Tier-2 reachability and multihoming
Hi there, I have been working on characterizing the internet hierarchy. I noticed that 27% of the total possible tier-2 provider node pairs are not connected i.e., they dont have any tier-1 node connecting them nor a direct peering link between them. Multihoming can be used as a predominant reason for the reachability of tier-3 nodes which are customers of these nodes, but what about the reachability of tier-2 nodes themselves and its customers which cannot afford to multihoming? How does BGP solve this reachability problem when it gets a request to a prefix unreachable? 1 tier-1 / 2 4 tier-2 / \ / \ 5 6 7 8 tier-3 here, nodes 2 and 4 have no reachability, 1 / | 2 3 4 / \ \/ \ 5 6 7 8 now, node 7 is reachable from 2 and its lower level nodes, but what about node 4 and 8, and as a typical case, suppose nodes 4 and 8 have no multihoming whatsoever, what then? Regards,
I have been working on characterizing the internet hierarchy. I noticed that 27% of the total possible tier-2 provider node pairs are not connected i.e., they dont have any tier-1 node connecting them nor a direct peering link between them.
It's quite simple. The Internet is not a tree hierachy; it is a partial mesh. Partial meshes can often be characterised as having some sort of hierarchy of connectedness, however the Internet does change continuously which means that an analysis of hierarchy done today will come up with different results from last year's analysis. The terminology of "tier 1" and "tier 2" only refers to a brief time in the evolution of the Internet in North America during the 1990s when the topology was much more treelike. That is all changed. Go to google and search the following line exactly as written. "internet topology" "partial mesh" --Michael Dillon
--On Wednesday, March 23, 2005 4:54 PM +0530 G Pavan Kumar <pavanji@cse.iitb.ac.in> wrote:
Hi there, I have been working on characterizing the internet hierarchy. I noticed that 27% of the total possible tier-2 provider node pairs are not connected i.e., they dont have any tier-1 node connecting them nor a direct peering link between them. Multihoming can be used as a predominant reason for the reachability of tier-3 nodes which are customers of these nodes, but what about the reachability of tier-2 nodes themselves and its customers which cannot afford to multihoming? How does BGP solve this reachability problem when
it gets a request to a prefix unreachable?
I think that likely you're looking at partial data (well i am sure you are, since i'm part of the internet and you didn't' get routing data from me...) and not seeing paths because of that. The BGP tables of a single node list all outward paths to other places. Thus from a single sample point it is totally impossible to 'map' the internet. Not to mention the *constant* change in routing.
On Wed, 23 Mar 2005, Michael Loftis wrote:
I think that likely you're looking at partial data (well i am sure you are, since i'm part of the internet and you didn't' get routing data from me...)
Duh !
and not seeing paths because of that. The BGP tables of a single node list all outward paths to other places. Thus from a single sample point it is totally impossible to 'map' the internet.
Not to mention the *constant* change in routing.
Actually, I am not doing what you think I am. I am using the RouteViews aggregation of the BGP routing tables. RouteViews is a project at the univ. of Oregon that peers with backbones and other ASes at interesting locations so as to make it as comprehensive as possible. Also, it updates the data every 2 hours of everyday. So, I am looking at almost full and fresh data :>
i don't thing an operator or seasoned researcher would characterize route-views or ripe ris as "almost full" data. they provide such a small and narrow peek as to require great caution when dealing with them. considering the topologies you suggest, folk may legitimately wonder if perhaps you have not exercised sufficient caution. randy
On Mar 24, 2005, at 12:06 AM, G Pavan Kumar wrote:
On Wed, 23 Mar 2005, Michael Loftis wrote:
I think that likely you're looking at partial data (well i am sure you are, since i'm part of the internet and you didn't' get routing data from me...)
Duh !
Not nice to make fun of people who are trying to help you.
and not seeing paths because of that. The BGP tables of a single node list all outward paths to other places. Thus from a single sample point it is totally impossible to 'map' the internet.
Not to mention the *constant* change in routing.
