PI vs PA in routing table
Hello! Is there a graphs shows how much prefixes in global route table are PA and how much - PI (maybe in percents)? And also this one year-by-year (monthly?)? Maybe there is no real "route table PI pollution" trouble in the real world? -- WBR, Max Tulyev (MT6561-RIPE, 2:463/253@FIDO)
Hi, On Tue, Sep 19, 2006 at 03:41:06PM +0000, Max Tulyev wrote:
Maybe there is no real "route table PI pollution" trouble in the real world?
Currently, there are much more PA entries than PI - but the problem that I see is that the PI growth rate is much higher than PA, and as such, "unlimited PI" might rise to be a problem. Nobody is proposing to disallow PI. Just balance "cost" vs. "convenience" a little bit better. Gert Doering -- NetMaster -- Total number of prefixes smaller than registry allocations: 94488 SpaceNet AG Mail: netmaster@Space.Net Joseph-Dollinger-Bogen 14 Tel : +49-89-32356-0 D- 80807 Muenchen Fax : +49-89-32356-234
On 19-sep-2006, at 17:41, Max Tulyev wrote:
Is there a graphs shows how much prefixes in global route table are PA and how much - PI (maybe in percents)?
And also this one year-by-year (monthly?)?
Maybe there is no real "route table PI pollution" trouble in the real world?
Ok a bit late but maybe this is useful: - there is no easy way to distinguish between a PI and a PA prefix - increase in RIR delegations last year was about 6%, increase in advertised prefixes around 16% - there are less than 20000 AS numbers and some 180000 prefixes, so at least 160000 of these prefixes aren't the result of multihoming as such (although multihomers can also advertise more than one prefix for traffic engineering reasons or no reason at all of course) (If multihoming is going to be a problem then look at AS numbers, the ASes used by ISPs won't be significant in that case)
Iljitsch van Beijnum wrote: [...]
- there is no easy way to distinguish between a PI and a PA prefix
I presume the RIPE NCC would assign those blocks from a "separate" address range, much like it is the habit for IPv4 PI. And of coure include that informtion in the *usual* reports and places. Maybe this should go into the proposal? Wilfried.
Iljitsch van Beijnum wrote:
Ok a bit late but maybe this is useful:
- there is no easy way to distinguish between a PI and a PA prefix
RIPE (and other RIRs) DB?
- increase in RIR delegations last year was about 6%, increase in advertised prefixes around 16%
- there are less than 20000 AS numbers and some 180000 prefixes, so at least 160000 of these prefixes aren't the result of multihoming as such (although multihomers can also advertise more than one prefix for traffic engineering reasons or no reason at all of course)
(If multihoming is going to be a problem then look at AS numbers, the ASes used by ISPs won't be significant in that case)
This is not PI specific, as I can see. There is a lot of PA holders (LIRs) having just one uplink channel. -- WBR, Max Tulyev (MT6561-RIPE, 2:463/253@FIDO)
participants (4)
-
Gert Doering -
Iljitsch van Beijnum -
Max Tulyev -
Wilfried Woeber, UniVie/ACOnet