Query: What policies do backbone providers use to determine IP ownership?
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 All, I'm curious to what extent everyone is checking to determine ownership of IP addresses when taking on new customers. Lately, multi-homing has become a very hot topic for even the smallest of providers. With that, customers are bringing along their IP addresses from their previous providers. Are we required, as providers, to determine if that block is actually owned by that customer, and facilitates good Internet routing? I've seen a trend lately where I'm finding out, after the fact, where pieces of larger CIDR blocks are being taken apart by a myriad of unaggregated routes. The other backbone providers freely allowed an announcement of that non-portable space to the Internet without regard to either the owning provider, or to general Internet routing. My concern is two fold: 1) This contributes to terrible Internet routing. By not addressing this with the customer right away, we'll continue to deal with a proliferation of /24s and Internet bloat. I realize the customer needs its address space to announce separately, but should we allow them to freely announce random /24's? This is only due to that the customer received IPs by growing over the years, rather than getting a single block up front. 2) It seems that other providers are allowing our customers to hijack our routing space piece by piece. I will happily participate in multihoming a customer, but I would hope it involves us. We can make a contiguous allocation from our CIDR blocks, and then work with the customer in a more consistent manner. Much of this is customer education about multihoming, but unfortunately we often find out too late. So the question becomes, what do providers do to determine where a block is coming from, and what is its implications on the global routing system? Just cutting and pasting an email from the customer into an access-list seems to be what we have now... I'd be interested to hear what others thoughts and experience are with this. Perhaps I'm just overly concerned with a normal happening on the Internet. Comments? Tony -----BEGIN PGP SIGNATURE----- Version: PGPfreeware 7.0.3 for non-commercial use <http://www.pgp.com> iQA/AwUBOwvHa89/cCqjBxGyEQLVUACfeyX9zbWUbhYQyRWP82f+jU5FhXUAoPHT pmjEerQRWEtEJHG0OvMUxDdP =t+ws -----END PGP SIGNATURE-----
Customers want to be able to multihome. This is not an unreasonable desire, and it is what they are, ostensibly, paying us for. Your main complaint seems to be advertisement of individual netblocks from larger aggregates. While this does contribute to the growth of the internet routing table, this is an artifact of the way routing processes and BGP work, rather than any fault from your customers. What are your customer's options for multihoming? 1) Advertise smaller blocks out of larger aggregates to other providers 2) Get their own, PI space, and advertise it - if they are large enough to obtain it. Of course, this causes one or more additional route announcements as well. 3) The customer sucks it up, and sticks with a single provider Clearly, most customers are going with #1 or #2, which will increase the size of the routing table. They can minimize this by renumbering into consecutive address space, but that assumes that this greatly disrupt their business, and that they can indeed get a quantity of consecutive space from an upstream. "It seems that other providers are allowing our customers to hijack our routing space piece by piece." Is your complaint that other providers aren't calling you to get permission to route your space? Although there may be some disagreement on the etiquette of routing other provider's blocks, if someone has a block swip'ed to them, that's pretty much license to route that block through a secondary provider, unless the policy of the original provider specifically forbids it. And in that case, the duty is on the customer to know his provider's policies, rather than on the second provider to somehow research the policies of the first provider. Of course, in the absence of affirmative WHOIS information stating that a customer has the right to advertise a block, it's wise to get written permission from the "owner of the block". The routing table will increase in size, as the number of issued AS numbers go up, as multihoming increases, as new address space is advertised. However, current core router hardware is more than capable of dealing with this growth. I certainly encourage customers to aggregate wherever possible. However, requiring renumbering is a heavy, and unnecessary burden. Multihoming has become part of basic transit service functionality. - Daniel Golding Tony Mumm Said....
All,
I'm curious to what extent everyone is checking to determine ownership of IP addresses when taking on new customers.
Lately, multi-homing has become a very hot topic for even the smallest of providers. With that, customers are bringing along their IP addresses from their previous providers. Are we required, as providers, to determine if that block is actually owned by that customer, and facilitates good Internet routing?
