© 2006 Cisco Systems, Inc. All rights reserved.ONT v Implement the DiffServ QoS Model Introducing Congestion Avoidance

© 2006 Cisco Systems, Inc. All rights reserved.ONT v Managing Interface Congestion with Tail Drop

© 2006 Cisco Systems, Inc. All rights reserved.ONT v Managing Interface Congestion with Tail Drop Router interfaces experience congestion when the output queue is full: Additional incoming packets are dropped. Dropped packets may cause significant application performance degradation. Tail drop has significant drawbacks.

© 2006 Cisco Systems, Inc. All rights reserved.ONT v Managing Interface Congestion with Tail Drop Router interfaces experience congestion when the output queue is full: Additional incoming packets are dropped. Dropped packets may cause significant application performance degradation. Tail drop has significant drawbacks.

© 2006 Cisco Systems, Inc. All rights reserved.ONT v Managing Interface Congestion with Tail Drop Router interfaces experience congestion when the output queue is full: Additional incoming packets are dropped. Dropped packets may cause significant application performance degradation. Tail drop has significant drawbacks.

© 2006 Cisco Systems, Inc. All rights reserved.ONT v Tail Drop Limitations

© 2006 Cisco Systems, Inc. All rights reserved.ONT v Tail Drop Limitations Tail drop should be avoided because it contains significant flaws: TCP synchronization TCP starvation No differentiated drop

© 2006 Cisco Systems, Inc. All rights reserved.ONT v TCP Synchronization Multiple TCP sessions start at different times. TCP window sizes are increased. Tail drops cause many packets of many sessions to be dropped at the same time. TCP sessions restart at the same time (synchronized).

© 2006 Cisco Systems, Inc. All rights reserved.ONT v TCP Delay, Jitter, and Starvation Constant high buffer usage (long queue) causes delay. Variable buffer usage causes jitter. More aggressive flows can cause other flows to starve. No differentiated dropping occurs.

© 2006 Cisco Systems, Inc. All rights reserved.ONT v TCP Delay, Jitter, and Starvation Constant high buffer usage (long queue) causes delay. Variable buffer usage causes jitter. More aggressive flows can cause other flows to starve. No differentiated dropping occurs.

© 2006 Cisco Systems, Inc. All rights reserved.ONT v TCP Delay, Jitter, and Starvation Constant high buffer usage (long queue) causes delay. Variable buffer usage causes jitter. More aggressive flows can cause other flows to starve. No differentiated dropping occurs.

© 2006 Cisco Systems, Inc. All rights reserved.ONT v TCP Delay, Jitter, and Starvation Constant high buffer usage (long queue) causes delay. Variable buffer usage causes jitter. More aggressive flows can cause other flows to starve. No differentiated dropping occurs.

© 2006 Cisco Systems, Inc. All rights reserved.ONT v Random Early Detection

© 2006 Cisco Systems, Inc. All rights reserved.ONT v Random Early Detection Tail drop can be avoided if congestion is prevented. RED is a mechanism that randomly drops packets before a queue is full. RED increases drop rate as the average queue size increases. RED result: –TCP sessions slow to the approximate rate of output-link bandwidth. –Average queue size is small (much less than the maximum queue size). –TCP sessions are desynchronized by random drops.

© 2006 Cisco Systems, Inc. All rights reserved.ONT v RED Profiles

© 2006 Cisco Systems, Inc. All rights reserved.ONT v RED Modes RED has three modes: –No drop: When the average queue size is between 0 and the minimum threshold –Random drop: When the average queue size is between the minimum and the maximum threshold –Full drop (tail drop): When the average queue size is above the maximum threshold Random drop should prevent congestion (prevent tail drops).

© 2006 Cisco Systems, Inc. All rights reserved.ONT v TCP Traffic Before and After RED

© 2006 Cisco Systems, Inc. All rights reserved.ONT v Weighted Random Early Detection

© 2006 Cisco Systems, Inc. All rights reserved.ONT v Weighted Random Early Detection WRED can use multiple different RED profiles. Each profile is identified by: –Minimum threshold –Maximum threshold –Mark probability denominator WRED profile selection is based on: –IP precedence (8 profiles) –DSCP (64 profiles) WRED drops less important packets more aggressively than more important packets. WRED can be applied at the interface, VC, or class level.

© 2006 Cisco Systems, Inc. All rights reserved.ONT v WRED Building Blocks

© 2006 Cisco Systems, Inc. All rights reserved.ONT v Class-Based WRED Class-based WRED is available when configured in combination with CBWFQ. Using CBWFQ with WRED allows the implementation of DiffServ assured forwarding PHB. Class-based configuration of WRED is identical to stand- alone WRED.

