CCNP ONT Notes4 Apr 2008
Chapter 4: Congestion Management and QueuingThe default queuing method on an interface faster than 2.048 Mbps is First In, First Out (FIFO). Interfaces operating at 2.048 Mbps or slower perform Weighted Fair Queuing (WFQ). Each physical interface has hardware and software queuing mechanisms; software queues are only used when the hardware queue is congested. Tail-drop occurs when all queues are full and a packet is dropped. Hardware queue sizes can be configured with Simple QueuingFirst-In-First-Out (FIFO)Packets are transmitted in the order they are received with no preference (no QoS). FIFO is the default mechanism for interfaces >2.048Mbps. Priority Queuing (PQ)PQ provides four queues: high, medium, normal, and low. All packets in a higher priority queue will be processed before any packets in a lower priority queue. Lower priority queues can be starved if higher priority queues consume all available bandwidth. PQ is implemented by defining and applying priority lists:
Round Robin (RR)All queues are equal priority; one packet is taken from each queue per cycle. Round robin does not provide for traffic prioritization, and queues with larger packets will consume more bandwidth than queues with smaller packets. Weighted Round Robin (WRR)WRR is a modification to RR which allows for disproportionate allowance of bandwidth to queues. Custom Queuing (CQ) is an example of WRR; it specifies a certain number of bytes to be processed from each queue. Weighted Fair QueuingWFQ is the default mechanism on and only supported on interfaces less than or equal to 2.048 Mbps. WFQ queues are created per flow and are not configurable. Each flow is assigned to a dynamic FIFO queue by source/destination IP address, protocol number, ToS value, or source/destination port number. The maximum number of dynamic queues is configurable between 16 - 4096 (256 by default). Packets are dropped from aggressive flows more frequently than from less aggressive flows. The hold queue is the sum of all memory available to the WFQ system; all packets are aggressively dropped while the hold queue is full. Each queue has a Congestive Discard Threshold (CDT) which allows for early dropping of packets before the queue is completely full. WFQ can be disabled on an interface with Queue information can be viewed with Class-Based Weighted Fair Queuing (CBWFQ)CBWFQ is similar to WFQ but with user-defined queue classes instead of dynamically created flow-based queues. CBWFQ supports a maximum of 64 queues. Each queue is allotted a certain amount or percentage of the available bandwidth. The default queue named Bandwidth can be allocated in Kbps, percentage, or remaining percentage. All classes within a policy map must use the same unit of measure (Kbps or percentage). The default maximum reserved bandwidth is 75%; this can be modified with Fair queuing (instead of FIFO) can be enabled for the default class with The queue size for each class can be adjusted with Configuration example:
Low Latency Queuing (LLQ)LLQ implements a strict-priority queue which is favored over all other queues. LLQ is typically used for delay-sensitive traffic like VOIP. The priority queue is policed to a certain bandwidth to prevent starvation of other queues. Priority queues are created under a class with Configuration example:
|
Navigation
Armory
Online Toolbox
|