APNIC Member Meeting Seoul, Korea, 3 March 2000

APNIC Routing SIG

Discussion Paper: IRR Scalability

Circumstances:

IRR, Internet Routing Registry, is a database which accumulates routing information in the Internet. And some ISPs have been doing route filtering using these databases. These routing information databases in the Internet are very important for operation. It is also useful that we check a correctness of routing information in shooting troubles.

So far, these databases have been managed by three organizations, Merit RADB, RIPE/NCC and MCI. Merit RADB has most number of objects in the database. Merit RADB started to charge last November.

And they began to suggest that each ISP should start to manage IRR database themselves to use IRRd which is available distributed management.

The way of distributed management makes a database to set at many points of the Internet. These databases has already reached 26 databases (on 16 Feb. 2000, the present).

Searching necessary information from all these distributed data is getting more difficult day by day. However, these distributed databases are increasing day by day.

Proposal:

The dispersion of IRR relates to the dispersion of the administration information for routing. However, the administration information for routing should be able to search from ONE database. And its result must be correct from a viewpoint of the Internet operation. That's why the dispersion of the administration information for routing makes getting difficult to satisfy this requirement.

If we satisfy this requirement from a viewpoint of operation, IRR users should be going to satisfy two requirements. One of the requirement is that we can get a result by searching from just one database system which is gathered data with data of another database. Another requirement is that result is correct.

I would like to propose that we make that kind of database which satisfies those requirements. We have a good point that in each region it has RIR which manage the administration information for routing for each region. These IR manages a database which has been registered the administration information for routing that was assigned/allocated by IR. It means that RIR have correct administration information for routing.

And we expect to exchange the data of IRR database smoothly, because every RIR has relation to another RIRs. It means that the best organization are every RIRs that can satisfy two requirements I mentioned before. I propose this.

Expecting effect:

The administration information and priority information for routing are managed at many points of the Internet. In this situation, it has a possibility to be left out a searching object when we have to search. If this trend is going to be accelerated, it has a possibility we can't search operation information we need.

On the other hand, IRR have been managing a priority information for routing on IX. Obviously these information is out of range of managing for RIR.

However, IRR manages priority information and administration information separately. So, RIR can manage these separated two information.

And, it can be expected the effect which makes a limitation of a tendency of dispersing by starting IRR. It makes advantage that we have to search for few databases. And also reliability of information is going to become higher, because we will search for data managed by IR.

In addition, if we can get the priority information together with the administration information, then we can get an advantage from a viewpoint of operation.

Top  |  SIGs