© 2006 Cisco Systems, Inc. All rights reserved.ONT v WRED Profiles

© 2006 Cisco Systems, Inc. All rights reserved.ONT v IP Precedence and Class Selector Profiles

© 2006 Cisco Systems, Inc. All rights reserved.ONT v DSCP-Based WRED (Expedited Forwarding)

© 2006 Cisco Systems, Inc. All rights reserved.ONT v Configuring CBWRED

© 2006 Cisco Systems, Inc. All rights reserved.ONT v Configuring CBWRED random-detect router(config-pmap-c)# Enables IP precedence-based WRED in the selected class within the service policy configuration mode. Default service profile is used. Command can be used at the interface, perVC (with random-detect-group), or at the class level (service policy). Precedence-based WRED is the default mode. WRED treats non-IP traffic as precedence 0. policy-map Policy1 class mission-critical bandwidth percent 30 random-detect class transactional bandwidth percent 20 random-detect class class-default fair-queue random-detect

© 2006 Cisco Systems, Inc. All rights reserved.ONT v Changing the WRED Traffic Profile random-detect precedence precedence min-threshold max- threshold mark-prob-denominator Changes WRED profile for specified IP precedence value. Packet drop probability at maximum threshold is: 1 / mark-prob-denominator Nonweighted RED is achieved by using the same WRED profile for all precedence values. router(config-pmap-c)#

© 2006 Cisco Systems, Inc. All rights reserved.ONT v CBWFQ Using IP Precedence with CBWRED: Example Enable CBWFQ to prioritize traffic according to the following requirements: –Class mission-critical is marked with IP precedence values 3 and 4 (3 is high drop, 4 is low drop) and should get 30% of interface bandwidth. –Class bulk is marked with IP precedence values 1 and 2 (1 is high drop, 2 is low drop) and should get 20% of interface bandwidth. –All other traffic should be per-flow fair-queued. Use differentiated WRED to prevent congestion in all three classes.

© 2006 Cisco Systems, Inc. All rights reserved.ONT v CBWFQ Using IP Precedence with CBWRED: Example (Cont.)

© 2006 Cisco Systems, Inc. All rights reserved.ONT v WRED Profiles: DSCP-Based WRED (Assured Forwarding)

© 2006 Cisco Systems, Inc. All rights reserved.ONT v Configuring DSCP-Based CBWRED Enables DSCP-based WRED. Command can be used at the interface, perVC (with random detect group), or at the class level (service policy). Default service profile is used. The WRED random-detect command and the WFQ queue-limit command are mutually exclusive for class policy. random-detect dscp-based router(config-pmap-c)#

© 2006 Cisco Systems, Inc. All rights reserved.ONT v Changing the WRED Traffic Profile random-detect dscp dscpvalue min-threshold max-threshold mark-prob-denominator router(config-pmap-c)# Changes WRED profile for specified DSCP value Packet drop probability at maximum threshold is: 1 / mark-prob-denominator

© 2006 Cisco Systems, Inc. All rights reserved.ONT v CBWRED Using DSCP with CBWFQ: Example Enable CBWFQ to prioritize traffic according to the following requirements: –Class mission-critical is marked using DSCP AF2 and should get 30% of interface bandwidth. –Class bulk is marked using DSCP AF1 and should get 20% of interface bandwidth. –All other traffic should be per-flow fair-queued. Use differentiated WRED to prevent congestion in all three classes. Make sure that the new configurations still conform to the design and implementation from the previous example.

© 2006 Cisco Systems, Inc. All rights reserved.ONT v CBWRED Using DSCP with CBWFQ: Example (Cont.)

© 2006 Cisco Systems, Inc. All rights reserved.ONT v Monitoring CBWRED

© 2006 Cisco Systems, Inc. All rights reserved.ONT v Monitoring CBWRED show policy-map interface interface-name router# Displays the configuration of all classes configured for all service policies on the specified interface router#show policy-map interface Ethernet 0/0 Ethernet0/0 Service-policy output: Policy1 Class-map: Mission-critical (match-all) 0 packets, 0 bytes 5 minute offered rate 0 bps, drop rate 0 bps Match: ip precedence 2 Match: ip dscp Weighted Fair Queueing Output Queue: Conversation 265 Bandwidth 30 (%) Bandwidth 3000 (kbps) (pkts matched/bytes matched) 0/0 (depth/total drops/no-buffer drops) 0/0/0 exponential weight: 9 mean queue depth: 0 Dscp Transmitted Random drop Tail drop Minimum Maximum Mark (Prec) pkts/bytes pkts/bytes pkts/bytes threshold threshold probability 0(0) 0/00/00/ /10 1 0/00/0 0/ /10 2 0/00/0 0/ /10

© 2006 Cisco Systems, Inc. All rights reserved.ONT v Summary TCP uses windowing and the TCP slow-start mechanism as its means of controlling congestion. Tail drop causes significant issues, including TCP synchronization, starvation, and delay. TCP synchronization decreases the average utilization of network links. RED is a mechanism that randomly drops packets before a queue is full, preventing congestion and avoiding tail drop. RED operates by increasing the rate at which packets are dropped from queues as the average queue size increases.

© 2006 Cisco Systems, Inc. All rights reserved.ONT v Summary (Cont.) RED has three modes of operation: no drop, random drop, and full drop (tail drop). With RED, TCP global synchronization is eliminated and the average link utilization increases. WRED combines RED with IP precedence or DSCP and performs packet dropping based on IP precedence or DSCP markings. Each WRED profile defines the minimum and maximum threshold and the maximum drop probability. Profiles are already defined by default for IP precedence and DSCP.

© 2006 Cisco Systems, Inc. All rights reserved.ONT v