I've seen a trend lately where I'm finding out, after the fact, where pieces of larger CIDR blocks are being taken apart by a myriad of unaggregated routes. The other backbone providers freely allowed an announcement of that non-portable space to the Internet without regard to either the owning provider, or to general Internet routing.
My concern is two fold:
1) This contributes to terrible Internet routing. By not addressing this with the customer right away, we'll continue to deal with a proliferation of /24s and Internet bloat. I realize the customer needs its address space to announce separately, but should we allow them to freely announce random /24's? This is only due to that the customer received IPs by growing over the years, rather than getting a single block up front.
2) It seems that other providers are allowing our customers to hijack our routing space piece by piece. I will happily participate in multihoming a customer, but I would hope it involves us. We can make a contiguous allocation from our CIDR blocks, and then work with the customer in a more consistent manner. Much of this is customer education about multihoming, but unfortunately we often find out too late.
So the question becomes, what do providers do to determine where a block is coming from, and what is its implications on the global routing system? Just cutting and pasting an email from the customer into an access-list seems to be what we have now...
I'd be interested to hear what others thoughts and experience are with this. Perhaps I'm just overly concerned with a normal happening on the Internet.
Comments?
Tony
*** END PGP VERIFIED MESSAGE ***
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 I certainly agree with your statements that customers want to multihome, and that we will have to accommodate them. It is quite reasonable, and we encourage it where high availability is required. I will also accept that renumbering can be a great amount of work for a customer. I think that amount of work, however, is quite justified if the customer has received portable space from the registrar. I think the customer would be just as motivated to get out of ISP provided IP space...just to be flexible the day they decide to add/drop a provider. On the technical front, with routers being what they are, I can see that we are not in danger of hurting core routers with bloat. Well, most of us anyway...I know of small shops that have had to swap for more memory, etc as the table grew. I can now say that their method of multihoming is the whole reason they need to buy a new router (just kidding). Secondary to that however, is that if I don't know what's happening with the customer, how can we set this up correctly. If no BGP peering session exists for that customer, they will announce their smaller block as the only path, until its withdrawn and only the CIDR remains. This is where I feel the word hijacking applies. We would work with them to set it up correctly so an AS path for this smaller block is passed through us as well as the other customer. I guess it comes down to etiquette in the end. If I whois the block, and its non-portable space allocated to someone, I feel they should know I'm going to announce it. This may have technical repercussions on their network that they may want to know about. Customer education about the policies of the ISP who allocates the IP space certainly will help I think, however customers (bless their soul) will often do what suits them best regardless of policies. Knowing the policy is probably the only way to have reproach if a problem situation comes up. The only check and balance in this whole process is the person that is adding the route to their accept policies. It seems a small inconvenience to check with the owning provider, and a great courtesy to both ISPs. In the end, communication will not only benefit the ISP, but also the customer in the form of better routing (and understanding). - -----Original Message----- From: Daniel Golding [mailto:dan@netrail.net] Sent: Wednesday, May 23, 2001 11:14 AM To: Tony Mumm; nanog@nanog.org Subject: RE: Query: What policies do backbone providers use to determine IP ownership? Customers want to be able to multihome. This is not an unreasonable desire, and it is what they are, ostensibly, paying us for. Your main complaint seems to be advertisement of individual netblocks from larger aggregates. While this does contribute to the growth of the internet routing table, this is an artifact of the way routing processes and BGP work, rather than any fault from your customers. What are your customer's options for multihoming? 1) Advertise smaller blocks out of larger aggregates to other providers 2) Get their own, PI space, and advertise it - if they are large enough to obtain it. Of course, this causes one or more additional route announcements as well. 3) The customer sucks it up, and sticks with a single provider Clearly, most customers are going with #1 or #2, which will increase the size of the routing table. They can minimize this by renumbering into consecutive address space, but that assumes that this greatly disrupt their business, and that they can indeed get a quantity of consecutive space from an upstream. "It seems that other providers are allowing our customers to hijack our routing space piece by piece." Is your complaint that other providers aren't calling you to get permission to route your space? Although there may be some disagreement on the etiquette of routing other provider's blocks, if someone has a block swip'ed to them, that's pretty much license to route that block through a secondary provider, unless the policy of the original provider specifically forbids it. And in that case, the duty is on the customer to know his provider's policies, rather than on the second provider to somehow research the policies of the first provider. Of course, in the absence of affirmative WHOIS information stating that a customer has the right to advertise a block, it's wise to get written permission from the "owner of the block". The routing table will increase in size, as the number of issued AS numbers go up, as multihoming increases, as new address space is advertised. However, current core router hardware is more than capable of dealing with this growth. I certainly encourage customers to aggregate wherever possible. However, requiring renumbering is a heavy, and unnecessary burden. Multihoming has become part of basic transit service functionality. - - Daniel Golding Tony Mumm Said....
