The FireBrick has Constant Quality Monitoring. When used with the FireBrick acting as an LNS the CQM graphs play an important role.

Per connection monitoring. Each connection is assigned a CQM graph name. This is normally set based on the circuit ID passed by the carrier, or if not present, the username used. A long graph name (over 20 characters) is reduced to a hash. The name can also be set by RADIUS response (Chargeable User Identity attribute). Each graph shows the send and receive throughput for each 100 seconds. The graph also shows the loss and latency with minimum, average, and maximum per 100 seconds. This is based on an LCP echo sent every second on every connection. The interval can be configured to be lower if you wish (either in the config or by RADIUS).

The per connection graphs also have a downlink speed setting. This is set based on the connection speed from L2TP connection. This can also be set in the RADIUS response. This limits the speed of traffic to the line. This is usually done so that the LNS is in control of the speed of the line as the FireBrick will drop larger packets before smaller packets, which helps VoIP and many other protocols work well even on a full link. The speed control can also be used to provide slower services.

In addition to the per-connection graphs, there is also an aggregate graph based on the incoming L2TP connection settings - e.g. typically for a whole carrier. This tracks the overall throughput for all of the lines. This is useful simply for reporting and tracking, but the aggregate graph can also have a speed setting. This allows rate limiting to meet commit levels with carriers, which can be very important where, for example, there is 100th percentile billing.

This also allows a damping setting to be used. Where the aggregate is hitting the limit, all lines within that aggregate are reduced in their shaper settings by a percentage to damp the overall throughput. The continued hitting of the aggregate increases the percentage level. Individual lines can be tagged high or low priority by RADIUS which affects the level of damping applied, and so allows three grades of service when an aggregate link is full. At each stage, aggregate and per line, the shaping still drops larger packets first making for a very effective way to manage overall traffic levels.

It is also possible to set a third level of aggregation, where each connection can be placed in a group which is, itself, another CQM graph. This can be useful for tracking and shaping at a per wholesale customer or customer grouping in some way.