Information About BGP Inbound Optimization Using Performance Routing
BGP Inbound Optimization
The PfR BGP Inbound Optimization feature introduced the ability to support inside prefixes. Using BGP, PfR can select inside prefixes to support best entrance selection for traffic that originates from prefixes outside an autonomous system destined for prefixes inside the autonomous system. Company networks advertise the inside prefixes over the Internet using an Internet service provider (ISP) and receive advertisements for outside prefixes from an ISP.
BGP inbound optimization provides the ability to manually configure or automatically learn inside prefixes. The resulting prefixes can be monitored using link utilization threshold or link utilization range techniques. Link policies defining traffic load or range performance characteristics can be applied against PfR-managed entrance links. BGP inbound optimization provides the ability to influence inbound traffic by manipulating eBGP advertisements to influence the best entrance selection for traffic bound for inside prefixes.
Note |
Although PfR can learn an inside prefix, PfR will not try to control an inside prefix unless there is an exact match in the BGP routing information base (RIB) because PfR does not advertise a new prefix to the Internet. |
Prefix Traffic Class Learning Using PfR
The PfR master controller can be configured, using NetFlow Top Talker functionality, to automatically learn prefixes based on the highest outbound throughput or the highest delay time. Throughput learning measures prefixes that generate the highest outbound traffic volume. Throughput prefixes are sorted from highest to lowest. Delay learning measures prefixes with the highest round-trip response time (RTT) to optimize these highest delay prefixes to try to reduce the RTT for these prefixes. Delay prefixes are sorted from the highest to the lowest delay time.
PfR can automatically learn two types of prefixes:
-
outside prefix--An outside prefix is defined as a public IP prefix assigned outside the company. Outside prefixes are received from other networks.
-
inside prefix--An inside prefix is defined as a public IP prefix assigned to a company. An inside prefix is a prefix configured within the company network. The maximum number of inside prefixes that can be learned in a monitoring period is 30.
The PfR BGP Inbound Optimization feature introduced the ability to learn inside prefixes. Using BGP, PfR can select inside prefixes to support best entrance selection for traffic that originates from prefixes outside an autonomous system destined for prefixes inside the autonomous system. Company networks advertise the inside prefixes over the Internet using an Internet service provider (ISP) and receive advertisements for outside prefixes from an ISP.
PfR Link Utilization Measurement
Link Utilization Threshold
After an external interface is configured for a border router, PfR automatically monitors the utilization of the external link (an external link is an interface on a border router that typically links to a WAN). Every 20 seconds, by default, the border router reports the link utilization to the master controller. Both egress (transmitted) and ingress (received) traffic utilization values are reported to the master controller. If the exit or entrance link utilization is above the default threshold of 75 percent, the exit or entrance link is in an OOP state and PfR starts the monitoring process to find an alternative link for the traffic class. The link utilization threshold can be manually configured either as an absolute value in kilobytes per second (kbps) or as a percentage.
Link Utilization Range
PfR can also be configured to calculate the range of utilization over all the links. Both egress (transmitted) and ingress (received) traffic utilization values are reported to the master controller. In the figure below there are two border routers with exits links to the Internet through two ISPs. The master controller determines which link on one of the border routers--either BR1 or BR2 in the figure below--is used by a traffic class.
PfR range functionality attempts to keep the exit or entrance links within a utilization range, relative to each other to ensure that the traffic load is distributed. The range is specified as a percentage and is configured on the master controller to apply to all the exit or entrance links on border routers managed by the master controller. For example, if the range is specified as 25 percent, and the utilization of the exit link at BR1 (in the figure above) is 70 percent, then if the utilization of the exit link at BR2 (in the figure above) falls to 40 percent, the percentage range between the two exit links will be more than 25 percent and PfR will attempt to move some traffic classes to use the exit link at BR1 to even the traffic load. If BR1 (in the figure above) is being configured as an entrance link, the link utilization range calculations work in the same way as for an exit link, except that the utilization values are for received traffic, not transmitted traffic.
PfR Link Policies
PfR link policies are a set of rules that are applied against PfR-managed external links (an external link is an interface on a border router on the network edge). Link policies define the desired performance characteristics of the links. Instead of defining the performance of an individual traffic class entry that uses the link (as in traffic class performance policies), link policies are concerned with the performance of the link as a whole.
The BGP Inbound Optimization feature introduced support for selected entrance (ingress) link policies.
The following performance characteristics are managed by link policies:
-
Traffic Load (Utilization)
-
Range
-
Cost—Cost policies are not supported by the BGP Inbound Optimization feature. For more details about cost policies, see the "Configuring Performance Routing Cost Policies" module.
Traffic Load
A traffic load (also referred to as utilization) policy consists of an upper threshold on the amount of traffic that a specific link can carry. Cisco IOS PfR supports per traffic class load distribution. Every 20 seconds, by default, the border router reports the link utilization to the master controller, after an external interface is configured for a border router. Both exit link and entrance link traffic load thresholds can be configured as an PfR policy. If the exit or entrance link utilization is above the configured threshold, or the default threshold of 75-percent, the exit or entrance link is in an OOP state and PfR starts the monitoring process to find an alternative link for the traffic class. The link utilization threshold can be manually configured either as an absolute value in kilobytes per second (kbps) or as a percentage. A load utilization policy for an individual interface is configured on the master controller under the border router configuration.
