|> From: Randy Bush [mailto:randy@psg.com] |> Sent: Monday, August 27, 2001 4:58 PM |> |> >>> The whole problem seems to me to be a lack of a |> micro-allocation policy, |> >>> and an agreement from providers that they will not |> filter that space. |> >> judicious use of this might be helpful. the problem is |> the 'judge' in |> >> judicious. |> > indeed. but it would still disassociate the space conservation and |> > table growth problems to some extent; the "judge" being the same |> > as today. |> |> ahhh, somebody understood the comment. |> |> the problem is that there is no obvious detent on the knob. |> to start the |> usual posturing and flamage, i propose min 2xDS3 multihomed. I would sincerely like to know why not 2xDS1 multi-homed. Most businesses can't cost-justify dual DS3's and many (UNited Airlines) gate their entire company through a T1 (not counting the ecommerce stuff, which is colo'd as it should be).
On Mon, 27 Aug 2001, Roeland Meyer wrote:
|> the problem is that there is no obvious detent on the knob. |> to start the usual posturing and flamage, i propose min 2xDS3 |> multihomed.
I would sincerely like to know why not 2xDS1 multi-homed.
A quick scan through the customer database shows customers on 2*256 links (approx) and there might be smaller ones. I wouldn't be surprised to see multihomed people with 64k or smaller links. Please remember that that bandwidth costs a lot more in most countries than it does in the US. -- Simon Lyall. | Newsmaster | Work: simon.lyall@ihug.co.nz Senior Network/System Admin | Postmaster | Home: simon@darkmere.gen.nz ihug, Auckland, NZ | Asst Doorman | Web: http://www.darkmere.gen.nz
2xds3 as a minimum, that seems a bit steep. I personally would have no problem with someone going with frac ds1 with 2 different providers in order to protect themselves from one provider's hardware, circuit, or routing failure. Sometimes people for security or ease of physical access reasons want hardware at their premises. Brian "Sonic" Whalen Success = Preparation + Opportunity On Tue, 28 Aug 2001, Simon Lyall wrote:
On Mon, 27 Aug 2001, Roeland Meyer wrote:
|> the problem is that there is no obvious detent on the knob. |> to start the usual posturing and flamage, i propose min 2xDS3 |> multihomed.
I would sincerely like to know why not 2xDS1 multi-homed.
A quick scan through the customer database shows customers on 2*256 links (approx) and there might be smaller ones. I wouldn't be surprised to see multihomed people with 64k or smaller links.
Please remember that that bandwidth costs a lot more in most countries than it does in the US.
-- Simon Lyall. | Newsmaster | Work: simon.lyall@ihug.co.nz Senior Network/System Admin | Postmaster | Home: simon@darkmere.gen.nz ihug, Auckland, NZ | Asst Doorman | Web: http://www.darkmere.gen.nz
the point of 2xDS3 was specifically to get major services, and not to get every basement dual-homer. do the latter and you have the same grazing of the commons as we have today. randy
On Tue, 28 Aug 2001, Randy Bush wrote:
the point of 2xDS3 was specifically to get major services, and not to get every basement dual-homer.
Please explain why the "basement dual-homer" should not have the same right to diversity as the "major services." And please, be specific.
Patrick Greenwell wrote:
Please explain why the "basement dual-homer" should not have the same right to diversity as the "major services."
And please, be specific.
The quesetion is bogus, there is no such thing as a right to have a route in my router without paying me for it. If I choose to extend that privilege to people who meet certain minimum requirements because I believe the benefits will outweight the costs, then that's *my* right. All others can pay me to do it if they want me to. Your rights end at my network. DS
On Mon, 27 Aug 2001, David Schwartz wrote:
Patrick Greenwell wrote:
Please explain why the "basement dual-homer" should not have the same right to diversity as the "major services."
And please, be specific.
The quesetion is bogus, there is no such thing as a right to have a route in my router without paying me for it.
If I'm not mistaken, we were talking about the standards for micro-allocations with respects to the RIRs and not your network specifically. If I was speaking about dictating your rights to set your own network policy, I would have mentioned it.
On Mon, 27 Aug 2001, David Schwartz wrote:
Patrick Greenwell wrote: If I'm not mistaken, we were talking about the standards for micro-allocations with respects to the RIRs and not your network specifically. If I was speaking about dictating your rights to set your own network policy, I would have mentioned it.
What good will a microallocation from the RIR do if it's not routable? If you think you can talk about allocation policy without considering routing policy, you are mistaken. DS
On Mon, 27 Aug 2001, David Schwartz wrote:
On Mon, 27 Aug 2001, David Schwartz wrote:
Patrick Greenwell wrote: If I'm not mistaken, we were talking about the standards for micro-allocations with respects to the RIRs and not your network specifically. If I was speaking about dictating your rights to set your own network policy, I would have mentioned it.
