You are on page 1of 35

QoS Management in IP Networks:

Integrated and Differentiated


Services

Lund University / Presentation 2012


Review

Demands on IP-based internets are rising


IP-based internets were designed for elastic applications
that tolerate variations in throughput and loss
Now, they are used to support high volumes and various
traffic mix including real-time and non real-time applications
These are sensitive to delay and throughput variations and require
high quality of service (QoS)
Thus, they now need to provide service differentiations for
different applications like ATM network!

Lund University / Presentation 2012


Introduction
New additions to Internet increasing traffic
High volume client/server application
Web
Real time voice and video
Need to manage traffic and control congestion
Two complementary IETF standards:
Integrated services (IntServ)
Provides collective service to set of traffic demands placed
in a domain
Limit demand per capacity & reserve resources to meet
QoS
Differentiated services (DiffServ)
Classify traffic in groups
Different group traffic handled differently

Lund University / Presentation 2012


Internet Traffic
Elastic
Can adjust to changes in delay and throughput
E.g. common TCP and UDP application like email, FTP, web
Inelastic
Does not easily adapt to changes in delay and throughput
real time traffic such as web streaming, voice over IP
(VoIP)
Requires minimum throughput, bounded delay and jitter
(i.e. variation of delay)

Lund University / Presentation 2012


IntServ Architecture

IPv4 header fields for precedence/priority and type of service


usually ignored
ATM only network designed to support TCP, UDP and real-time
traffic from inception
not available everywhere, costly reconfiguration
Need to support Quality of Service (QoS) within TCP/IP architecture
Requires adding functionality to routers
Means of requesting QoS

Lund University / Presentation 2012


IntServ Approach

Enable provision of QoS over IP (RFC2211,2212)


Enables sharing available capacity when congested
Currently, routers have these mechanisms:
Dynamic Routing Algorithms
Select to minimise delay to balance load
Active Queue Management (AQM)
Causes TCP sender to back off and reduce load

These are not sufficient, and are enhanced by IntServ

Lund University / Presentation 2012


IntServ Functions
Admission control
For specific QoS, reservation required for new flow
Resource reSerVation Protocol (RSVP) used
Routing algorithm
Base decision on QoS parameters, not shortest
path only
Queuing discipline
Take account of different flow requirements
Meet QoS
AQM policy
Manage congestion

Lund University / Presentation 2012


IntServ Implementation in Router

Background
Functions

Forwarding
functions

Lund University / Presentation 2012


IntServ Components
Background Functions
Reservation Protocol
Reserve resources for new flows
Admission control
Determines whether current resources enough to support new request
Management agent
Can use agent to modify traffic control database and direct admission
control
Routing protocol
Directs next hop for each address and flow

Lund University / Presentation 2012


IntServ Components
Forwarding
Classifier and route selection
Incoming packets mapped to classes
Single flow or set of flows with same QoS
E.g. all video flows
Based on IP header fields
Determines next hop
Packet scheduler
Manages one or more queues for each output
Order in which queued packets sent
Based on class, traffic control database, current and past activity on
outgoing port
Policing
Determine whether flow exceed its requested capacity

Lund University / Presentation 2012


IntServ Services
Service defined on two levels
General categories of service:
Guaranteed
Controlled load
Best effort (default)
Particular flow within category
Service for a flow is specified by certain
parameters known as traffic specification
(TSpec)
TSpec is part of the traffic contract

Lund University / Presentation 2012


IntServ Services Guaranteed
Service

Most demanding service


Provides assured data rate
Has specific upper bound on queuing delay
through network
Must be added to propagation delay to get total
delay
May be wise to set high to accommodate rare long
queue delays
Has no queuing losses
i.e. no buffer overflow

Lund University / Presentation 2012


IntServ Services Controlled
Load
Tightly approximates to best effort under unloaded
conditions
No upper bound on queuing delay
High percentage of packets do not experience delay over
minimum transit delay
Propagation delay plus router processing with no queuing
delay
Very high percentage delivered
Almost no queuing loss
Useful for adaptive (or soft) real time applications

To provide these service categories, routers adopt


suitable queuing discipline

Lund University / Presentation 2012


Queuing Disciplines
Traditionally FIFO or FCFS at each router port
No special treatment to high priority packets
(flows)
Small packets held up by large packets ahead
of them in queue
Larger average delay for smaller packets
Flows of larger packets get better service
Greedy TCP connection can crowd out
altruistic (i.e. unselfish) connections
If one connection does not back off, others may
back off more

Lund University / Presentation 2012


FCFS

Work Conserving (if packet waiting, serve)


