
MVTX2804
Data Sheet
31
Zarlink Semiconductor Inc.
best effort classes to be used for non-essential traffic, because we provide no assurances about best effort
performance. However, in a typical network setting, much best effort traffic will indeed be transmitted, and with
an adequate degree of expediency.
Because we do not provide any delay assurances for best effort traffic, we do not enforce latency by dropping
best effort traffic. Furthermore, because we assume that strict priority traffic is carefully controlled before
entering the MVTX2804, we do not enforce a fair bandwidth partition by dropping strict priority traffic. To
summarize, dropping to enforce quality of service (i.e. bandwidth or delay) does not apply to strict priority or
best effort queues. We only drop frames from best effort and strict priority queues when global buffer resources
become scarce.
7.5 Weighted Fair Queuing
In some environments - for example, in an environment in which delay assurances are not required, but precise
bandwidth partitioning on small time scales is essential - WFQ may be preferable to a delay-bounded
scheduling discipline. The MVTX2804 provides the user with a WFQ option with the understanding that delay
assurances cannot be provided if the incoming traffic pattern is uncontrolled. The user sets eight WFQ “weights”
such that all weights are whole numbers and sum to 64. This provides per-class bandwidth partitioning with
error within 2%.
In WFQ mode, though we do not assure frame latency, the MVTX2804 still retains a set of dropping rules that
helps to prevent congestion and trigger higher level protocol end-to-end flow control.
As before, when strict priority is combined with WFQ, we do not have special dropping rules for the strict priority
queues, because the input traffic pattern is assumed to be carefully controlled at a prior stage. However, we do
indeed drop frames from SP queues for global buffer management purposes. In addition, queues P1 and P0 are
treated as best effort from a dropping perspective, though they still are assured a percentage of bandwidth from
a WFQ scheduling perspective. What this means is that these particular queues are only affected by dropping
when the global buffer count becomes low.
7.6 Shaper
Although traffic shaping is not a primary function of the MVTX2804, the chip does implement a shaper for
expedited forwarding (EF). Our goal in shaping is to control the peak and average rate of traffic exiting the
MVTX2804. Shaping is limited to class P6 (the second highest priority). This means that class P6 will be the
class used for EF traffic. (By contrast, we assume class P7 will be used for control packets only.) If shaping is
enabled for P6, then P6 traffic must be scheduled using strict priority. With reference to Table 2, only the middle
two QoS configurations may be used.
Peak rate is set using a programmable whole number, no greater than 64 (register QOS-CREDIT_C6_Gn). For
example, if the setting is 32, then the peak rate for shaped traffic is 32/64 1000 Mbps = 500 Mbps. Average rate
is also a programmable whole number, no greater than 64, and no greater than the peak rate. For example, if
the setting is 16, then the average rate for shaped traffic is 16/64 1000 Mbps = 250 Mbps. As a consequence of
the above settings in our example, shaped traffic will exit the MVTX2804 at a rate always less than 500 Mbps,
and averaging no greater than 250 Mbps.
Also, when shaping is enabled, it is possible for a P6 queue to explode in length if fed by a greedy source. The
reason is that a shaper is by definition not work-conserving; that is, it may hold back from sending a packet
even if the line is idle. Though we do have global resource management, we do nothing to prevent this situation
locally. We assume SP traffic is policed at a prior stage to the MVTX2804.