QoS Overview
Objective
Introduction
QoS Service Models
Traffic Characterization
Different QoS Techniques
Congestion Mangement
Congestion - Avoidance
What is QoS ?
QoS
represents the set of techniques
necessary to manage network bandwidth,
delay, jitter and packet loss.
From
a business perspective, it is essential to
assure that the critical applications are
guaranteed the network resources they need,
despite varying network traffic load.
QoS Service Models
Best effort (No QoS)
Integrated services
Differentiated services
Best Effort Model
Well do the best we can
But messages may be lost on route.
Traditional datagram model
Integrated Services Model
Flow
based QoS model
A kind of virtual circuit is created between
source and destination
Strict Bandwidth Reservations
Needs Signaling to accomplish Path
Reservation
Resource Reservation Protocol
(RSVP)
Must be configured on every router along the
path.
Integrated Services Model(Cont.)
Advantages:
1. It is simple.
2. Discrete per flow QoS makes it suitable
for voice calls.
Disadvantages:
1. Has issues with scaling with large
number of
flows.
2. Requires devices to retain state
information.
Differentiated Services Model
DiffServ Architecture
Class based QoS model
Creates a means for traffic conditioning (TC)
Defines per-hop behavior (PHB)
Edge nodes perform TC
Differentiated Services Model (Cont.)
Advantages:
1. Scales well with large flows as no state or
flow
information is required to be
maintained.
2. The main processing was moved from the
core
to the edge. Thus high performance.
Disadvantages:
No end to end bandwidth reservations are
present.
DiffServ Code Point (DSCP):
IPv4 Type of Service Byte
(IP precedence bit and DSCP)
Per Hop Behaviour(PHB):
There are 4 types of standard PHBs:
Default PHB
DSCP value is 000000.
Class-Selector PHB
Traditional best effort services will be provided .
Provides code points that can be used for backward
compatibility with IP Precedence.
DSCP value in the form xxx000, where x is either 0 or 1.
Per Hop Behaviour(PHB):
Assured
Forwarding:
The AF PHB defines four AF classes.
Within each AF class, there are three drop
precedence (dP) values: 1, 2, and 3.
Expedited Forwarding:
Provides a strict priority service.
It provides low loss,low latency,low jitter,
assured bandwidth service.
and
Traffic Characterization
Traffic can be classified into following types:
Voice :
Voice traffic should be marked to DSCP EF.
Call Signalling Traffic:
Marked as DSCP CS3.
It is intended for voice and video signalling traffic such as SIP .
Video :
Interactive-Video :
It should be marked to DSCP AF41.
Streaming Video:
It should be Marked as DSCP CS4.
Non Organizational video be marked as scavenger.
Traffic Characterization
Data:
1. Best Effort Data
Marked as DSCP 0.
It is the default class.
2. Bulk Data
Marked as DSCP AF11
3.Transactional Data/Interactive Data
Marked as DSCP AF21.
4. Mission Critical Data
Marked as DSCP AF31.
Only few applications should be assigned to
this
class.
Traffic Characterization (Cont)
Control Plane Data:
1. IP Routing
Marked as DSCP CS6.
2. Network Management:
Marked as DSCP CS2
Scavenger Class:
Marked as DSCP CS1.
Generally entertainment oriented in nature.
Should be given lowest queuing services.
BANDWIDTH DISTRIBUTION
QoS Triangle
CLASSIFICATION AND MARKING
Objective:
For defining QoS policy Identify the traffic that is to be treated differently.
Classifiers :
1. Inspect fields in a packet
2.Identify the type of traffic.
3. Direct the traffic to the applicable policy-enforcement mechanism
for that traffic type
Markers:
Markers Write a field within the packet to preserve the classification
decision that was reached at the trust boundary
.
POLICING AND SHAPING
AIM:
To identify and respond to traffic violations.
Policers and shapers identify traffic
violations in an identical manner. However,
they differ in how they respond to the
violations.
POLICERS
It Performs:
1. Checks for traffic Violations.
2. Takes prescribed action(Remarking or Dropping of Packets )
for violations when occur.
SHAPERS
Objective:
To send all the traffic offered to an interface, but to smooth it
out so that it never exceeds a given rate.
