At 10:17 PM 4/3/96 -0500, Curtis Villamizar wrote:
Provider X takes on some number of customers N that want prefixes and think they may later dual home or want to leave the option of changing providers without renumbering open. Substitute for X as you see fit.
If provider X insists that small providers or small to medium business customers must renumber to leave a CIDR aggregate the smaller organization go off and get "portable" address allocations which put them in the unaggregatable toxic waste dump (TWD). If so, they will also try as hard as they can to get a /19.
Some of the small prefixes go out of business. Some grow and become dual homed. Some switch providers. Most just don't change.
In either case, TWD allocation or out of a provider aggregate, a dual homed customer requires an additional prefix (to get routing right).
If a small prefix changes providers and is TWD allocated, they already have a unique route. If they were allocated from a large provider aggregate, one more prefix is needed. If they were allocated from a large provider aggregate and are given a generous grace period, some will renumber quickly, some not at all (continuous requests to extend the grace period). Lets assume they are never forced (grace period extensions are granted).
If the number of small prefixes that resort to the TWD as a result of strong renumbering policies exceeds the number of small prefixes that move out of aggregates without eventually renumbering, then there the strong renumbering policy actually promotes more growth in the routing table size.
In the short term, the difference may not be all that substantial. Longer term, if the provider community can cooperate to aggregate better then many of the extra routes caused by prefixes changing providers can be aggregated back together over a multple AS aggregation boundary.
Since you made the comment "And the global routing table grows", do you feel what I described above is invalid? If so, what assumptions are you making differently? Do you feel people will never renumber if given a grace period, even if renumbering becomes easier with time?
Curtis
I think that its a fair description. And honestly, I don't think a substantial percentage of end-networks will renumber if there are not substantial incentives. If renumbering becomes less-painful, with time & better tools, perhaps more will renumber, but again I personally don't foresee a substantial number doing so without some incentive(s). The scenario that was previously described by Michael Dillon, I believe, was one in which a singularly-homed [to provider 'a'] end-network [x] moved to another provider [provider 'b'] and wanted to take their provider [a] allocated address(es) with them. This is a case where, if a larger aggregate is being announced by [a], then a specific component announced from the [a] CIDR block would be announced by [b]. Of course, this happens anyway if [x] is dual-homed. I think we can all agree that the peace-of-mind obtained by [x] in becoming dual-homed is less than optimal for the Routing Table Watchers (tm). :-) This just happens to be a Catch-22 with multihomed end-networks. - paul
On Wed, 3 Apr 1996, Paul Ferguson wrote:
And honestly, I don't think a substantial percentage of end-networks will renumber if there are not substantial incentives.
This is true. One incentive could be peer pressure created by industry magazines publishing how-to articles with tools being readily available and the stigma of toxic waste dumps attached to those who don't renumber. Of course, use of RFC1918 addresses like 10/8 coupled with proxy firewalling solves the renumbering problem quite handily but ISP's who are trying to justify /16 allocations would be loathe to recommend use of RFC1918 to their customers. Negative incentives, eh?
The scenario that was previously described by Michael Dillon, I believe, was one in which a singularly-homed [to provider 'a'] end-network [x] moved to another provider [provider 'b'] and wanted to take their provider [a] allocated address(es) with them. This is a case where, if a larger aggregate is being announced by [a], then a specific component announced from the [a] CIDR block would be announced by [b].
One problem ISP's run into is that if they allocate addresses to customer networks and then move to another provider they either need to keep their IP allocation or force their customers to renumber. Many customers may choose to switch ISP's or other nasty things, therefore the ISP would like a way to keep the allocation. I'm not sure my idea was terribly great, the real solution is probably to keep the old T1 for the old customers and buy a new T1 for expansion with the new NSP and a new set of addresses. It's not neccessary to run BGP4 in order to have two T1's from two providers with two different CIDR blocks. Has anyone ever proposed this as a solution to an ISP?
