IP_QOS PROJECT OBJECTIVE -----------------

1. Configure the following QOS-specific technologies:

1.1. Queuing schemes:

- First In First Out (FIFO), - Priority Queuing (PQ), - Custom Queuing (CQ), and - Weighted Fair Queuing (WFQ).

1.2. Committed Access Rate mechanism (CAR).

1.3. IPv6 QoS.

2. Provide example scenarios highlighting the benefits of these technologies. SCENARIO DESCRIPTION --------------------

This project includes nine scenarios. The role of each of the scenario is as follows:

1. None :

This scenario is a reference scenario. No queuing is defined at the IP layer.

2. FIFO :

This scenario illustrates FIFO queuing at the IP layer.

3. Priority_Queuing :

This scenario illustrates Priority Queuing at the IP layer.

4. Custom_Queuing :

This scenario illustrates Custom Queuing at the IP layer.

5. Custom_Queuing_with_LLQ :

This scenario illustrates the impact of using a low

latency queue in Custom Queuing.

6. WFQ :

This scenario illustrates WFQ Queuing at the IP layer.

7. WFQ_with_LLQ :

This scenario illustrates the impact of using a low latency queue in WFQ.

8-9. CAR_disabled / CAR_enabled : These scenarios illustrate the impact of CAR in a congested network.

10. MDRR_IPv6_IPv4: profiles.

This scenario illustrates queuing of IPv6 and IPv4 traffic at the same interface. The scenario uses local QoS

When/Why should I use QoS in my network? ----------------------------------------

Queuing schemes provide predictable network service by providing dedicated bandwidth, controlled jitter and latency, and improved packet loss characteristics.

The basic idea is to pre-allocate resources (e.g., processor and buffer space) for sensitive data. Each of the following schemes require customized configuration of output interface queues.

- Priority Queuing (PQ) assures that during congestion the highest priority data does not get delayed by lower priority traffic. However, lower priority traffic can experience significant delays.

(PQ is designed for environements that focus on mission critical data, excluding or delaying less critical traffic during periods of congestion.)

- Custom Queuing (CQ) assigns a certain percentage of the bandwidth to each queue to assure predictable throughput for other queues. It is designed for environments that need to guarantee a minimal level of service to all traffic.

- Weighted Fair Queuing (WFQ) allocates a percentage of the output bandwidth equal to the relative weight of each traffic class during periods of congestion.

RED is a dropping mechanism based upon the premise that adaptive flows such as TCP will back off and retransmit if they detect congestion. By monitoring the average queue depth in the router and by dropping packets, RED aims to prevent the rampup of too mamy TCP sources at once and minimize the effect of that congestion.

CAR is a traffic regulation mechanism, which defines a traffic contract in routed networks. CAR can classify and set policies for handling traffic that exceeds a certain bandwidth allocation.

CAR can be also used to set IP precedence based on application, incoming interface and TOS. It allows considerable flexibility for precedence assignement.

STATISTICS ----------

Global Statistics

-----------------

These statistics get updated by all the nodes/links in the network.

1) Links throughput

2) Application response time

3)

Application traffic sent and received

------------------------------------------------------Node Statistics ---------------

Statistics collected by all the nodes in the network.

1) Buffer size (packets) Size of a queue (annotated by "Q") on an interface (annotated by "IF")

2) Queuing delay (secs) Delay experienced by a packet from the time it arrives in the queue until it leaves it

3) Traffic received (bits/sec or packets/sec) Amount of traffic received by a queue (annotated by "Q") on an

interface (annotated by "IF")

4) Traffic sent (bits/sec or packets/sec) Amount of traffic sent from a queue (annotated by "Q") on an interface (annotated by "IF")

5) Traffic dropped (bits/sec or packets/sec) Amount of traffic dropped by a queue (annotated by "Q") on an interface (annotated by "IF"). Traffic can be dropped because of the queue size or RED/WRED

6) RED average Queue Size (packets) Average queue size of a queue (annotated by "Q") on an interface (annotated by "IF"). This size is computed for RED.

7) CAR incoming traffic dropped (bits/sec or packets/sec) Amount of incoming traffic dropped by CAR policies on an interface (annotated by "IF").