Shapers are traffic-smoothing tools work in conjunction with
queuing mechanisms
Token Bucket Algoritham
Policers and shapers are modelled after token bucket
algorithms.
Metering Engines.
A token permits the algorithm to send a single bit/
byte of traffic.
Single-Rate Two-Color Policer
Example:
CIR =8000 bps
8000 tokens are placed in a "bucket" at the beginning of the time period.
Each time a bit of traffic is offered the bucket is checked for tokens.
Conform :
If there are tokens in the bucket, the traffic is passed. One token is removed
from the bucket for each bit of traffic that is passed.
The specified action for conforming traffic is taken(Transmitted).
Exceed:
When the bucket runs out of tokens, any additional offered traffic is viewed
to
exceed the rate.
Exceed action is taken(remarked /dropped)
Single-Rate Three-Color Marker/Policer
Single-Rate Three-Color Marker/Policer
Advantages:
Tolerance to temporary burst according to
credits in the second bucket.
Fewer TCP retransmissions and, thus, more
efficient bandwidth utilization.
Single-Rate Three-Color Marker/Policer
Drawbacks:
Accumulated excess burst credits could cause
a degree of unpredictability in traffic flows.
To improve on this, a two-rate three-color
marker/policer was defined in RFC 2698.
Two-Rate Three-Color Marker/Policer
Two-Rate Three-Color Marker/Policer
The first bucket is filled with the PIR number of
tokens
Second bucket is filled with the CIR number of
tokens.
Initial check :whether the traffic is within the PIR ?
Then compare whether traffic is within the CIR?
Sustained excess bursts are permitted in excess
of the CIR, but only until the PIR is reached
Two-Rate Three-Color Marker/Policer:
Congestion/Queuing-Management:
Why congestion?
Whenever packets enter a device faster than
they can exit it, the congestion occurs.
when congestion occurs, packets must be
buffered, or queued, to mitigate dropping.
Scheduling and Queuing
Queuing: Ordering packets in linked output buffers.
Scheduling :The process of deciding which packet to
transmit next using schedulinng algorithams.
QUEING MECHANISMS:1.Priority queuing (PQ)
QUEING MECHANISMS: 2 .Custom queuing (CQ)
QUEING MECHANISMS: 3 .WFQ
QUEING MECHANISMS: 3 .WFQ
Whenever
packet arrives, then the classifier snoops its
header and using this information (source address and
port, destination address, ip precedence, protocol, etc.)
calculates a number between "1" and "number of
queues".
Then, it locates the packet in the queue identified by
this number.
WFQ allots skewing bandwidth by favouring higherpriority flows (priority being a function of IPP marking).
Certain flows are preferred over others, WFQ lost CQ's
capability to provide bandwidth guarantees because
bandwidth allocations continuously changed as flows
were added or ended.
QUEING MECHANISMS: 4 .CBWFQ
CBWFQ enables the creation of up to 256 classes of
traffic, each with its own reserved queue.
Major differences between WFQ and CBWFQ :
Each queue is serviced based on the bandwidth
assigned to each class.
With WFQ, the bandwidth is calculated
instantaneously; but with CBWFQ, a minimum
bandwidth explicitly is defined and enforced
Drawbacks:
Highly efficient algorithm for data applications, but it lacks
the capability to provide strict-priority to real-time
applications because the scheduler can serve other queues
when a VoIP or video packet is waiting in a queue.
QUEING MECHANISMS: 5 .LLQ
Two components:
PQ for the real-time sensitive traffic as PQ
is the optimal queuing mechanism for realtime traffic,
CBWFQ is the best queuing algorithm for
data applications.
QUEING MECHANISMS: 5 .LLQ
Congestion-Avoidance
Global Synchronisation:
Random Early Detect (RED)
Designed
to work with a transport protocol like
TCP
Avoids global synchronization of many
connections
It randomly drops packets before queues fill to
capacity.
Monitors buffer length and perform early
discard on random packet when queue
threshold has exceeded.
Weighted RED (WRED)
WRED
is an enhancement to RED
It influences the randomness of the selection
of packets to be dropped.
WRED factors the weight or IPP of the packet
into the drop selection process.
WRED based on IPP
WRED drops packets with lower IPP values
sooner than packets with higher IPP values
WRED based on DSCP
WRED uses AF drop preference values to drop
packets.
Thank you