This just happens to be a Catch-22 with multihomed end-networks.
It really is about time that some of the larger ISP's started following the lead of folks like netaxs.com and become aggregate providers for local ISP's in their cities. This way the aggregator can be doubly and triply homed and deal with all the BGP4 nastiness. The ISP's gain the benefit of that multihoming to their city and in addition can get some of the redundancy-in-case-of-failure by buying a T1 and frame relay, or a T1 and ISDN dialup to their aggregate provider. Every ISP wants to have a backup connection and right now most assume that multi-homing is the only way to achieve this. I believe that a middle-tier between the ISP and the NSP is the best way to achieve this and could very well decrease global routing table size. Michael Dillon Voice: +1-604-546-8022 Memra Software Inc. Fax: +1-604-546-3049 http://www.memra.com E-mail: michael@memra.com
sooner or later we will have to kill off the /24's, which make up 70% of the routing table but offer way less than 10% of the total reachable destinations. perhaps now that address ownership has been put to bed, the gang of big providers can agree on a date after which they will all stop listening to or exporting any prefixes longer than /23? THAT would be the incentive the industry needs to look at private addressing and aggressive renumbering. who's willing to risk collusion lawsuits and lost customers? step right up and sign the register please.
On Wed, 3 Apr 1996, Paul A Vixie wrote:
sooner or later we will have to kill off the /24's, which make up 70% of the routing table but offer way less than 10% of the total reachable destinations. perhaps now that address ownership has been put to bed, the gang of big providers can agree on a date after which they will all stop listening to or exporting any prefixes longer than /23? THAT would be the incentive the industry needs to look at private addressing and aggressive renumbering. who's willing to risk collusion lawsuits and lost customers? step right up and sign the register please.
I would recommend that the PIER group work with providers on this; PIER would be a great organization to take the huge ACTIVE table of /24's and mail the listed contacts for the network to offer tools, easier renumbering methods, etc., to minimize impact to the network's customers. Once all the mails are sent out and a semi-generous grace period is set, PIER should recommend a date providers should stop listening to /24 announcements. Granted, ISP's wouldn't have to follow this recommendation and could cut off such announcements at any given time; they follow the risk of more impact to their customers as Paul mentions above. I would ALSO recommend to ISP's who wish to implement this that they not be hypocritical. We've heard the ISP stories where particular ISPs want to filter out routes for larger prefixes, but are GLAD to advertise a /23 if it gains that particular ISP money. /cah
On Wed, 3 Apr 1996, Paul A Vixie wrote:
sooner or later we will have to kill off the /24's, which make up 70% of the routing table but offer way less than 10% of the total reachable destinations. perhaps now that address ownership has been put to bed, the gang of big providers can agree on a date after which they will all stop listening to or exporting any prefixes longer than /23? THAT would be the incentive the industry needs to look at private addressing and aggressive renumbering. who's willing to risk collusion lawsuits and lost customers? step right up and sign the register please.
I'm not sure if this is the most completely wrong place to ask this question, so please forgive me if it is, but I'm not sure where else to ask it... As someone who's about to renumber a public school district from a /24 to something else, what would be the smallest network to get (from InterNIC) that would pretty much be guaranteed to be routed for the next few years? I'm thinking a /22 at the moment, but am not sure. Granted the best solution would be go to our provider (all the schools in Santa Clara County, CA go through the county office of education for internet access) and have them get an /18 or something and distribute that, but they don't seem to want that. Should I push them for this solution? Thanks in advance, and apologies for the 'dumb' traffic. -Sven Nielsen Dalvenjah FoxFire, the Teddy Dragon (also known as Sven Nielsen to some :) dalvenjah@dal.net --- dalvenjah on IRC Remember: if you're not on DALnet, you're on the wrong IRC server!! (/serv irc.dal.net 7000 or telnet telnet.dal.net to try it out) -- ____ _ _ _ "I had the dagger in my hand, and he has | _ \ __ _| |_ _____ _ _ (_)__ _| |_the indecency to start dying on his own!" | |_) / _` | \ V / -_) ' \ | / _` | ' \ --Ambassador G'kar, Babylon 5 |____/\__,_|_|\_/\___|_||_|/ \__,_|_||_| FoxFire -- dalvenjah@dal.net -- (SN90) |__/
On Wed, 3 Apr 1996, Paul A Vixie wrote:
sooner or later we will have to kill off the /24's, which make up 70% of the routing table but offer way less than 10% of the total reachable destinations. perhaps now that address ownership has been put to bed, the gang of big providers can agree on a date after which they will all stop listening to or exporting any prefixes longer than /23? THAT would be the incentive the industry needs to look at private addressing and aggressive renumbering. who's willing to risk collusion lawsuits and lost customers? step right up and sign the register please.