What good will a microallocation from the RIR do if it's not routable?
None of course. The question is would you honor microallocations from a RIR if they said "we have designated this particular space for microallocations, would you please accept routes for these netblocks with these prefixes?" It is of course completely up to you, and I don't believe anyone was questioning or challenging the sovereignty of your network. Interne routing works due to cooperation, not coersion.
Patrick Greenwell:
None of course.
The question is would you honor microallocations from a RIR if they said "we have designated this particular space for microallocations, would you please accept routes for these netblocks with these prefixes?"
That's what this discussion was about, until the right to have it was questioned. (That's what the 'who decides' questions are about, aren't they?) ARIN could decide to issue /32's to dialup customers so they could change providers without renumbering and it would do no good since nobody would carry those routes. Providers generally filter on allocation boundaries (or are more generous) because they trust the RIRs to set sane allocation policies. It would be a waste of IP space and harm the net as a whole if RIRs adopted a microallocation policy that was too generous and resulted in allocating non-routable IP space. Hence the desire to discuss amount network operators what a reasonable microallocation policy would look like. This is why comments like:
Please explain why the "basement dual-homer" should not have the same right to diversity as the "major services."
And please, be specific.
Don't make any sense. That was my point in replying to you. There is no right to a route in my router. If you want a route in my router, you better find out what routes I'm willing to carry and under what terms. DS
Hi, At 12:14 AM 8/28/2001 -0700, David Schwartz wrote:
It would be a waste of IP space and harm the net as a whole if RIRs adopted a microallocation policy that was too generous and resulted in allocating non-routable IP space.
One could argue that the RIRs are wasting address space by allocating on arbitrary boundaries, e.g., /20s, instead of allocating according to documented requirements. One could also argue that the "sane" allocation policies of the RIRs have resulted in ISPs not being forced to figure out how to apply effective mechanisms to limit route prefix growth and as a direct result created a tragedy of the commons in the DFZ. Of course, I wouldn't argue those positions... :-) Rgds, -drc Speaking for no one (and/or nothing) but myself
One could argue that the RIRs are wasting address space by allocating on arbitrary boundaries, e.g., /20s, instead of allocating according to documented requirements.
If someone were to argue that, someone could reply that unless people cheat, no IP address space is wasted because the registries still only allocate based upon demonstrated need. One could even argue that a smaller allocation policy saves IP space because it stops people from cheating by asking for more IP space than they need.
One could also argue that the "sane" allocation policies of the RIRs have resulted in ISPs not being forced to figure out how to apply effective mechanisms to limit route prefix growth and as a direct result created a tragedy of the commons in the DFZ.
I'm not sure I believe that this tragedy of the commons exists where people route on allocation boundaries. If I make Sprint carry an extra route just for my little network, that helps all Sprint customers reach my little network. I may not have many hosts, but Sprint has many, and each of those reach my just a bit better. A distinct route for a distinct network of at least some minimal value doesn't create a tragedy of the commons. Where you do have a tragedy of the commons is where people place routes without technical justification. A sane microallocation policy shouldn't exacerbate this. In any event, historically the dog has wagged the tail and the tail has wagged the dog. DS
David, At 08:07 AM 8/28/2001 -0700, David Schwartz wrote:
If someone were to argue that, someone could reply that unless people cheat, no IP address space is wasted because the registries still only allocate based upon demonstrated need.
While "demonstrated need" is easy to say, it is much more difficult to actually verify, particularly when the demonstrated need is projected into the future.
One could even argue that a smaller allocation policy saves IP space because it stops people from cheating by asking for more IP space than they need.
Exactly. The RIRs are forced to balance conservation of the remaining free pool of addresses (the only thing the RIRs really have any control over and even that is pretty tenuous) with the number of route prefixes in the default free zone (something the RIRs have no control over but which ISPs do). Historically (since CIDR and 2050), the balance has been swung towards limiting the number of prefixes in the DFZ, primarily by restricting the number of new prefixes allocated (there were other policies, e.g., APNIC's policy permitting the return of multiple prefixes for a single prefix of the next largest CIDR block with no questions asked, but most of the focus has been on preventing new prefixes from being allocated). From my perspective, the whole point of micro-allocations is to try to move the balance back towards neutral a bit. Address space would be allocated for those applications that need to be announced in the DFZ but which don't represent a large amount of address space. Of course, figuring out exactly what those applications are will be a bit of a challenge for the policy makers, but hey, that's what they get paid for (well, if they got paid for doing it, of course).
I'm not sure I believe that this tragedy of the commons exists where people route on allocation boundaries.
