RA Project Update

Elise Gerich

Randy Bush was kind enough to provide me a copy of his notes. I have formatted them for html, but did not change any of the content in anyway.

Table of Contents

• Randy Bush's Notes
• Stan Barber's Notes
• September 1995 NANOG Home Page

Randy Bush's Notes

• RA project is collab between ISI, MERIT, Cisco, and U Mich
• What it was supposed to be
  • Routing Arbiter Database
  • Route Servers
  • Tool Development
• RADB Service Description
  Based on RIPE-81 now RFC 17
  • provide mechanisms to register routing informaton
    • note that RADB now has PGP authentication for updates
  • encourage registrants to maintain integrity of information
    • now have more tools, i.e. rrc2r
    • consistency checking tools
  • provide mechanisms to access then information
    • whois server
  • provide information for the configuration of the route servers
    • route servers serve up RADB+RIPE+MCI+CANET
• Currently 42182 routes with 1724 withdrawn
  • dupes with CANET have been resolved, but some remain with MCI and RIPE
  • it is recommended to select one registry and register only there
  • Kent England asks if /16s and /24s could be further aggregated. Brian says yes, many could be
  • Andrew Partan points out that there are only about 30.000 routes on the internet, so the numbers above are off in some way
• Route Servers
  • project is to put four at the four NSF NAPs
    • in DC, there are two servers peering with 16 providers, about half of the organizations at MAE-East. it is not considered not to be fully operational because the out of band dialup modem line is bad
    • in SF at PacBell, two servers and six peers of 22 possible. not fully operational because the out-of-band access via modem is not yet working.
    • Chicago one of two servers installed, the next coming in a week. Two peers out of a possibly 13.
    • Sprint Pensauken NAP peering with four of ?, fully operational as of July. None of the peers are Sprint.
    • MAE-West has one route server, but there is a financial support glitch. AlterNet peers with it.
    Check this URL for more information: http://www.ra.net/routing.arbiter/RA/RADB.tools.docs/radbserv.html
• Next Steps
  • All peers are already inolved in bi-laterals, thereby obviating the route servers. Volunteers sought to actually rely on the route servers. Sean Doran said he would if Elise would take responsibility for all problems. Elise said she would if Sean would take responsibility for Sprint outages
  • Complete population of maintainer objects
  • Get all servers fully operational
• Sean Doran
  • When will the AS690 Advisories go away
  • Peter's tool dies on AS690 advisories
  • Will other advisory lines be allowed
• Discussion
  • Advisory lines may go away in the next 60-90 days.
  • Curtis notes that the syntax checking with the advisories are a problem. Curtis is trying to get it all stable before doing the cutover. Diffs are still not exactly matching.
  • Tony Bates notes that advisory lines are for local information, and the current use is merely an example. All that is defind is the ASN and the rest is free form advice to that AS
  • Andrew Partan wants to know how accurate is working on a dynamic tool which could shove RADB updates as fast as the net flaps. Does the RA project wish updates at that rate.
  • Randy Bush asks if the RADB is proscriptive or descriptive?
  • Sean wants it to be used for debugging. But does not believe it should be completely proscriptive. Wants to know who should be announcing the route. Sean says he wants a change log of routes so he can see how things were when they might have been correct.

Stan Barber's Notes

Just over 42,000 route-object pairs are registered.

Router Servers are installed at various interconnection points. At MAE-EAST, there are two RSes that are peering with a number of providers at MAE-EAST. They are peering with 16 folks they are peering which is slightly more than 1/2 of the organizations connected at that point.

At PAC-BELL, there are 2 route servers which peer with six organizations. There are 22 total organizations interconnected there.

At Chicago, there is an RS installed with the second one coming this Friday. There are 2 peers up and 13 organizations connected.

At Sprint, there are four organizations peering. This one is fully operational.

At MAE-WEST, there is one installed, but it is not operational at this time. It has one peer with Alternet. Once the agreements are finalized, this server will be made operational.

Everyone that is involved has a bilateral arrangement with whomever they peer with.

Now, the RS-team wants find volunteers to start using RS as primary information source and their current direct peer as a secondary.

They want to get the RAs maintainer objects all population of the maintainer objects.

Finally, they want to declare the RSes operational.

Sean asks when the AS690 advisory lines will go away. Curtis says that work is moving forward on this and work is going well. However, it is not ready for prime time just yet. As soon as these bugs are worked out, then the advisory lines can go away. Tony said that that advisory lines are just that "advisory" and specific for a particular AS. Elise said that advisory lines were created to permit local information specific to an AS.

Andrew said that he is working on a tool that speaks BGP on one side and objects out the otherside. Elise and Randy ask if the RADB should be proscriptive or descriptive. Sean does find the data in the RADB to be useful in debugging routing problems. [It appears that there are two schools of thought....should the RADB describe a policy or just document how it is working at the time the query is made.] Sean argues that a routing revision tool would be useful. Sean believes that the prescriptive database is not as useful for this kind of work. Matt notes that PSC has done things like this (comparing routing changes to routing policy).