Understanding Enabling Export of Subscriber Statistics and Queue Statistics for Dynamic Interfaces and Interface-Sets
You can use subscriber statistics and queue statistics for dynamic interfaces and interface-sets to support remote analytics and monitor Juniper devices that operate as a Broadband Network Gateway (BNG). Using these statistics, you can model and condition traffic flows in a subscriber access network.
About Subscriber and Queue Statistics
Subscriber statistics include the per-IP protocol family (IPv4 or IPv6) packet information (receive and transmitted packets and bytes) for a subscriber interface. They will only include subscriber data forwarded by the system. Filtered and dropped packets and control traffic are factored out and not delivered.
ON-CHANGE subscription support for interface meta-data sends
asynchronous notifications when interfaces are created and deleted.
After an initial baseline of delivering create notifications
for all existing interfaces, only notifications for interfaces that
are being created or deleted are sent to an external collector.
Use queue statistics to determine oversubscription levels, the mix of forwarding-class traffic, or traffic rates for a given CoS-enabled interface or interface-set.
To receive subscriber statistics, you also must enable RADIUS accounting. See 802.1X and RADIUS Accounting.
Enabling Export of Statistics
To receive statistics, you enable both meta-data and statistical data for export on your Juniper device through the Junos CLI. Meta-data for the interface is provided because the interface key is a dynamic integer, a session identifier (SID), which conveys no context to an external server. The meta-data provides more tangible context (such as the user name, a profile name VLAN tags, etc.) to the SID. An external collector associates the statistical data to a persistent reference.
A subscription for both statistical data and meta-data can be made from the external collector (in Figure 1, the JTI collector). In this way, the two streams are “merged” and a correlation is made between the statistical data and the meta-data. The dynamic SID is matched with the more permanent attributes such as user name and location.