If that happens, at least some businesspeople who read articles or the 'net will simply start demanding /23s and will shop around until they find someone willing to give one to them, even if they have 2 hosts.
I'm not sure if this is the most completely wrong place to ask this question, so please forgive me if it is, but I'm not sure where else to ask it...
As someone who's about to renumber a public school district from a /24 to something else, what would be the smallest network to get (from InterNIC) that would pretty much be guaranteed to be routed for the next few years? I'm thinking a /22 at the moment, but am not sure.
Unless they have a real plan & need to get about 128 buildings full of 100 machines/building online, it's a lost cause. You already need a /18 to be heard from Sprint in >= 207/8. Unless you're a customer of Sprint. But Sprint/Sean's position is that other providers should do the same thing (filter on routes in new address space to preserve status quo but stomp on the growth of the announcement of routes which are "not worth" the expense vs. reachability tradeoff of being inserted in every router with "full routes").
-Sven Nielsen
Avi
On Thu, 4 Apr 1996, Dalvenjah FoxFire wrote: Anybody want to recommend a list (or create one) to discuss this kind of thing on? I'll join it, promote it and leave NANOG alone for a while.
As someone who's about to renumber a public school district from a /24 to something else, what would be the smallest network to get (from InterNIC) that would pretty much be guaranteed to be routed for the next few years? I'm thinking a /22 at the moment, but am not sure.
No. Read RFC1918 and use those addresses in conjunction with a proxy firewall. You'll neve need to renumber again and you gain the added benefit that nobody can contact the outside world without going through your proxy which can block the kiddies from http://www.penthouse.com et al.
Granted the best solution would be go to our provider (all the schools in Santa Clara County, CA go through the county office of education for internet access) and have them get an /18 or something and distribute that, but they don't seem to want that. Should I push them for this solution?
Doesn't matter if you use a *COORDINATED* scheme for allocating RFC1918 addresses. As long each school in the county uses different blocks from 10/8 there will be no problems going it alone today and merging at some future date. Michael Dillon Voice: +1-604-546-8022 Memra Software Inc. Fax: +1-604-546-3049 http://www.memra.com E-mail: michael@memra.com
On Wed, 3 Apr 1996, Paul A Vixie wrote:
sooner or later we will have to kill off the /24's, which make up 70% of the routing table but offer way less than 10% of the total reachable destinations. perhaps now that address ownership has been put to bed, the gang of big providers can agree on a date after which they will all stop listening to or exporting any prefixes longer than /23? THAT would be the incentive the industry needs to look at private addressing and aggressive renumbering. who's willing to risk collusion lawsuits and lost customers? step right up and sign the register please.
