Table of Contents

Planning Intermediate Route Selection

When a PNNI network node receives a call request, there can be multiple routes available that meet the Quality of Service (QoS) requirements for the call. This chapter describes how PNNI selects a route from multiple acceptable routes, and it describes parameters that you can modify to control route selection.

How MGX and SES Nodes Select Routes

MGX and SES nodes provide for the following PNNI route sources:

Preferred routes allow the switch administrator to define a specific route between the source and destination nodes, and then specify this route as the preferred route when defining SPVCs. If the connection is configured as a directed route, no other route is allowed, even if the route fails. If the connection is not configured as a directed route, other routes are considered when the directed route is available. When other routes are required, the switch can use the pre-calculated routing tables or dynamic routing.

The PNNI routing protocol automatically builds routing tables that list optimized routes for each destination node. The routing tables include tables for the ATM service classes listed in Table 4-1.

Withing each service class, additional routing tables list up to five entries per destination based on Administrative Weight (AW), Cell Transfer Delay (CTD), or Cell Delay Variation (CDV). The number of entries is determined by the optimum path value. If there is just one path within range of the lowest AW for CBR traffic, only one path is listed. If there are three paths within range of the lowest rtVBR CTD, all three paths are listed in the rtVBR CTD routing table. Table 4-2 shows the 10 precalculated class of service routing tables that are optimized for AW, CTD, and CDV traffic metrics.

The PNNI database entries are a collection of PNNI Topology State Elements (PTSEs), which are PNNI messages that switches use to share and collect PNNI topology data. The precalculated routing tables are calculated by analyzing the PTSEs. For many connections, the precalculated routes are acceptable. However, when a CPE requests a connection to a destination, it can specify a service class and any or all of the following:

The information in the above list is described in more detail later in this section. For the purpose of understanding how routes are selected, it is important to know that different service classes require different types of information. The information required for a service class establishes the Quality of Service (QoS) required for a call. Table 4-3 shows the types of information that are required for each service class.

When a PNNI controller chooses a route, it searches for routes that meet all of the requirements for the requested class of service. If one or more precalculated routes are working and meet all of the specified criteria, one of those routes is chosen.

When none of the precalculated routes for a particular service class and destination are acceptable, the switch performs dynamic routing. During dynamic routing, the switch searches the PNNI database for routes that match the specified criteria. As a switch administrator, you can choose what action the controller takes when it discovers the first acceptable dynamic route. You can configure the controller for first fit, which produces the fastest route selection, or you can configure the switch for best fit. When the controller is configured for best fit dynamic route selection, it identifies all routes that meet the call requirements, and then it chooses the route based on the setting of the load balancing option.

The load balancing option is a configurable parameter that allows you to choose whether the best-fit dynamic route is chosen randomly or according to the AvCR of the route. If you select the random method, the PNNI controller considers the conforming routes equal and balances the load by randomly assigning calls to each. If you choose the best-fit route based on the AvCR, the route with the highest available cell rate is chosen.

When two parallel links are available along the route, the controller chooses a link based on the configuration of the switch. The link selection options are:

The following sections provide additional information on each one of the route selection parameters listed in Table 4-3.

Tip The priority routing feature allows administrators to influence the order in which connections are routed or rerouted when network events require connection rerouting. The priority feature doesn't change how connections are routed. It controls the sequence in which connections are rerouted.

Destination Address

When an MGX or SES node receives a call request, it compares the destination ATM address with the addresses and address prefixes in the node's routing tables. The node looks for a match between the first 19 bytes of the destination address and the address prefixes in its database. The longest match determines the routes that are eligible. If there is just one route for the longest matching entry, and if that route meets the QoS requirements for the call, that is the route selected.

When multiple routes are available for the longest match, the other route selection parameters are used to determine the optimum route.

Note   Border nodes can be configured with a 0-length prefix, which matches all ATM addresses. This 0-length prefix serves as a default destination or route for all calls that cannot terminate within the peer group.

Available Cell Rate

Available cell rate (AvCR) is a dynamically generated value that indicates how much of the link bandwidth is available for the requested service class. AvCR is measured in cells per second (cps).

You cannot configure the AvCR, but you can configure a parameter called the overbooking factor, which can change how the AvCR is advertised for new calls. The overbooking factor is described in more detail later in this section.

For CBR, rtVBR, and nrtVBR, the advertised AvCR represents the bandwidth available for calls. For ABR, AvCR is the capacity available for minimum cell rate (MCR) reservation. If the AvCR meets the call requirements, the route is eligible.