The tragedy of the commons exist because there is a limited resource, incentive to do the wrong thing, and disincentives to do the right thing. Until there are disincentives to do the wrong thing, e.g., filter routes, apply a charge to routes in the DFZ to encourage aggregation, etc., incentives to do the right thing, and/or the limitations in the DFZ are removed, you _will_ get a tragedy of the commons.
A distinct route for a distinct network of at least some minimal value doesn't create a tragedy of the commons.
Of course it can.
Where you do have a tragedy of the commons is where people place routes without technical justification.
Technical justification does not remove the limitations on a resource, it merely allows triage as to who gets to use the resource. Micro-allocations and filtering are treating symptoms. The underlying disease (rational route announcement policy) could conceivably be treated by applying standard market economics to the problem, but there hasn't yet been enough incentive to figure out how to do it (and/or get over the historical resistance to doing it). Rgds, -drc Speaking only for myself
drc writes:
Micro-allocations and filtering are treating symptoms. The underlying disease (rational route announcement policy)
I agree with David's analysis of the problem right up to this point; from my pespective the underlying disease is the route convergence/routing table growth issue. The route announcement policy's rationality has to be judged in the light of this greater pathology. It *is* the real problem, though, and we have to keep it in mind to understand that any invocation of market forces, draconian filters, or community-mindedness only delays the progress of the underlying pathology. In a sense, they too are symptoms, however rational they may appear in today's lights. Or as Randy puts, we're playing whack-a-mole here, and we're doing it knowing it only buys us time. The underlying problem is real, and it will take time to develop a real solution. Encourage your vendors to participate in the appropriate working groups, tell them you'll spend money on products that fix it, and be ready as an operator to help debug potential solutions. It's important. End of pleading.... regards, Ted (Not speaking for Equinix; speaking for David. No wait, I meant *to* David.)
The tragedy of the commons exist because there is a limited resource, incentive to do the wrong thing, and disincentives to do the right thing. Until there are disincentives to do the wrong thing, e.g., filter routes, apply a charge to routes in the DFZ to encourage aggregation, etc., incentives to do the right thing, and/or the limitations in the DFZ are removed, you _will_ get a tragedy of the commons.
Rgds, -drc Speaking only for myself
The limited resource is the fixed upper bound on numbers. There are concerns w/ the current technological limitations on mgmt of the route table and the weaknesses in the current routing assumptions. As friend Bush has indicated, the IRTF and in the IETF, much thought is being given to how to migrate from BGP to something new. Perhaps heirarchical routing itself is flawed and we need something new. This problem is not new. As a data point, I would ask that those whom are allowed to participate in the design discussions and are willing to be active in them, to take this request into those discussions. I would like to see the ability to have the routing system support 2-32nd entries in the "DFZ" (whatever that is... :) --bill
Bill, Your comments regarding the hierarchical routing may be valid, I believe, -- especially taking into consideration that the whole subject of this thread is just one of the drivers inevitably flattening the Internet topology :) But look what's going on in MANET, for example (LANMAR being particularly amazing). This at least suggests that some scenarios may exist where you have neither strictly hierarchical routing nor '2-32nd entries in the "DFZ"'. -- dima.
-----Original Message----- From: owner-nanog@merit.edu [mailto:owner-nanog@merit.edu]On Behalf Of bmanning@vacation.karoshi.com Sent: Tuesday, August 28, 2001 6:22 PM To: David R. Conrad Cc: nanog@merit.edu Subject: Re: multi-homing fixes
The tragedy of the commons exist because there is a limited resource, incentive to do the wrong thing, and disincentives to do the right thing. Until there are disincentives to do the wrong thing, e.g., filter routes, apply a charge to routes in the DFZ to encourage aggregation, etc., incentives to do the right thing, and/or the limitations in the DFZ are removed, you _will_ get a tragedy of the commons.
Rgds, -drc Speaking only for myself
The limited resource is the fixed upper bound on numbers. There are concerns w/ the current technological limitations on mgmt of the route table and the weaknesses in the current routing assumptions. As friend Bush has indicated, the IRTF and in the IETF, much thought is being given to how to migrate from BGP to something new. Perhaps heirarchical routing itself is flawed and we need something new. This problem is not new.