Actually, I am not doing what you think I am. I am using the RouteViews aggregation of the BGP routing tables. RouteViews is a project at the univ. of Oregon that peers with backbones and other ASes at interesting locations so as to make it as comprehensive as possible. Also, it updates the data every 2 hours of everyday. So, I am looking at almost full and fresh data :>
Unfortunately, the paragraph above shows me that there are errors in your base assumptions about how the Internet works. A couple of people have tried to point this out to you, you should listen instead of telling them why they are wrong. It is bad to base conclusions on incorrect assumptions. It is even worse to assume those of whom you ask for help know less than you do about the topic at hand. I am very sorry that you spent a lot of time probably doing good work digging through the route-views archives but have seem to come to false conclusions. It can be difficult to admit hard work has come to a bad end. However, it might not have been a waste. You seem to have the motivation, time, and energy to research the topic, perhaps your research can be quickly applied to different data, or in a different way? Might I suggest a Google search for past research on Internet topology? I believe the University of Oregon has done some. :) And CAIDA. And many others. Many are still doing research and happy to collaborate. Good luck in your research. -- TTFN, patrick
On Thu, 24 Mar 2005, G Pavan Kumar wrote: > Actually, I am not doing what you think I am. I am using the RouteViews > aggregation of the BGP routing tables. RouteViews is a project at the > univ. of Oregon that peers with backbones. Really? Could you tell us more about it? I thought there was just one Internet backbone. > I am looking at almost full and fresh data. So what value do you assign to "almost full?" There's a difference between "best" and "complete," which you may not be entirely appreciating. -Bill
On Thu, Mar 24, 2005 at 12:18:34PM -0800, Bill Woodcock wrote:
On Thu, 24 Mar 2005, G Pavan Kumar wrote: > Actually, I am not doing what you think I am. I am using the RouteViews > aggregation of the BGP routing tables. RouteViews is a project at the > univ. of Oregon that peers with backbones.
Really? Could you tell us more about it? I thought there was just one Internet backbone.
Bill... Stop it!!! shooting fish in a barrel is no sport at all.
> I am looking at almost full and fresh data.
So what value do you assign to "almost full?" There's a difference between "best" and "complete," which you may not be entirely appreciating.
-Bill
almost full == just after dessert and as you (and almost every one else on this list) know, there is zero chance of "complete" ... and "best" is always in the eye/routing-table of the beholder. --bill
On Fri, 25 Mar 2005 bmanning@vacation.karoshi.com wrote:
On Thu, Mar 24, 2005 at 12:18:34PM -0800, Bill Woodcock wrote:
On Thu, 24 Mar 2005, G Pavan Kumar wrote:
Actually, I am not doing what you think I am. I am using the RouteViews aggregation of the BGP routing tables. RouteViews is a project at the univ. of Oregon that peers with backbones.
Really? Could you tell us more about it? I thought there was just one Internet backbone.
Bill... Stop it!!! shooting fish in a barrel is no sport at all.
You think I am a fish in a barrel? Well, guess what, I didnt think it through while entering your mouth that you're dumb enough to prefer it rather in a barrel!!
I am looking at almost full and fresh data.
So what value do you assign to "almost full?" There's a difference between "best" and "complete," which you may not be entirely appreciating.
-Bill
almost full == just after dessert
and as you (and almost every one else on this list) know, there is zero chance of "complete" ... and "best" is always in the eye/routing-table of the beholder.
--bill
On Thu, 24 Mar 2005, Bill Woodcock wrote:
On Thu, 24 Mar 2005, G Pavan Kumar wrote:
Actually, I am not doing what you think I am. I am using the RouteViews aggregation of the BGP routing tables. RouteViews is a project at the univ. of Oregon that peers with backbones.
Really? Could you tell us more about it? I thought there was just one Internet backbone.
Would you excuse me if I didnt predict that you couldnt improvize and make out of the context?
I am looking at almost full and fresh data.
So what value do you assign to "almost full?" There's a difference between "best" and "complete," which you may not be entirely appreciating.
-Bill
On Mar 25, 2005, at 12:25 AM, G Pavan Kumar wrote:
On Thu, 24 Mar 2005, Bill Woodcock wrote:
On Thu, 24 Mar 2005, G Pavan Kumar wrote:
Actually, I am not doing what you think I am. I am using the RouteViews aggregation of the BGP routing tables. RouteViews is a project at the univ. of Oregon that peers with backbones.
Really? Could you tell us more about it? I thought there was just one Internet backbone.
Would you excuse me if I didnt predict that you couldnt improvize and make out of the context?
Okie, this has gone on long enough. If you would like some help, please stop, take a deep breath, count to ten slowly, then ask nicely and some people here might teach you something. Woody's sarcasm might have annoyed you, but your repeated flames (and not even good ones!) at the people you asked to help you annoy all of us. If you do not want any help, you are welcome to continue in your misunderstanding of how the Internet works. -- TTFN, patrick
On Fri, 25 Mar 2005, Patrick W Gilmore wrote:
Okie, this has gone on long enough.
If you would like some help, please stop, take a deep breath, count to ten slowly, then ask nicely and some people here might teach you something.
May be you should spend more time on networking than your partime job of yoga teaching!