You don't risk collusion lawsuits by announcing that you are considering this action and strongly urging people to look at RFC1918 and renumbering. You could probably even get away with something like: "NANOG agrees that the only way to avert the pending collapse of the Internet (see From The Ether, Infoworld, Apr 1 1996) is to stop routing the old Class C networks to reduce global routing table sizes. To this end NANOG is strongly urging organizations using Class C addresses to either switch to using RFC1918 private network addresses or to renumber their Class C address into a provider aggregate. By doing this you will maintain uninterrupted global connectivity. Action to cease carrying the old Class C addresses is contemplated to begin Sept 1st 1996. If you are unsure whether you Class C addresses are part of a provider's CIDR aggregate we suggest that you contact your provider immediately. Before the explosive growth of the Internet occurred 3 years ago many people felt that Internet network addresses could be assigned permanently and stay with an organization even if it moved from one provider to another. Unfortunately the sheer size of the Internet is now outpacing the capabilities of state-of-the-art routing equipment. NANOG has for some time been encouraging new network number allocations to be made out of topologically based aggregates so that the global routing tables need only maintain a single route pointing towards a local provider. This has helped slow down routing table growth, but it is not enough. The Internet already has periodic outages caused by the size of the global routing tables and more drastic action needs to be taken. NANOG is recommending that major network providers limit the size of the routes they carry. This means that they would no longer carry routes to the smaller networks which currently take up 70% of the global routing table space. Information on how to plan for and accomplish renumbering of your network is available at http://www.isi.edu/div7/pier/ You may prefer to renumber using private network addresses http://info.internet.isi.edu:80/in-notes/rfc/files/rfc1918.txt in order to avoid renumbering in the future." Did I say anything that indicates collusion? Michael Dillon Voice: +1-604-546-8022 Memra Software Inc. Fax: +1-604-546-3049 http://www.memra.com E-mail: michael@memra.com
It really is about time that some of the larger ISP's started following the lead of folks like netaxs.com and become aggregate providers for local ISP's in their cities. This way the aggregator can be doubly and triply homed and deal with all the BGP4 nastiness. The ISP's gain the benefit of that multihoming to their city and in addition can get some of the redundancy-in-case-of-failure by buying a T1 and frame relay, or a T1 and ISDN dialup to their aggregate provider.
Not just netaxs.com... Also tlg.net in CA And cais.net in Baltimore/The DC area And new-york.net in the NY/Jersey area Of course, all of us are selling these connections, so it's not strictly that we're waking up in the morning saying "We need to be aggregate providers for all of the local ISPs to preserve global routing table space". I think the motivation is more to: a) Enhance reachability to local content/customers by slapping them on our network; b) Make it easier (as you say) for new/existing ISPs to get most of the advantages of being multiply-connected to the 'net without paying the cost or having to buy or earn the clues; c) Support the infrastructures we'd like to/need to have; and d) Yes, even make some money.
Every ISP wants to have a backup connection and right now most assume that multi-homing is the only way to achieve this.
When someone connects into us with a dedicated line we encourage them to get a 56k or T1 frame mapped into us for use strictly as a backup - and preferably from another LEC.
I believe that a middle-tier between the ISP and the NSP is the best way to achieve this and could very well decrease global routing table size.
Michael Dillon Voice: +1-604-546-8022 Memra Software Inc. Fax: +1-604-546-3049 http://www.memra.com E-mail: michael@memra.com
Avi
On Thu, 4 Apr 1996, Avi Freedman wrote:
Not just netaxs.com... Also tlg.net in CA And cais.net in Baltimore/The DC area And new-york.net in the NY/Jersey area
Add ixa.net in the Seattle, Wa area, and rain.net in the Portland, OR area. I'm sure there are others in both cities selling resellable connections, but these are the ones I know for sure. Christopher E Stefan http://www.ironhorse.com/~flatline System Administrator Home: (206) 706-0945 Ironhorse Software, Inc. Work: (206) 783-6636 flatline@ironhorse.com finger for PGP key
participants (7)
-
Avi Freedman -
Christopher E. Stefan -
Craig A. Huegen -
Dalvenjah FoxFire -
Michael Dillon -
Paul A Vixie -
Paul Ferguson