As a data point, I would ask that those whom are allowed to participate in the design discussions and are willing to be active in them, to take this request into those discussions. I would like to see the ability to have the routing system support 2-32nd entries in the "DFZ" (whatever that is... :)
--bill
On Tue, 28 Aug 2001 bmanning@vacation.karoshi.com wrote:
As a data point, I would ask that those whom are allowed to participate in the design discussions and are willing to be active in them, to take this request into those discussions. I would like to see the ability to have the routing system support 2-32nd entries in the "DFZ" (whatever that is... :)
4 billion routes is not impossible, although I don't think one out of every two people on the entire planet is going to multihome. 100 million seems more reasonable. In either case, this means we have to find a completely new way to look at routes. The current paradigm is that every route is very important, so we should store as much information about it as possible. This will have to change. If we remove all non-essential information from a route, we finally arrive at the single thing that must always be encoded for each route individually: whether it is reachable or not. If we assign a bit of memory to every possible route, it is possible to store the reachability state of the entire Internet as /24s in just two megabytes. Or as individual /32s in 512 MB. Obviously, a lot of work has to be done to apply this to the real world. An idea would be to assign /16s to geographic areas. Each ISP that has customers in that area would announce the /16, just like they would do now, but with an attached bitmap that indicates for which /24s this announcement is valid and for which it isn't. So 10 ISPs in one area would all announce the /16 with a 256 bit bitmap, so just 10 routes end up in the default-free zone instead of 500. Iljitsch van Beijnum
* Thus spake David Schwartz (davids@webmaster.com): [snip]
The quesetion is bogus, there is no such thing as a right to have a route in my router without paying me for it. If I choose to extend that privilege to people who meet certain minimum requirements because I believe the benefits will outweight the costs, then that's *my* right. All others can pay me to do it if they want me to. Your rights end at my network.
DS
I agree that there is no 'right' to have a route in someone else's router. Different providers, different policies etc. etc. However, if I choose to filter on allocation boundaries but advertise prefixes to peers that I myself would filter based on my own policy is that considered hypocritical? Bad form? Acceptable? Just wondering aloud.
I agree that there is no 'right' to have a route in someone else's router. Different providers, different policies etc. etc. However, if I choose to filter on allocation boundaries but advertise prefixes to peers that I myself would filter based on my own policy is that considered hypocritical? Bad form? Acceptable?
normal business. you're doing that for which folk PAY you. randy
On Tue, 28 Aug 2001, Randy Bush wrote:
I agree that there is no 'right' to have a route in someone else's router. Different providers, different policies etc. etc. However, if I choose to filter on allocation boundaries but advertise prefixes to peers that I myself would filter based on my own policy is that considered hypocritical? Bad form? Acceptable?
normal business. you're doing that for which folk PAY you.
Curious that this entire discussion is justified by delivering what your customers pay you for, when what is proposed couldn't be further from that. If this is about what customers pay for, then we would be disussing how to accommodate, and even encourage effective multi-homing at a more granular level. Customers pay for the network to work end-to-end. More choices mean better performance, more reliability. The entire premise for this discussion goes directly against that. Let me guess, this /is/ for the good of the users, because if we don't do it the world will blow up with too many routes. Uh huh. And everyone is turning down customers who want to multi-home a /24. I pay my network providers to reach all those multi-homed /24's quickly and reliably. Filtering devalues your network, I buy from your non-filtering competitor instead. BTW, your sales people (if you are a major carrier) are salivating over my RFP. Your CEO sweats bullets over next quarter's numbers. Filtering /24's doesn't seem important to them. Where did the 'you don't pay me, so you can't use my route table' argument come from? A multi-organizational, ubiquitous, globally-reachable, resilient network presumes that the majority of routes in my router are /not/ my customers, and /that's/ why the network is valuable. I'm not saying there isn't a problem, or that we shouldn't be doing anything about it. But it's one thing to talk about the problem (technology needs to improve to allow individuals and small companies to have better reliability), and quite another for networks to be hypocritically preaching/enforcing the 'pay or be filtered' principle while violating the principle themselves. Pete.
Randy Do you have any Idea of the cost of a T3 from overseas to the US? (or even the cost of a local loop T3 outside the US) Hint: A T3 from Tel-Aviv to Chicago NAP is 6 figures (US$) a _month_ And this is after de-regulation dropped the price of a fiber link to under satellite link price(2-3 years ago a fiber terrestrial link was ~ 3 times a satellite link with an minimum RTT of ~ 550 msec ;-( ) -- Rafi Sadowsky rafi@oumail.openu.ac.il Network/System/Security VoiceMail: +972-3-646-0592 FAX: +972-3-646-0454 Mangler ( :-) | FIRST-REP for ILAN-CERT(CERT@CERT.AC.IL) Open University of Israel | (PGP key -> ) http://telem.openu.ac.il/~rafi On Tue, 28 Aug 2001, Randy Bush wrote:
the point of 2xDS3 was specifically to get major services, and not to get every basement dual-homer. do the latter and you have the same grazing of the commons as we have today.
randy
participants (14)
-
bmanning@vacation.karoshi.com
-
Brian Whalen
-
Craig Pierantozzi
-
David R. Conrad
-
David Schwartz
-
Dmitri Krioukov
-
hardie@equinix.com
-
Iljitsch van Beijnum
-
Patrick Greenwell
-
Pete Kruckenberg
-
Rafi Sadowsky
-
Randy Bush
-
Roeland Meyer
-
Simon Lyall