the switch
> to v6, but in the end, v4 will go away.

That assumes that there isn't a solution that requires constant presence
in the global table, instead of a
tell-me-about-this-prefix-when-I-need-it-and-not-before method. I admit
that there hasn't been a good solution to the problem yet, but that
doesn't mean there isn't one. I'm not sure it has been seriously
researched in recent years.

> Having all that said, I am not knocking the 'dreamers' out there one
> bit. I encourage new ideas to help solve issues that we've discussed
> in this very thread. But at this point, there's more dreaming than
> solutions and revenue. And de-aggreation is one of the biggest
> problems with global routing today. Add v6 and the possibility of
> /48's being permitted into the global table, and most folks with a
> router from any vendor today couldn't support a full global table.

No, but providers having to upgrade software or hardware to support the
needs of the network in 3, 5, or 7 years isn't anything new, and neither
is router vendors coming up with incremental software or hardware
upgrades to make boxes do what they can't do now.

-Dave


marka at isc

Nov 3, 2009, 4:02 PM

Post #10 of 10 (128 views)
Permalink
Re: small site multi-homing (related to: Small guys with BGP issues) [In reply to]

In message <5414.1257270127[at]turing-police.cc.vt.edu>, Valdis.Kletnieks[at]vt.edu w
rites:
> On Tue, 03 Nov 2009 08:11:15 PST, Mike said:
> >
> > Small-site multi-homing is one of the great inequities of the
> > Internet and one that can, and should, be solved. I envision an Internet
> > of the future where anyone with any mixture of any type of network
> > connections can achieve, automatically, provider independence and
> > inbound/outbound load sharing across disparate links.
>
> 400 million Joe Sixpacks and their counterparts around the globe, all wanting
> to run BGPto multihome the /29 in their basement.
>
> Be careful what you ask for, you may get it.

With a protocol to distribute which prefixes (with weighting) are
viable, a end node could just select a appropritate source address
out of several provider assigned ones and use source address routing
to find a appropropiate exit path which doesn't break BCP 38. This
is as good as the NAT solutions for small-site multi-homing today.

Mark
--
Mark Andrews, ISC
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742 INTERNET: marka[at]isc.org

NANOG users RSS feed   Index | Next | Previous | View Threaded
 
 


Interested in having your list archived? Contact lists@gossamer-threads.com
 
  Web Applications & Managed Hosting Powered by Gossamer Threads Inc.