Automatic adjustment of RSVP LSP bandwidth based on measured traffic rate into the tunnel requires the LSP to be configured for egress statistics collection at the ingress LER. The following CLI shows an example:
config router mpls lsp name
egress-statistics
accounting-policy 99
collect-stats
no shutdown
exit
All LSPs configured for accounting, including any configured for auto-bandwidth based on traffic measurements, must reference the same accounting policy. An example configuration of such an accounting-policy is shown below: in the CLI example below.
config log
accounting-policy 99
collection-interval 5
record combined-mpls-lsp-egress
exit
exit
The record combined-mpls-lsp-egress command in the accounting policy has the effect of recording both egress packet and byte counts and bandwidth measurements based on the byte counts if auto-bandwidth is enabled on the LSP.
When egress statistics are enabled the CPM collects stats from of all XCMs or IOMs involved in forwarding traffic belonging to the LSP (whether the traffic is currently leaving the ingress LER via the primary LSP path, a secondary LSP path, an FRR detour path or an FRR bypass path). The egress statistics have counts for the number of packets and bytes forwarded per LSP on a per-forwarding class, per-priority (in-profile vs. out-of-profile) basis. When auto-bandwidth is configured for an LSP the ingress LER calculates a traffic rate for the LSP as follows:
Average data rate of LSP[x] during interval[i] = F(x, i)—F(x, i-1)/sample interval
F(x, i) — The total number of bytes belonging to LSP[x], regardless of forwarding-class or priority, at time[i]
sample interval = time[i] — time [i-1], time[i+1] — time[i], and so on.
The sample interval is the product of sample-multiplier and the collection-interval specified in the auto-bandwidth accounting policy. A default sample-multiplier for all LSPs may be configured using the config>router>mpls>auto-bandwidth-defaults command but this value can be overridden on a per-LSP basis at the config>router>mpls>lsp>auto-bandwidth context. The default value of sample-multiplier (the value that would result from the no auto-bandwidth-defaults command) is 1, which means the default sample interval is 300 seconds.
Over a longer period of time called the adjust interval the router keeps track of the maximum average data rate recorded during any constituent sample interval. The adjust interval is the product of adjust-multiplier and the collection-interval specified in the auto-bandwidth accounting-policy. A default adjust-multiplier for all LSPs may be configured using the config>router>mpls>auto-bandwidth-multiplier command but this value can be overridden on a per-LSP basis at the config>router>mpls>lsp>auto-bandwidth context. The default value of adjust-multiplier (the value that would result from the no auto-bandwidth-multiplier command) is 288, which means the default adjust interval is 86400 seconds or 24 hours. The system enforces the restriction that adjust-multiplier is equal to or greater than sample-multiplier. It is recommended that the adjust-multiplier be an integer multiple of the sample-multiplier.
The collection-interval in the auto-bandwidth accounting policy can be changed at any time, without disabling any of the LSPs that rely on that policy for statistics collection.
The sample-multiplier (at the mpls>auto-bandwidth level or the lsp>auto-bandwidth level) can be changed at any time. This has no effect until the beginning of the next sample interval. In this case the adjust-interval does not change and information about the current adjust interval (such as the remaining adjust-multiplier, the maximum average data rate) is not lost when the sample-multiplier change takes effect.
The system allows adjust-multiplier (at the mpls level or the lsp>auto-bandwidth level) to be changed at any time as well but in this case the new value shall have no effect until the beginning of the next adjust interval.
Byte counts collected for LSP statistics include layer 2 encapsulation (Ethernet headers and trailers) and therefore average data rates measured by this feature include Layer 2 overhead as well.