Other Implementations

This section discusses two Cisco Systems, Inc. proprietary network protocols and control-plane implementations. The following sections cover a monolithic version and a portable version of AutoRoute.

AutoRoute

AutoRoute (or Automatic Routing) is a very optimized and robust Cisco Systems, Inc. proprietary networking protocol that runs on BPX-8600 and IGX-8400 switches. AutoRoute (AR) makes use of a distributed topology database called the Network Information Base (NIB) that is used by the distributed Connection Management (CM) protocol. Processor cards in the nodes (BCC [Broadband Controller Card] in BPX-8600 and NPM [Network Processor Module] in IGX-8400) communicate with one another using network messages that need to be acknowledged by the receiving node. Each node maintains its own representation of the network topology and is aware of the current network state (including alarm conditions and the amount of traffic routed over a trunk, also called trunk-based loading). AutoRoute allocates bandwidth and routes. It reroutes connections over optimal paths through the network following either a hop-based route or a cost-based route. Route selection is source-based and is based on Dijkstra's Shortest Path Algorithm and the distributed database.

AutoRoute offers a wide range of added features, such as automatic connection management, automatic alternate routing, optional route cache, the ability to specify preferred paths or to avoid certain trunks for some connections, priority bumping, network clock distribution, time of day and date distribution, and in-band management. With its distributed intelligence, AR eliminates central points of failure. A connection's master endpoint is responsible for rerouting it, not a centralized route processor.

As far as scalability, the AR limit is 223 routing nodes. AR has proven itself in large networks. Each routing node can support up to 16 feeder nodes (which do not participate in the distributed topology), expanding the network node count to more than 1000. However, the long-term solution to "infinite scalability" in routing nodes is to migrate to PNNI.

NOTE

PNNI nodes also support feeder nodes that are not topology-aware.

AutoRoute is included in the software image that BCC and NPM processor cards run, called Switch Software (SwSW). The first implication is that ATM cards (BXM and UXM) always have an AutoRoute partition, controlled by SwSW and not VSI. Apart from the AR partition, using SwSW you can configure partitions to be controlled by the VSI slave in those cards.

NOTE

One of the differences between the AR partition and VSI partition management is that VSI connection setups are always acknowledged by the VSI slave, whereas normally AR connections are not. Switch Software sends the connection setup command to the line or trunk card and assumes it is successful. Switch Software checks Connection Admission Control (CAC). (In AutoRoute, Switch Software can set overlap checking flag and ACK required flag in a connection setup, specifically for VPI = 0.)

Portable AutoRoute

Portable AutoRoute (PAR), as the acronym clearly indicates, separates AutoRoute from the platform software. It is implemented in PXM-1-based MGX-8230 and MGX-8250. It uses VSI as the interface between the networking software (PAR) and the platform software.

The PXM-1 card houses the centralized VSI slave for all service modules in the MGX-8230 or MGX-8250, as well as the PAR application and VSI master. The VSI slave also manages logical ports in the PXM-1 for external clock interfaces and IP connectivity. All connections and resource management go through the centralized slave.

In a PXM-1-based MGX-8250 implementation, PAR has four functions:

• Connection provisioning Performs feeder connections and local connections (DAX connections). The current implementation does not support AutoRoute routing functionality. It also maintains database synchronization.

• Connection alarm management PAR sends messages to condition connections when it receives a failure event.

• Annex.G In a feeder configuration, PAR manages Annex.G LMI communication between the MGX-8230 or MGX-8250 and the BPX-8600.

• Clocking PAR handles the clocking selection in the edge switch.

An MGX-8230 or MGX-8250 node can be connected as a feeder node to a BPX-8600 or a PXM-45-based MGX-8850 or can act as a standalone node. A feeder node does not share topology information. In both cases, it supports single-height service modules such as FRSM, AUSM, CESM, and VISM and also double-height modules such as RPM/B and RPM-PR.

PXM-45-based MGX-8850 also implements PAR as an internal controller in the PXM-45 card.