After the PNNI controller calculates the AvCR for a route, it applies the overbooking factor to the AvCR before advertising the AvCR. The purpose of the overbooking factor is to allow you to purposely under book or over book a link. If link users are reserving more bandwidth than they actually need, bandwidth is being wasted. Overbooking allows you to make the wasted bandwidth available to other users. For example, if you estimate that 30% of the link bandwidth is not being used, you can configure the overbooking factor so that the advertised AvCR is 30% higher than the actual value. This enables the PNNI controller to route more calls for the link. Of course, if link users suddenly start using all link resources, some user-compliant traffic may be discarded when congestion occurs.

Bandwidth overbooking may be configured on a per-interface or per-service-class basis.

Maximum Cell Rate

The Maximum Cell Rate (maxCR) is a static value that is configured for each logical interface and can be configured separately for each service class. The maxCR represents the maximum throughput available for PNNI routes and cannot be modified by the overbooking factor.

As Table 4-3 shows, maxCR must be specified for ABR and UBR call requests, and is optional for other service class types. To block traffic for a particular service class over a link, set the maxCR for that service class to 0.

Administrative Weight

Administrative Weight (AW) is a cost factor that is assigned in both directions to every PNNI link. The cumulative AW for each direction of a network route is the sum of the administrative weights assigned to all links in that direction on that route. All routes known to the PNNI controller have a cumulative AW or cost assigned to them.

When the PNNI controller searches for routes, the chosen route must be equal to or lower than the requested AW. When best-fit route selection is enabled and multiple links are available between two nodes, the links can be configured so that the link with the lowest AW is chosen.

For SPVC connections, a Max Cost parameter can be defined to prohibit selection of a link above a specified cost, which is equivalent to the cumulative AW.

The default link AW is 5040, so if there are two links between the source and the destination, the combined AW is 1040 for the route. If you leave the AW unchanged on all network nodes, the AW operates as a hop counter.

You can change the AW on links to control network traffic. For example, you can reduce traffic on a backup link by increasing the AW to more than that on the preferred route. If the preferred route fails, the backup link becomes the lowest cost route and becomes available.

Cell Transfer Delay

Cell Transfer Delay (CTD) is the measure of the delay an ATM cell encounters as it passes through switch interfaces. CBR, rtVBR, and nrtVBR traffic may request a maximum acceptable CTD for a route. The route CTD is the sum of the CTD values for all interfaces through which the route passes and represents the time interval between a cell exiting the source node and entering the destination node.

The CTD used in MGX and SES nodes is a static value that is set by Cisco and is based on the speed of the interface. Faster interfaces will have lower CTD values, and slower interfaces will have higher values. When the PNNI controller is selecting a route, any route with a CTD equal to or lower than the value requested for the call will be eligible.

Note   The PNNI controller does not choose routes based on the lowest CTD. The route CTD is used to determine which routes are eligible for selection; the route is chosen from eligible routes based on the lowest AW.

Cell Delay Variation

Cell Delay Variation (CDV) is a measurement of the variation in CTD over links and through nodes. CBR and rtVBR calls can request a maximum acceptable CDV for a route.

The CDV used in MGX and SES nodes is a static value that is set by Cisco and is based on the type of interface and node. When the PNNI controller is selecting a route, any route with a CDV equal to or lower than the value requested for the call will be eligible.

Note   The PNNI controller does not choose routes based on the lowest CDV. The route CDV is used to determine which routes are eligible for selection; the route is chosen from eligible routes based on the lowest AW.

CLR0

The Cell Loss Ratio (CLR) of a route is the ratio of the number of lost cells to the total number of cells transmitted by the source endpoint. CLR0 is the cell loss ratio for cells with the Cell Loss Priority bit set to 0. Table 4-3 shows that CBR, rtVBR, and nrtVBR calls request specific CLR0 levels.

For each link along a route, static values are assigned to CLR0 for each class of service. All routes with a CLR0 value that is less than the requested value are eligible for selection.

CLR0+1

CLR0+1 is the cell loss ratio for all cells with the Cell Loss Priority bit set to either 0 or 1. Table 4-3 shows that CBR, rtVBR, and nrtVBR calls request specific CLR0+1 levels.

For each link along a route, static values are assigned to CLR0+1 for each class of service. All routes with a CLR0+1 value that is less than the requested value are eligible for selection.

Configurable Parameters for Route and Link Selection

There are many parameters that influence route and link selection. Some of these parameters are configurable and some are not. The following list summarizes the configurable parameters and features:

Most of the parameters and features listed above are described in the previous section. The following sections describe the nodal parameters that can be configured.

Nodal Transit Restriction

As a switch administrator, you can configure MGX and SES nodes to support or deny connections that pass through the node. If you chose to deny transit or pass-through connections, the node will only accept calls that terminate on one of the node's interfaces. Other nodes will not be able to establish routes through the blocked node to other nodes.

Nodal Point-to-Multipoint Branch Restriction

MGX and SES nodes support point-to-multipoint connections by default. The software allows you to disable this feature if you want to block point-to-multipoint calls through a node.