Demands routed within a single AS have a specified source node and destination node where traffic originates and terminates. Demands routed between two connecting internal ASes are specified in the same manner: with a source node in the first AS and a destination node in the second.

Cisco Crosswork Planning routes within an AS, to and from the border exit point, are determined by the IGP protocols. The selection of border exit point is modeled by the Routing policy, which is set to either Shortest Exit or Respect MEDs. This property is set in the Edit AS Relationships window, which is accessed through the Edit AS window.

• Shortest Exit—The border exit node is selected, which is closest to the source node, within the IGP of the source AS. If there is a tie, the exit node with the lowest BGP ID is used.

• Respect MEDs—The border exit node is selected, which is closest to the destination node, within the IGP of the destination AS. If there is a tie, the exit node with the lowest BGP ID is chosen.

Table 1 lists typical routing configurations that can be constructed by applying different combinations of routing policies for traffic in both directions between two ASes.

• In a peer relationship, routing in both directions is Shortest Exit, which means each controls its own border exit points.

• For a customer relationship, the customer determines the border exit points for traffic in both directions.

• For a transit relationship, the transit AS provides paid transit to the internal AS, so the internal AS determines all border exit points.

Like traffic routed within an AS, traffic routed between ASes is represented by demands. However, for demands from and/or to external ASes, the external AS is defined as the source or destination of the demand. Optionally, the specific node in the external AS from which the traffic enters or exits the internal AS is also specified.

Failover between nodes in the external AS can be modeled. For example, if the traffic is sourced from an external AS and if the peering circuit from which traffic is entering the internal AS under normal operation fails, the traffic can enter the internal AS from a different interface or peering node in the same external AS. In the Demands table, the sources and destinations are represented as follows.

AS{<ASN>}:if{node_name|interface_name}

Example: AS{33287}:if{cr01.newyork.ny|POS3/7/0/0}

For more detailed information on demand sources and destinations, see Simulate Traffic Flow from Source to Destination Using Demands.

The AS that controls the routing chooses which peering node to use. If the internal AS controls the routing, then because the topology of the internal AS is known, you can simulate the routing to the peering node. However, because Cisco Crosswork Planning has limited knowledge of the external AS topology, if the external AS controls the routing, you cannot predict how traffic will be distributed among the exit points.

The AS that controls the routing is determined by the AS type, direction of the demand, and the Routing policy property as described in Table 2.

Two ASes can be in one of four different routing relationships to one another, depending on which of the two routing policies is chosen in each direction (Table 3).

• If traffic is routed to an external AS when it has control and there is no knowledge of its topology, a set of demands is created from the source in the internal AS (or from another external AS), each with a destination set to one of the border nodes in the external AS. This way, any division of traffic between the exit points can be modeled.

• If traffic is routed to an external AS when an internal AS has control, a single demand is created from the source to the AS itself. Cisco Crosswork Planning simulations determine the correct exit point for this single demand based on the source.

• If traffic is to be routed from an external AS when it has control, a demand is created from each node in the external AS to each node in the internal AS.

• If traffic is to be routed from an external AS when an internal AS has control, a demand is created to each node in the internal AS using the external AS as the source. The demand originates from one or multiple nodes in the external AS, depending on the topology and the metric cost to reach the destination node. For example, a single demand from an external AS to a specific node could be sourced from two different nodes in the external AS, each carrying 50% of the demand traffic.

An external mesh consists of two or more external ASes with a Type property of external. An internal AS typically restricts advertisement of BGP routes for some external ASes to other external ASes. For example, destinations reachable through the transit network would not be advertised to a peer, or vice versa. In Cisco Crosswork Planning, these restrictions are represented by the absence of demands between the two external ASes.

Each AS has a property called External mesh, which Cisco Crosswork Planning uses when inserting demand meshes into a plan. Demands are created for external ASes only if one or both ASes have External mesh set to include. If both ASes are set to exclude, no demands are created for the external AS. For example, in External Mesh Control the peer and transit ASes are both set to Exclude, so no demands are created between those ASes. All other external AS demands are included in the demand mesh. Table 1 shows the External Mesh settings for common AS relationships.

The External mesh property is set in the Edit AS window.

For internal ASes, the External mesh property is ignored. More complex route advertisement policies cannot be represented by these simple External mesh settings. In this case, demand mesh creation must be performed in several steps, possibly using a script.

Cisco Crosswork Planning automatically constructs BGP pseudonodes where necessary when BGP multihops are detected.

Cisco Crosswork Planning models the nodes in external ASes that are directly connected, for example, through eBGP, to nodes in internal ASes. One exception is that you can model BGP multihops by setting the node Type property to psn (pseudonode), such as might occur at a peering exchange. This pseudonode can represent the switch that connects a number of external AS nodes to the same internal AS node. In this instance, multiple external AS nodes are connected by circuits to a BGP psn node, and this node is connected to a node in the internal AS.

Note	In all cases, eBGP multipaths across parallel border circuits is assumed.

In networks, there are two common configurations for the BGP next-hop IGP metric used in the path selection. One is to set the next-hop self on the iBGP peers (next-hop self = on). The other is to configure IGP metrics on eBGP interfaces, and to inject the interface prefix into the IGP database by setting the interface to be a passive IGP interface (next-hop self = off).

Cisco Crosswork Planning does not have an explicit next-hop self setting, so it simulates paths as if next-hop self is off. That is, the IGP metric of the egress peering interface is included in the IGP distance to the peering router and is used in the iBGP path selection. However, next-hop self to an external AS can effectively be simulated by setting the metrics on all egress interfaces to that external AS to 0. You can set the IGP metric in either the Edit Interface or Edit Circuit window.