Max. no. of BGP prefixes supported in ASR9K Platform

[3alee]

Hello

If the requirement is to support 6 hundred thousand IPv4 BGP routing table + 1 Million IPv6 BGP routing table, can the

A9K-RSP-8G support it? Or do I need A9K-RSP440-TR (6GB) or even 440-SE?

Thanks.

[Alexei Kiritchenko]

FIB is shared between IPv4 and IPv6 and one IPv6 route occupies twice more size compare to IPv4.

For trident LC the limit is 1.3M and for Typhoon 4M (note, the limit is per LC and not per RSP)

Trident:                IPv4 + 2*IPv6 <=1.3M

Typhoon:             IPv4 + 2*IPv6 <=4M

For RPS memory it is challenging to say as it is linked to the amount of memory the paths are going to consume.

Each process in XR can allocate up to 4G of memory, in the worst case scenario ipv6 RIB and ipv4 RIB may want to allocate 4G of mem, then the TR is not good enough (6G) and you will want the SE version (12G).

Regards,

/A

[3alee]

Thanks a lot!

But can the A9K-RSP-8G support that?

Any supporting doc for reference.

Thanks again.

[Alexei Kiritchenko]

Yes, A9K-RSP-8G can support it. but i'd like to emphasize that for your requirement we need to have a new generation Typhoon Line Cards.

(and if you invest into a new LC generation, that would be better to go for RSP440, what would enable your more new functionality compare to RSP2).

Regards,

/A

[3alee]

ok, but if for RSP440, I should choose SE version (12G) in this case, right?

[Alexei Kiritchenko]

Most likely 6G would be OK for your amount of prefixes. As I’ve mentioned, pure number of prefixes is not enough to confirm it. All would depend on your real system configuration and mem usage.

Again, if you can afford 12G, I’d go for it as there is no on fly memory option upgrade.

Regards,

/A

[Bradley Gould]

For others reading this, Trident LC's can scale to that number of routes, but you need to configure it first.

https://supportforums.cisco.com/docs/DOC-25489

[YUEN WAI NAM]

IOS XR ( RSP440-TR) seem default max BGP allowed prefix of a neighbor is 524288, how to change this default setting in stead per neighbour set the max allowed prefixes.

Prefix advertised 337, suppressed 1, withdrawn 34

  Maximum prefixes allowed 524288

  Threshold for warning message 75%, restart interval 0 min

Moreover, IOS has a command to check the default max routes TCAM on IPv4, IPv6 route table size, how about IOS XR

7600# sh mls cef maximum-routes

FIB TCAM maximum routes :

=======================

Current :-

-------

IPv4 + MPLS         - 512k (default)

IPv6 + IP Multicast - 256k (default)

[Alexei Kiritchenko]

we have a number of bgp defaults for protection, one of these defaults is the number of BGP prefixes we are allowed to receive from a peer.

these are only threshold default configs across all XR platforms and DON'T indicate the maximum capacity of a per platform maximum.

                PRE 4.2.1   4.2.1

IPv4 Unicast:     524288 -> 1048576

IPv6 Unicast:     131072 -> 524288

VPNv4 Unicast:    524288 -> 2097152

VPNv6 Unicast:    524288 -> 1048576

http://www.cisco.com/en/US/docs/routers/crs/software/crs_r4.2/routing/command/reference/b_routing_cr42crs_chapter_01.html#wp3192417938

Please note! As it was explained earlier, the number of routes limitation comes from LCs, not RSP!

You can use "show cef platform resource location " to find out max and currently used number of entries.

A memory usage on RSP is not linked to a number of prefixes, but to amount of attributes linked to those prefixes and number of features you run on the router. Amount of prefixes is just one of many things which contributes to the RSP mem usage.