Cisco BPX 8650, Руководство по установке и настройке

Страница: 72 / 606

1-22
Cisco BPX 8600 Series Installation and Configuration
Release 9.3.30, Part Number 78-12907-01 Rev. E0, May 2005
Chapter 1 The BPX Switch: Functional Overview
Traffic and Congestion Management
Cost-Based Routing Management
You can selectively enable cost-based route selection as the route selection per node. With cost-based
routing management, a trunk cost is assigned to each trunk (physical and virtual) in the network. The
routing algorithm then chooses the lowest-cost route to the destination node. The lowest cost routes are
stored in a cache to reduce the computation time for on-demand routing.
Cost-based routing can be enabled or disabled at anytime. There can be a mixture of cost-based and
hop-based nodes in a network.
For more detailed information about cost-based Automatic Routing Management, see the Cost-Based
Connection Routing section.
Priority Bumping
Priority Bumping (PB) allows BPX and IGX switch connections classified as more important (through
the CoS value) to “bump” (that is, set aside) existing connections of lesser importance. While the
Automatic Routing Management feature is capable of automatically redirecting all failed connections
onto other paths, priority bumping lets you prioritize and sustain more important connections when
network resources are diminished to a point that all connections cannot be sustained. Network resources
are reclaimed for the more important connections by bumping (derouting) the traffic on less important
connections.
Priority bumping is triggered by insufficient resources (such as bandwidth), resulting from any number
events, including changes to the network made by using the commands
addcon , upcon , cnfcon , cnnfcos ,
cnfpref , cnftrk , and deltrk . Other triggers include trunk line/card failure, node failure, and
communication failure. The most prominent event is a trunk failure.
For information on setting up Priority Bumping, refer to the Cisco WAN Switching Command Reference,
Release 9.3.30 .
Concurrent Routing
CR is an enhancement to the Automatic Routing Management feature and does not alter the Automatic
Routing Management messaging protocol. The Automatic Routing Management functionality is
operational whether or not CR is enabled on a node. If CR is disabled, the node exhibits preswitch
software Release 9.3.30 behavior, which includes collisions and back off. When CR is enabled,
collisions occur less frequently.
Concurrent Routing (CR) allows multiple routing requests to be processed simultaneously on a node.
For example, a node can initiate (master node) one or more routes while simultaneously accepting other
routes that pass through it (via node) or terminate at it (slave node).
If CR is not enabled on a node, routing requests received while a connection is in the process of being
routed, is routed sequentially. As a result, only one bundle at a time can be routed on a node. This
sequential routing algorithm under utilizes the computational power of the switch. Sequential routing is
illustrated in Figure 1-9.
CR allows the processor of the switch to be more effectively utilized by allowing multiple routes to be
in progress concurrently. The result is better overall reroute performance. Performance improvement is
not realized for individual or topologically disjoint reroutes. The key performance metric that is
improved by CR is network settling time. Network settling time is defined by the longest settling time
for any single node, assuming all of the nodes start routing at the same time. The number of nodes and
connections in the network, network topology, and other configureable routing parameters all affect
network settling time. CR is illustrated in Figure 1-10.