Explanation/Reference:
Section: Considerations for Expanding an Existing Network Explanation
Explanation:
Low-latency queuing (LLQ) provides strict-priority queues and prevents bandwidth starvation. LLQ supports the creation of up to 64 user-defined traffic classes as well as one or more strict-priority queues that can be used specifically for delay-sensitive traffic, such as voice and video traffic. Each strict-priority queue can use up to the maximum bandwidth available but can only use its guaranteed minimum bandwidth when other queues have traffic to send, thereby avoiding bandwidth starvation. Cisco recommends limiting the strict-priority queues to a total of 33 percent of the link capacity. Because LLQ can provide guaranteed bandwidth to delay-sensitive packets, such as Voice over IP (VoIP) packets, without monopolizing the available bandwidth on a link, LLQ is recommended for handling voice, video, and mission-critical traffic.
First-in-first-out (FIFO) queuing does not provide strict-priority queues or prevent bandwidth starvation. By default, Cisco uses FIFO queuing for interfaces faster than 2.048 Mbps. FIFO queuing requires no configuration, because all packets are arranged into a single queue. As the name implies, the first packet received is the first packet transmitted without regard for packet type, protocol, or priority.
Although you can implement priority queuing (PQ) on an interface to prioritize voice, video, and mission- critical traffic, you should not use it when lower-priority traffic must be sent on that interface. PQ arranges packets into four queues: high priority, medium priority, normal priority, and low priority. Queues are processed in order of priority. As long as the high-priority queue contains packets, no packets are sent from other queues. This can cause bandwidth starvation.
Custom queuing (CQ) is appropriate for voice, video, and mission-critical traffic, but it can be difficult to balance the queues to avoid bandwidth starvation of lower-priority queues. CQ is a form of weighted round robin (WRR) queuing. With round robin (RR) queuing, you configure multiple queues of equal priority and you assign traffic to each queue. Because each queue has equal priority, each queue takes turns sending traffic over the interface. With WRR queuing, you can assign a weight value to each queue whereby each queue can send a number of packets relative to their weight values. CQ allows you to configure each queue with a specific byte value whereby each queue can send that many bytes before the next queue can send traffic.
Although weighted fair queuing (WFQ) can be used for voice, video, and mission-critical traffic, it does not provide the bandwidth guarantees or the strict-priority queues provided by LLQ. WFQ is used by default on Cisco routers for serial interfaces at 2.048 Mbps or lower. WFQ addresses the jitter and delay problems inherent with FIFO queuing, and it addresses the bandwidth starvation problem inherent with PQ. Traffic flows are identified by WFQ based on source and destination IP addresses, port number, protocol number, and Type of Service (ToS). Although WFQ is easy to configure, it is not supported on high-speed links.
Class-based WFQ (CBWFQ) can be used for voice, video, and mission-critical traffic; however, it does not provide the delay guarantees provided by LLQ, because CBWFQ does not provide support for strict-priority queues. CBWFQ improves upon WFQ by enabling the creation of up to 64 custom traffic classes, each with a guaranteed minimum bandwidth. Bandwidth can be allocated as a value in Kbps, by a percentage of bandwidth, or by a percentage of the remaining bandwidth. Unlike with PQ, bandwidth starvation does not occur with CBWFQ.
Reference:
CCDA 200-310 Official Cert Guide, Chapter 6, Low-Latency Queuing, p. 235 Cisco: Enterprise QoS Solution Reference Network Design Guide: Queuing and Dropping Principles Cisco: Congestion Management Overview: Low Latency Queueing