All,
I'm curious to what extent everyone is checking to determine ownership of IP addresses when taking on new customers.
Lately, multi-homing has become a very hot topic for even the smallest of providers. With that, customers are bringing along their IP addresses from their previous providers. Are we required, as providers, to determine if that block is actually owned by that customer, and facilitates good Internet routing?
I've seen a trend lately where I'm finding out, after the fact, where pieces of larger CIDR blocks are being taken apart by a myriad of unaggregated routes. The other backbone providers freely allowed an announcement of that non-portable space to
the
Internet without regard to either the owning provider, or to general Internet routing.
My concern is two fold:
1) This contributes to terrible Internet routing. By not addressing this with the customer right away, we'll continue to deal with a proliferation of /24s and Internet bloat. I realize the customer needs its address space to announce separately, but should we allow them to freely announce random /24's? This is only due to that the customer received IPs by growing over the years, rather than getting a single block up front.
2) It seems that other providers are allowing our customers to hijack our routing space piece by piece. I will happily participate in multihoming a customer, but I would hope it involves us. We can make a contiguous allocation from our CIDR blocks, and then work with the customer in a more consistent manner. Much of this is customer education about multihoming, but unfortunately we often find out too late.
So the question becomes, what do providers do to determine where a block is coming from, and what is its implications on the global routing system? Just cutting and pasting an email from the customer into an access-list seems to be what we have now...
I'd be interested to hear what others thoughts and experience are with this. Perhaps I'm just overly concerned with a normal happening on the Internet.
Comments?
Tony
*** END PGP VERIFIED MESSAGE ***
-----BEGIN PGP SIGNATURE----- Version: PGPfreeware 7.0.3 for non-commercial use <http://www.pgp.com> iQA/AwUBOwvpcs9/cCqjBxGyEQLvugCeO1qsZwf6en5EeJ0iL2nAahv9EkYAn21e g4p4N3vva30X9qDb8tKXfu5D =b5mb -----END PGP SIGNATURE-----
On Wed, 23 May 2001, Tony Mumm wrote:
I've seen a trend lately where I'm finding out, after the fact, where pieces of larger CIDR blocks are being taken apart by a myriad of unaggregated routes. The other backbone providers freely allowed an announcement of that non-portable space to the Internet without regard to either the owning provider, or to general Internet routing.
