Sorry for being vauge.

We are an ISP and we have a very big project coming on test in the coming months. We will be upgrading all our IXP links to 1GB to ensure a 1GB link through to the project network. As i said - we have two paths to different IXPs in the UK and we are peering with about5 ISPs in each. We would ideally like all traffic destined for the project network (currently a /29 public range) to come in over one IXP link in paticular.

If that fails then we would like all traffic to come in over the other.

Also, it is the case that all inbound traffic is expected to be quite light HTTP requests but the outbound traffic is going to be images. We would ideally like to send the traffic back out over the second link. to somewhat load balance the traffic to and from the project network.

I understand that other ISPs tend to reject prefixes under /23 at the moment (understandably)

I am wondering if there is a way to have this network advertised out in a way that we don't have to up it to a /23 or even a /24. While not comprimising the Internet routing table with tiny routes.

Will be reading my BGP books later :)

Cheers

Stephen

As far as I know the limit is /24 for Internet routing(not /23).

Refer to the following to have an overview.

http://www.cidr-report.org/

To my knowledge below /24 are not allowed.

Hi,

I think you can achive the inbound route selection by simply doing as path prepending on your outgoing updates to your bgp peers.

So say you BGP AS is 2000, then you would advertise your routes to all BGP peers, except the preferred incoming BGP peer path, with a BGP path of 2000 2000, instead of the normal 2000, which is what is advertised to the desired incoming BGP peer.

With regard to outgoing then you will really need to be able to identify where the source traffic is coming from, BGP AS, specific subnets etc, then you could use weight or local preference to engineer the outgoing path from your AS.

Heres a link to BGP path selection @ Cisco.

http://www.cisco.com/en/US/tech/tk365/technologies_tech_note09186a0080094431.shtml

Basically, for incoming to your AS you manipulatte MED or AS PATH length, for outgoing from your AS you use either WEIGHT or LOCAL PREF.

HTH

LR

Hi Guys,

thanks for info. Read somewhere that it was a /23, anyway i should ask peers.

So, by the rules of BGP, by prepending the our AS-Path to all peers except desired incoming path, BGP will see longer incoming as path and use shortest path first. Using other path as backup. This is simple. I think we will have to use the /24 to advertise the network tho. Thats ok tho... i'm a bit tight with handing out IPs willy nilly. :)

Thanks guys

Stephen

Hi,

Please need to know what cisco router model is prefered if an ISP is to peer with 10 other ISP in an internet exchange point. I've choosing an ASR1001 and cisco ISR 3945, but it seems like it does not implement BGP multi sessions, neither QoS...

please need a router for an IXP we are building, with this minimum criteria:

- 3 x 10/100/1000Mbit ethernet interfaces

-some SFP or FO interfaces

-slot to add interfaces

thks