Starting with version 4.0 IRP is able to perform loss and latency optimization of inbound traffic.
This feature allows improving the inbound traffic performance by deflecting it from the worst-performing provider. The improvement actions do not guarantee the best path for the inbound traffic but minimize the usage of the worst path. For the traffic deflection consistency, IRP uses provider traffic engineering capabilities.
Some network providers give customers the ability to influence their traffic based on the announced BGP community, the so-called traffic engineering capability. Customers of such providers can check if the upstreams support this functionality by reviewing the upstreams BGP routing policies, formerly published in Whois/RDAP for provider’s Autonomous System, or by contacting providers directly. In other words, traffic engineering capability, therefore, represents a list of BGP communities that a provider configures on their side for the customer’s use. When prefixes tagged with a specific BGP community from such lists get announced by clients, the upstream provider applies the corresponding action to those prefixes.The requirements for the Inbound performance feature are:
- Provider traffic engineering capability using BGP community
- Ability to re-announce providers’ BGP table to IRP. Configuring the BGP monitoring Protocol (BMP monitoring station) is recommended
- Flow export (The feature is not supported in the SPAN-only setups)
Calculation of an inbound performance improvement takes more time, when compared to the regular outbound improvement. The reason is that the inbound traffic is affected much slower than the outbound traffic and the calculation method is different.For the Inbound performance feature to work, IRP must know what traffic should be monitored and improved. For that, the Inbound Performance rule(s) must be configured. More about Inbound performance rules: Inbound Performance Rules
IRP performs the analysis of network performance indicators and applies improvements per individual preconfigured rules in the following order:
- Probes prefixes that generate inbound traffic to build a statistical model of the route performance per each inbound rule
- Reprobes prefixes according to the configured inbound reprobing interval (irpinperfd.probing.interval)
- Compares the reprobed results with the statistical model
- In case the poor inbound traffic metrics are obtained – IRP makes an improvement by announcing the customer prefixes with the corresponding TE community
- Once IRP applies an inbound improvement, its relevance is checked on subsequent probing sessions
There are two modes in which the inbound performance improvements can be enabled: #
• Automated – IRP enables/disables all the inbound performance improvements automatically once an improvement is calculated
• Moderated – stands for manual confirmation of the inbound performance improvements. Once IRP detects an improvement, a suggestion to enable it gets displayed in the GMI interface. IRP does not perform any improvements without user confirmation. If an improvement is not enabled in time before a subsequent cycle of checks, IRP reviews the proposal and adjusts it as needed, keeping the recommendation, making a different one, or removing the improvement proposal
The inbound performance improvements can be enabled manually as per a particular rule. This is done at the discretion of the user. IRP does not make any calculations/proposals in such casesTo view how to configure the Inbound Performance feature, go to: Inbound performance optimization configuration