8) CAR outgoing traffic dropped (bits/sec or packets/sec) Amount of outgoing traffic dropped by CAR policies on an interface (annotated by "IF"). ###############################################################

Baseline ######### Network configuration =======================

The network is composed of four pairs of video clients. Each pair uses a distinct TOS (Type Of Service) for data transfer. The link between the two routers is a "potential" bottleneck. CONFIGURATION -------------

QoS configuration is performed on the following objects:

Routers support CAR and queuing functionnalities Router configuration ==================== QoS specification parameters are available on a per interface basis on every router. The sub-attribute called "Interface Information" in the "IP QoS Parameters" attribute is used to specify this information.

An incoming and outgoing CAR profile can be assigned to this interface as well as a queuing mechanism (FIFO, WFQ, PQ, CQ) with its queuing profile.

"Queuing Profiles" are special schemes defining different queue configuration options. There are two kinds of "Queuing Profiles":

a) Global profiles - defined on the "QoS Attribute Configuration" object. These can be applied netork-wide on any router. Global profiles can be configured only manually.

b) Local profiles - defined on a router under the "IP QoS Parameters" attribute. These can be applied only locally on the router where they are configured. Local profiles are created mainly via the Device Configuration Import (DCI), but can be also configured manually.

All scenarios except scenario MDRR_IPv6_IPv4 use global profiles. Check out the "QoS Attribute Configuration" object for a list of the default profiles. ----------------------------------Traffic sources assign type of service Traffic Specification ===================== Clients and servers are in charge of defining the traffic prioritization by setting the IP precedence in the TOS (Type of Service) or DSCP (Differentiated Sevice Codepoint) field of IP datagrams. Traffic is prioritized on session basis.

In clients the attribute "Application Configuration" defines for each application (Email, Ftp, Http ...) the type of service.

In servers the attribute "Supported Services" defines for each application the type of service as well.

Note that the type of service is not the only criterion to prioritize traffic in routers. Traffic can be also prioritized based on IP addresses, protocol, application ports ... ----------------------------------------QoS Attribute Configuration object defines global QoS profiles used in the network QoS Attribute Configuration Object ==================================== The QoS Attribute Configuration object defines global (as opposed to local) profiles for the following technologies: - CAR - FIFO - WFQ - Custom Queuing - Priority Queuing

Global profiles can be applied netork-wide on any router.

Each queuing-based profile (e.g., FIFO, WFQ, PQ, CQ) contains a table in which each row represents one queue. Each queue has many parameters such as queue size, classification scheme, RED parameters , etc.

Note that the classification scheme can also be configured to contain many different criteria by increasing the number of rows.

Some examples of setting queue priorities are: - Weight for WFQ profile. Higher priority is assigned to the queue with a higher weight. - Byte count for Custom Queuing profile. More traffic is served from the queue with a higher byte count. - Priority label for Priority Queuing. Higher priority is assigned to the queue with a higher priority label.

The CAR profile defines a set of classes of service (COS). Each row represents a COS for which CAR policies have been defined.

NOTE: Local QoS profiles can be configured on a router under the "IP QoS Parameters" attribute. These can be applied only locally on the router where they are configured. Local profiles are created mainly via the Device Configuration Import (DCI), but can be also configured manually.

##################################### FIFO Network configuration =======================

The network is composed of four pairs of video clients. Each pair uses a distinct TOS (Type Of Service) for data transfer. The link between the two routers is a "potential" bottleneck.

FIFO queuing can be enabled on each interface in "advanced" routers. Queuing profile and queuing processing mechanism are set in a sub-attribute called "Interface Information" in the "IP QoS Parameters" compound attribute. Queuing profile defines the number of queues and the classification scheme. Global queuing profiles are defined in the QoS configuration object. This object is found in "utilities" palette.

Results =========

Traffic is queued in "router A" because of the bottleneck. Since "router A" has limited buffer capacity, some packets are dropped when the buffer usage reaches its full capacity.

The application response time can be seen to reach a threshold because packet that arrive on a full queue always get dropped.

Note that the maximum delay that an arriving packet observes is the delay encountered as a result of servicing all the packets ahead of it in an almost full queue. ################################################################

Results =========

Traffic is queued in "router A" because of the bottleneck. Since "router A" has unlimited buffer capacity no packets get dropped.

The application response time keeps on increasing as the packets get queued indefinitely without ever getting dropped.

######################################### GLOSSARY --------

CAR

Committed Access Rate

COS

Class of Service

CQ

Custom Queuing

FIFO

First In First Out

LLQ

Low Latency Queue

PQ

Priority Queuing

QoS

Quality of Service

RED

Random Early Detection

TOS

Type Of Service

WFQ

Weighted Fair Queuing

WRED

Weighted Early Detection