RE: Multihoming - Resilience or Independence
The point is we will aggregate other ISP's prefixes if they are CIDR able. Most of the case are there are holes in the range so CIDR is out-of-question.
While the other camp insist to filter out any routes more specific than /20. They are doing nothing to re-allocate their PA assigned to multi-homed customers so CW can CIDR them into /20.
The funny thing is the solution are controlled by the other camp but they are all talk but no action. (Or trying to tell you multi-homed is a bad idea because they can't deal with the routing table ).
I am not sure I am following you here. Are you saying that we should re-adress customers so that we get larger blocks of PA space to give to mulithomed customers?
I think that re-addressing is pretty hard to do in any larger scale, although that would certanly help some other issues.
That's why filtering /20 only without re-aligning the holes in the block won't work. In the end /20 make the routing table looks pretty but they are black-holing lots of traffic. As we all agree that re-assign address is NOT practical at all thus just filtering /20 block is not a good solution.
another good start would be to start reclaiming space that has been handed out from various ISPs old B-space and later taken with the customer. I know a few operators that have done that (including us to some extent), and it's a rather demanding task but will help the operators and reudce the routing table as well.
- kurtis -
I am with you here. Starting from the beginning is much better than asking them to change later. Ping Lu Cable & Wireless USA Network Tools and Analysis Group W: +1-703-292-2359 E: plu@cw.net
That's why filtering /20 only without re-aligning the holes in the block won't work. In the end /20 make the routing table looks pretty but they are black-holing lots of traffic.
Well, it won't. The proposal for filtering is to filter on assigment boundary, and for swamp space filter on /24. Now, what will break is people punching holes in their PA space for multihomed users. Funny enough, this has never been discussed (to my knowledge) as a problem in PTOMAIN WG of the IETF. There the argument was for traffic engineering and commercial relationships...
As we all agree that re-assign address is NOT practical at all thus just filtering /20 block is not a good solution.
That is not the suggestion. See above. The assignments per block and RIR is well known. Best regards, - kurtis -
participants (2)
-
Kurt Erik Lindqvist KPNQwest
-
Lu, Ping