9.5. P-NNI

P-NNI routing and signaling protocols are used to discover an ATM network topology, create a database of topology information, and route ATM virtual circuits over the discovered topology [ATMF02]. Under the P-NNI model, the network is viewed as a multilevel hierarchy. Each level in the hierarchy consists of one or more entities known as the peer group. The constituents of the peer group vary depending on the level of the hierarchy. At the lowest level of the hierarchy, the peer groups consist of network elements, namely, the switches. As we move up in the hierarchy, the peer groups consist of abstract nodes representing peer groups at the lower layers. The structure of the P-NNI hierarchy, P-NNI peer groups, peer group leader, and border nodes were described in Chapter 5. P-NNI signaling was described in Chapter 7. In the following, we look at P-NNI routing briefly.

9.5.1. P-NNI Routing

P-NNI routing follows the link state paradigm, and it is hierarchical. Neighboring nodes form a peer group by exchanging their peer group identifiers (PGIDs) in Hello packets using a protocol that makes the nodes known to each other. If the nodes have the same PGID, they belong to the peer group defined by that particular ID. If their PGIDs are different, they belong to different peer groups. A border node has at least one link that crosses the peer group boundary. Hello protocol exchanges occur over logical links.

P-NNI defines the creation and distribution of a topology database that describes the elements of the routing domain as seen by a node. This database provides all the information required to compute a route from the node to any address that is reachable in or through that routing domain. Nodes exchange database information using P-NNI Topology State Elements (PTSE). PTSEs contain topology characteristics derived from link or node state parameter information. The state parameter information could be either metrics or attributes. PTSEs are grouped to form the P-NNI Topology State Packet (PTSP). PTSPs are flooded throughout the peer group so that all the nodes in a peer group can converge to identical databases.

As described in Chapter 5, every peer group has a peer group leader (PGL). There is at most one active PGL per peer group. The PGL will represent the peer group in the parent peer group as a single node. The PGL also floods the PTSEs received in the parent peer group into its own peer group. Apart from its specific role in aggregation and distribution of information for maintaining the P-NNI hierarchy, the PGL does not have any special role in the peer group.

P-NNI routing works in conjunction with P-NNI signaling during connection establishment. Specifically, the crank-back feature in the signaling protocol allows inaccuracies in summarized routing information to be compensated for during connection establishment. Connection establishment under P-NNI, which was described in Chapter 7, consists of two operations: the selection of a path and the setup of the connection state at each node along that path. If the computed path is not correct (due to inconsistencies in the summarized view and the actual state), the connection set-up attempt may be blocked en route. The crank-back feature allows the path to be partially or fully recomputed by nodes en route.

In summary, P-NNI routing is highly scalable due to the large number of hierarchical levels allowed. Furthermore, it incorporates multiple link and node metrics, and mechanisms to summarize peer group topology and attributes. Some of these concepts have been used to define interdomain routing in optical networks, as described in Chapter 12.