On 26.02.2009, at 14:22, Claudio Jeker wrote:
On Thu, Feb 26, 2009 at 12:07:12PM +0100, Andreas Fink wrote:
On 26.02.2009, at 11:27, Stanislav Sinyagin wrote:
...
For what its worth my router tells me this:
IPv6 routes: 1'577 entries, 1'194 AS numbers IPv4 routes: 274'504 Ientries, 30'488 AS numbers
If the wold would be all IPv6, our routers would need 10 times less memory ;-)
Do you think that if IPv6 is used worldwide the number of ASs would be smaller then today? The same question is also true for the number of networks.
This wont change much I guess. The "political" boundaries which exist in autonomous systems will stay the same. Maybe some multihomed customers would instead become dual homed and would not need an AS number.
This is the biggest lie of IPv6. The routing table will not get smaller and will gain the same exponential growth that IPv4 has now.
This is not true. ISP's today own dozens of subnets. The average AS number announces 8 routes. With IPv6 most of those who do announce dozens of nets would only announce one prefix.
If the world would be all IPv6 you would need at least 4 times as much memory on your routes. Most probably you need to replace most because their CAMs are to small or the ASICs do not support IPv6.
The current routers out there have no problem with IPv6. If they would have such issues, they have even bigger issues with IPv4 as it is currently. IPv6 has been designed from the ground up to allow easy routing in hardware. One of the reasons why IPv6 headers are always on 4 byte boundaries. And by the time we have all IPv6 ranges, routers with 4 times as much memory are widely available. My 10 year old Cisco7206VXR can still easily cope with a full routing table.
About the size reduction / increase, here's an example:
Below you see the BGP4 routing table of Swisscom (AS3303) for all routes ending with AS3303 (which excludes multihomed clients): This routing table has 147 entries. In IPv6 it would have ONE entry because that ONE subnet gives by far enough space. Also if a subnet would been taken over from another ISP (merger etc), renumbering is changing prefixes on routers but not changing all hosts in a subnet. So way easier. Also the prefix can be replaced 1:1. This is not possible in IPv4 and renumbering is a major hassle in that case (which is why no one does it) because the size constraints required to use optimal size allocation and over time what's optimal changes. IPv6 does not have that burden.
This example shows a 147 : 1 tradeoff in routing table entries. Assuming a table entry takes 4 times as much space (which I dont think because an entry holds more than just the IP... so it will be 24 bytes longer, not 4 times as big). you are still saving a factor of 1:36.
See for your self
*> 77.72.128.0/21 *> 78.110.128.0/20 *> 91.199.186.0/24 *> 91.208.130.0/24 *> 134.146.200.0/23 *> 138.187.128.0/18 *> 138.188.0.0 *> 138.190.0.0 *> 145.234.0.0 *> 145.250.128.0/17 *> 146.109.0.0 *> 146.159.0.0 *> 156.25.248.0/21 *> 156.106.0.0 *> 161.78.0.0 *> 163.168.0.0 *> 164.128.0.0 *> 192.53.104.0 *> 192.83.223.0 *> 192.102.95.0 *> 193.5.0.0 *> 193.5.3.0 *> 193.5.4.0/23 *> 193.5.38.0 *> 193.5.59.0 *> 193.5.61.0 *> 193.5.67.0 *> 193.5.224.0/20 *> 193.8.145.0 *> 193.8.167.0 *> 193.8.196.0 *> 193.8.198.0/23 *> 193.16.241.0 *> 193.47.232.0 *> 193.72.79.0 *> 193.73.106.0/23 *> 193.73.208.0 *> 193.134.32.0/22 *> 193.134.36.0/22 *> 193.134.131.0 *> 193.134.206.0 *> 193.134.210.0 *> 193.134.214.0 *> 193.134.248.0 *> 193.135.0.0/23 *> 193.135.46.0 *> 193.135.108.0/23 *> 193.135.128.0/22 *> 193.135.132.0 *> 193.135.143.0 *> 193.135.144.0/23 *> 193.135.156.0 *> 193.135.173.0 *> 193.135.214.0/23 *> 193.135.216.0/23 *> 193.135.218.0 *> 193.135.219.0 *> 193.135.255.0 *> 193.201.122.0/23 *> 193.222.64.0/19 *> 193.223.68.0 *> 193.223.112.0/20 *> 193.223.224.0/20 *> 193.246.0.0/23 *> 193.246.16.0/21 *> 193.246.48.0/23 *> 193.246.50.0 *> 193.246.56.0 *> 193.246.57.0 *> 193.246.62.0/23 *> 193.246.99.0 *> 193.246.100.0 *> 193.246.104.0 *> 193.246.113.0 *> 193.246.122.0 *> 193.246.127.0 *> 193.246.205.0 *> 193.246.246.0 *> 193.246.248.0/22 *> 193.246.252.0 *> 193.246.254.0 *> 193.247.36.0/22 *> 193.247.40.0 *> 193.247.44.0/22 *> 193.247.48.0/20 *> 193.247.86.0 *> 193.247.128.0/22 *> 193.247.132.0 *> 193.247.151.0 *> 193.247.154.0 *> 193.247.217.0 *> 193.247.224.0/21 *> 193.247.244.0/23 *> 193.247.247.0 *> 193.247.250.0 *> 194.6.160.0/19 *> 194.11.128.0/23 *> 194.11.144.0/21 *> 194.11.166.0/23 *> 194.11.223.0 *> 194.35.252.0 *> 194.40.244.0 *> 194.56.0.0 *> 194.56.3.0 *> 194.56.4.0 *> 194.56.127.0 *> 194.56.234.0 *> 194.93.112.0/22 *> 194.124.209.0 *> 194.124.232.0 *> 194.124.233.0 *> 194.124.242.0/23 *> 194.147.52.0/22 *> 194.147.96.0 *> 194.147.134.0/23 *> 194.169.219.0 *> 194.191.65.0 *> 194.209.0.0/16 *> 194.209.86.0/23 *> 195.8.108.0 *> 195.35.121.0 *> 195.47.231.0 *> 195.47.245.0 *> 195.65.0.0/16 *> 195.144.32.0/19 *> 195.158.230.0/23 *> 195.176.128.0/19 *> 195.176.192.0/19 *> 195.225.60.0/23 *> 195.234.37.0 *> 195.245.228.0 *> 195.248.91.0 *> 212.42.128.0/19 *> 212.90.214.0 *> 212.117.96.0/19 *> 212.243.0.0/16 *> 217.192.0.0/15
PS: above list could be optimized a bit by using proper aggregation / summarisation
-- :wq Claudio
swinog mailing list swinog@lists.swinog.ch http://lists.swinog.ch/cgi-bin/mailman/listinfo/swinog