Answer: B. Only the Integrated Services (IntServ) model made use of control plane signaling for resource allocation.

Answer: A. In the DiffServ model, a BA is a collection of packets with a shared code point. It is expected that each node will have the same PHB for a given BA, and therefore end-to-end performance can be modeled.

Answer: C. J-series WRED has four drop priority levels, but no TCP/UDP index. The M-series can support eight queues and per-unit scheduling with IQ PICs. Only the J-series offers true LLQ.

Answer: D. Although you could use multifield classification everywhere, this approach does not scale. Use a BA classifier and associated rewrite to convey PLP status between nodes.

Answer: B. The M7i has two hardware priority levels, and medium-high, high, and strict-high all map to the same value, which is high.

Answer: D. Because strict-high is given 100% of transmit weight, it should be used with a policer to ensure that other classes are not starved, especially on the J-series, where strict-high is an actual priority. You cannot use exact with a strict-high queue, but on the J-series you can use the shaping rate to cap total usage. However, a policer is preferred, as this allows excess bandwidth only when other queues are empty.

Answer: B. Supported on the J-series only, the shaping-rate limits the total amount of bandwidth available to the queue, regardless of activity in other queues. This in itself does not prevent starvation of lesser-priority queues, but it can help.

Answer: A. Many CoS configuration errors allow a commit, but they generate a log warning indicating that the configured values can be programmed. This means the default values are in effect.

Answer: C. Unless you shape at the logical interface level, each IFL can send up to line rate, and when multiple IFLs are active, they share available bandwidth. This is another benefit to using schedulers based on transmit percentage, rather than absolute values. The latter would result in one IFL getting 100 Mbps while the other receives a default scheduler configuration with 95%/5% assigned to queue 0 and 3, respectively.

Answer: A. The profile defines a 10% drop probability for fill levels between 70% and 80%. You cannot tell from the drop profile itself whether it affects PLP 0 or 1 (or UDP versus TCP), because the function of WRED against some criteria is performed via a scheduler-map, which was not shown.

Answer: B. Multifield overrides BA classification and generally is used only at network edges.

Answer: E. All of the methods listed can impact the PLP status of a given packet.