Tip |
When configuring load distribution, we recommend that you set the interface load calculation on external interfaces to 30-second intervals with the load-interval interface configuration command. The default calculation interval is 300 seconds. The load calculation is configured under interface configuration mode on the border router. This configuration is not required, but it is recommended to allow Cisco IOS PfR to respond as quickly as possible to load distribution issues. |
Range
A range policy is defined to maintain all links within a certain utilization range, relative to each other in order to ensure that the traffic load is distributed. For example, if a network has multiple exit links, and there is no financial reason to choose one link over another, the optimal choice is to provide an even load distribution across all links. The load-sharing provided by traditional routing protocols is not always evenly distributed, because the load-sharing is flow-based rather than performance- or policy-based. Cisco IOS PfR range functionality allows you to configure PfR to maintain the traffic utilization on a set of links within a certain percentage range of each other. If the difference between the links becomes too great, PfR will attempt to bring the link back to an in-policy state by distributing traffic classes among the available links. The master controller sets the maximum range utilization to 20-percent for all PfR-managed links by default, but the utilization range can be configured using a maximum percentage value. Both exit link and entrance link utilization ranges can be configured as a PfR policy.
Note |
If you are configuring link grouping, configure the no max-range-utilization command because using a link utilization range is not compatible with using a preferred or fallback set of exit links configured for link grouping. With CSCtr33991, this requirement is removed and PfR can perform load balancing within a PfR link group. |
PfR Entrance Link Selection Control Techniques
The PfR BGP inbound optimization feature introduced the ability to influence inbound traffic. A network advertises reachability of its inside prefixes to the Internet using eBGP advertisements to its ISPs. If the same prefix is advertised to more than one ISP, then the network is multihoming. PfR BGP inbound optimization works best with multihomed networks, but it can also be used with a network that has multiple connections to the same ISP. To implement BGP inbound optimization, PfR manipulates eBGP advertisements to influence the best entrance selection for traffic bound for inside prefixes. The benefit of implementing the best entrance selection is limited to a network that has more than one ISP connection.
To enforce an entrance link selection, PfR offers the following methods:
BGP Autonomous System Number Prepend
When an entrance link goes out-of-policy (OOP) due to delay, or in images prior to Cisco IOS Releases 15.2(1)T1 and 15.1(2)S, and PfR selects a best entrance for an inside prefix, extra autonomous system hops are prepended one at a time (up to a maximum of six) to the inside prefix BGP advertisement over the other entrances. In Cisco IOS Releases 15.2(1)T1, 15.1(2)S, and later releases, when an entrance link goes out-of-policy (OOP) due to unreachable or loss reasons, and PfR selects a best entrance for an inside prefix, six extra autonomous system hops are prepended immediately to the inside prefix BGP advertisement over the other entrances. The extra autonomous system hops on the other entrances increase the probability that the best entrance will be used for the inside prefix. When the entrance link is OOP due to unreachable or loss reasons, six extra autonomous system hops are added immediately to allow the software to quickly move the traffic away from the old entrance link. This is the default method PfR uses to control an inside prefix, and no user configuration is required.
BGP Autonomous System Number Community Prepend
When an entrance link goes out-of-policy (OOP) due to delay, or in images prior to Cisco IOS Releases 15.2(1)T1 and 15.1(2)S, and PfR selects a best entrance for an inside prefix, a BGP prepend community is attached one at a time (up to a maximum of six) to the inside prefix BGP advertisement from the network to another autonomous system such as an ISP. In Cisco IOS Releases 15.2(1)T1, 15.1(2)S, and later releases, when an entrance link goes out-of-policy (OOP) due to unreachable or loss reasons, and PfR selects a best entrance for an inside prefix, six BGP prepend communities are attached to the inside prefix BGP advertisement. The BGP prepend community will increase the number of autonomous system hops in the advertisement of the inside prefix from the ISP to its peers. Autonomous system prepend BGP community is the preferred method to be used for PfR BGP inbound optimization because there is no risk of the local ISP filtering the extra autonomous system hops. There are some issues, for example, not all ISPs support the BGP prepend community, ISP policies may ignore or modify the autonomous system hops, and a transit ISP may filter the autonomous system path. If you use this method of inbound optimization and a change is made to an autonomous system, you must issue an outbound reconfiguration using the clear ip bgp command.
PfR Map Operation for Inside Prefixes
The operation of a PfR map is similar to the operation of a route-map. A PfR map is configured to select an IP prefix list or PfR learn policy using a match clause and then to apply PfR policy configurations using a set clause. The PfR map is configured with a sequence number like a route-map, and the PfR map with the lowest sequence number is evaluated first.
The BGP Inbound Optimization feature introduced the inside keyword to the match ip address (PfR) command to identify inside prefixes. Inbound BGP only supports the passive mode which results in some configuration restrictions when using a PfR map. The following commands are not supported in a PfR map for inbound BGP; set active-probe , set interface , set mode monitor , set mode verify bidirectional , set mos threshold , set nexthop , set periodic , set probe frequency , and set traceroute reporting .
Note |
Match precedence priority is not supported in PfR maps. |