Here's a related question. Suppose provider A has a customer C who multihomes with a connection to A and provider B. C uses IP's assigned by A. C terminates service with A...but keeps announcing A's space to B. B propogates the routes to their peers. B and C ignore requests that they stop using A's space and renumber into B's. How does A reclaim their space? One obvious solution is dueling routes...A could announce more specific routes, fouling things up for B and C hoping this will serve as encouragement for C to renumber. -- ---------------------------------------------------------------------- Jon Lewis *jlewis@lewis.org*| I route System Administrator | therefore you are Atlantic Net | _________ http://www.lewis.org/~jlewis/pgp for PGP public key_________
I have seen dueling routes more often than I'd like to recall. No one wants to use space that isn't 100% routable. Assuming A has any peers, some traffic will go there and dead-end. Any customer in the block will be unhappy. B will eventually give up. A would be foolish to assign customers in the block until B has given up. If A is a much better connected network than B (since B is acting like an ass) the process is much more definitive. C can, and should be ignored. Just my opinion, I could be wrong. Deepak Jain AiNET -----Original Message----- From: owner-nanog@merit.edu [mailto:owner-nanog@merit.edu]On Behalf Of jlewis@lewis.org Sent: Wednesday, May 23, 2001 10:31 PM To: Tony Mumm Cc: nanog@nanog.org Subject: Re: Query: What policies do backbone providers use to determine IP ownership? On Wed, 23 May 2001, Tony Mumm wrote:
I've seen a trend lately where I'm finding out, after the fact, where pieces of larger CIDR blocks are being taken apart by a myriad of unaggregated routes. The other backbone providers freely allowed an announcement of that non-portable space to the Internet without regard to either the owning provider, or to general Internet routing.
Here's a related question. Suppose provider A has a customer C who multihomes with a connection to A and provider B. C uses IP's assigned by A. C terminates service with A...but keeps announcing A's space to B. B propogates the routes to their peers. B and C ignore requests that they stop using A's space and renumber into B's. How does A reclaim their space? One obvious solution is dueling routes...A could announce more specific routes, fouling things up for B and C hoping this will serve as encouragement for C to renumber. -- ---------------------------------------------------------------------- Jon Lewis *jlewis@lewis.org*| I route System Administrator | therefore you are Atlantic Net | _________ http://www.lewis.org/~jlewis/pgp for PGP public key_________
On Wed, 23 May 2001 jlewis@lewis.org wrote:
Here's a related question. Suppose provider A has a customer C who multihomes with a connection to A and provider B. C uses IP's assigned by A. C terminates service with A...but keeps announcing A's space to B. B propogates the routes to their peers. B and C ignore requests that they stop using A's space and renumber into B's. How does A reclaim their space?
One obvious solution is dueling routes...A could announce more specific routes, fouling things up for B and C hoping this will serve as encouragement for C to renumber.
And then C could announce more specific routes, and so forth, although you'll run into filters and start being ignored at some point. That's really the sort of problem better resolved by lawyers than by engineers. -Steve -------------------------------------------------------------------------------- Steve Gibbard scg@gibbard.org
On Thu, May 24, 2001 at 07:25:14PM -0700, Steve Gibbard wrote:
On Wed, 23 May 2001 jlewis@lewis.org wrote:
Here's a related question. Suppose provider A has a customer C who multihomes with a connection to A and provider B. C uses IP's assigned by A. C terminates service with A...but keeps announcing A's space to B. B propogates the routes to their peers. B and C ignore requests that they stop using A's space and renumber into B's. How does A reclaim their space?
One obvious solution is dueling routes...A could announce more specific routes, fouling things up for B and C hoping this will serve as encouragement for C to renumber.
And then C could announce more specific routes, and so forth, although you'll run into filters and start being ignored at some point.
That's really the sort of problem better resolved by lawyers than by engineers.
-Steve
-------------------------------------------------------------------------------- Steve Gibbard scg@gibbard.org
Why go through all that mess? You call up their new upstream, say you're a representative of XYZ, and customer ABC does not have permission to announce your netblocks. I know InterNAP requires that they get consent from the netblock contact, I don't know about other companies. And so it doesn't happen as a prank, while on phone with them, tell the person to send an e-mail to the netblock contact, you open up the e-mail and read it back to them for verification, it seriously can't be that complicated. kthkxbye. -- Omachonu Ogali missnglnk@informationwave.net http://www.informationwave.net
participants (6)
-
Daniel Golding -
Deepak Jain -
jlewis＠lewis.org -
Omachonu Ogali -
Steve Gibbard -
Tony Mumm