Woody's sarcasm might have annoyed you, but your repeated flames (and not even good ones!) at the people you asked to help you annoy all of us. well guess who wouldnt think that if not being helped a minuscule amount, why not be part of the fun!
If you do not want any help, you are welcome to continue in your misunderstanding of how the Internet works.
I am sorry, am I not ingratiating myself with the good graces of the father of Internet?!
On Fri, Mar 25, 2005 at 12:41:37PM +0530, G Pavan Kumar wrote:
Okie, this has gone on long enough.
If you would like some help, please stop, take a deep breath, count to ten slowly, then ask nicely and some people here might teach you something. May be you should spend more time on networking than your partime job of yoga teaching!
Woody's sarcasm might have annoyed you, but your repeated flames (and not even good ones!) at the people you asked to help you annoy all of us. well guess who wouldnt think that if not being helped a minuscule amount, why not be part of the fun!
If you do not want any help, you are welcome to continue in your misunderstanding of how the Internet works.
I am sorry, am I not ingratiating myself with the good graces of the father of Internet?!
<plonk> Cheers, -- jr 'someone had to do it' a -- Jay R. Ashworth jra@baylink.com Designer Baylink RFC 2100 Ashworth & Associates The Things I Think '87 e24 St Petersburg FL USA http://baylink.pitas.com +1 727 647 1274 If you can read this... thank a system adminstrator. Or two. --me
i suspect that, rather than being a sociologist trying to catorgize operators on the surliness scale, at which you seem to have succeeded splendidly, you may be tring to understand AS relationships. you may find a useful paper trail that kinda goes from <http://citeseer.ist.psu.edu/rd/22812611%2C453493%2C1%2C0.25%2CDownload/http%3AqSqqSqwww-unix.ecs.umass.eduqSq%7ElgaoqSqton.ps> to the more current and more intuitively believable <http://citeseer.ist.psu.edu/rd/59669334%2C702627%2C1%2C0.25%2CDownload/http%3AqSqqSqwww.ieee-infocom.orgqSq2004qSqPapersqSq34_2.PDF> randy
On Fri, Mar 25, 2005 at 07:43:22AM -0800, Randy Bush wrote:
i suspect that, rather than being a sociologist trying to catorgize operators on the surliness scale, at which you seem to have succeeded splendidly, you may be tring to understand AS relationships. you may find a useful paper trail that kinda goes from
to the more current and more intuitively believable
randy
thank you randy. and for the more energetic, who can tell me all the BGP relationships w/ AS 4555? (existing peers and operators of AS 4555 are not allowed to play :) --bill
On Fri, 25 Mar 2005 12:41:37 +0530 (IST), G Pavan Kumar <pavanji@cse.iitb.ac.in> wrote:
May be you should spend more time on networking than your partime job of yoga teaching!
Pavan, what you seem to know of networking is not just outdated, its plain wrong. And you've just managed to flame some of the people who understand it well enough to be the among the best teachers of bgp, routing and DNS that I know of - and I've seen them teach at many a netops conference. And some others who have enable on huge, worldwide networks that you'd kill to get an interview call from, let alone a "you're hired" email. I know that passing the JEE exam to get into an IIT means that you do have more brains than the average engineering student - so try to apply them for a change. Follow that first, original suggestion of googling for "partial mesh", "peering", etc. Then download some of Philip Smith's elementary bgp tutorials - google will find those for you too. Finally, google for the names of the people who have replied to you in this thread. Till then, do yourself a favor. Go right back to lurking on nanog
I am sorry, am I not ingratiating myself with the good graces of the father of Internet?!
No. All you are doing is pissing off the people who have been giving you the right answers. If what they tell you means you have to tear up a whole paper thats based on wrong assumptions, and start from scratch, tough. It'll save you the hassle of seeing your prof fling it right back at your face with an F scrawled on it when you turn it in. -- Suresh Ramasubramanian (ops.lists@gmail.com)
On Fri, 25 Mar 2005 21:15:25 +0530, Suresh Ramasubramanian <ops.lists@gmail.com> wrote:
And some others who have enable on huge, worldwide networks that you'd kill to get an interview call from, let alone a "you're hired" email.
speaking of hiring .. you'd better pray that your brief foray into nanog goes unnoticed by people who will be interviewing you for a job that lets you anywhere near a $90 linksys box let alone cisco or juniper kit :)
participants (9)
-
Bill Woodcock -
bmanning＠vacation.karoshi.com -
G Pavan Kumar -
Jay R. Ashworth -
Michael Loftis -
Michael.Dillon＠radianz.com -
Patrick W Gilmore -
Randy Bush -
Suresh Ramasubramanian