Klienrock Conservation Law
= occupancy
q = mean scheduler delay
C = a constant

If delay for one flow is lowered, the delay


for one or more other flows must increase

Lund University / Presentation 2012


Non Work Conserving
Scheduler can be idle even if packets waiting
Switches packets to
The right destination
At the right time
Reduces jitter
Makes traffic predictable
Mechanism known as shaping

Lund University / Presentation 2012


Fair Queuing (FQ)

Also work conserving


Multiple queues for each port
One for each source or flow
Queues serviced in round robin
Each busy queue gets exactly one packet per cycle
Achieves load balancing among flows
No advantage to being greedy
Your queue gets longer, increasing your delay
Drawback: Short packets penalized as each queue sends one
packet per cycle

Lund University / Presentation 2012


FIFO and FQ

Lund University / Presentation 2012


Priority Queuing
321
12345678
54
K queues
876
1kK
Queue k+1 higher prio. than queue k
Higher prio. served first
Simple implementation
Low processing overhead
No fairness, low prio. queues can be starved

Lund University / Presentation 2012


Processor Sharing (PS)

Not practical but same principle adopted in


another scheme, BRFQ
Multiple queues as in FQ
Send one bit from each queue per round
Longer packets no longer get an advantage
Work out virtual start and finish time for a
given packet (of queue )

However, we wish to send packets, not bits


in reality

Lund University / Presentation 2012


Bit-Round Fair Queuing (BRFQ)

Based on PS
Each flow gets 1/nth of bandwidth (n flows)
Compute virtual start and finish time as in PS
When a packet finished, the next packet sent is the one with the
earliest virtual finish time
Good approximation to performance of PS
Throughput and delay of queues converge as time increases

Lund University / Presentation 2012


Generalised Processor Sharing
(GPS)
BRFQ cant provide different capacities to different
flows
Enhancement called weighted fair queuing (WFQ),
based on generalised PS
From PS, allocate weighting to each flow that
determines how many bits are sent during each round
If weighted 5, then 5 bits are sent per round

Gives means of responding to


different service levels => The
concept of service differentiation!

Lund University / Presentation 2012


Weighted Fair Queuing

Emulates GPS
Same strategy as BRFQ
Enables a router to assign weight to each flow and guarantee
bound on delay
Max buffer size needed proportional to defined max delay

Lund University / Presentation 2012


Example: BRFQ vs
WFQ
Received these packets all at about the Packet Size Flow

same time and same output link 1 100 1

Service with BRFQ: 2 100 1

3 60 2

Packet Size Flow Fi 4 120 2

5 60 2
1 100 1 100

2 100 1 200

3 60 2 60

4 120 2 180

5 60 2 240 Packet Size Flow Fi

1 100 1 50
pkt tx sequence: 3,1,4,2,5
2 100 1 100

Service with WFQ 3 60 2 20

Assume flow 2 gets 1.5 of flow 1 4 120 2 60


So, weight ratio 2 : 3
5 60 2 80

Sequence:
3,1,4,5,2

Lund University / Presentation 2012


CBQ
Root

Assigns fractions of BW X 100 % Y


to class nodes 40 %

Values minimum RT 60 %
NRT

Nodes can borrow 40 % 20 %

unused BW
Priority to flows within a
class

Lund University / Presentation 2012


Differentiated Services
(DiffServ)
IntServ is complex to deploy!
May not scale well for large volumes of traffic
Amount of control signals ( overhead)
Maintenance of state information at routers
Intserv has only two classes
DiffServ (RFC2475) designed to provide simple, easy to
implement, low overhead tool
simple functions in network core, relatively complex functions at edge
routers (or hosts)
Doesnt define service classes, provide functional components to build
service classes

Lund University / Presentation 2012


Characteristics of DiffServ
Use IPv4 header Type of Service or IPv6
Traffic Class field (called DS field)
So, no change to IP!
Service level agreement (SLA) established
between provider and customer prior to use of
DiffServ
All traffic with same DS field treated same
E.g. multiple voice connections

Lund University / Presentation 2012


DiffServ Architecture
Edge/Boundary
router: scheduling
- per-flow traffic
marking
management
..
- marks packets as in- .
profile and out-profile

Core/Interior
router:
- per class traffic
management
- buffering and scheduling
based on marking at edge
- preference given to
in-profile packets

J.F Kurose and K.W. Ross

Lund University / Presentation 2012


Edge-router Packet Marking
profile:pre-negotiated rate
packet marking at edge based on
per-flow profile
Possible usage of
marking:
class-based marking: packets of different
classes marked differently
intra-class marking: conforming portion of flow
marked differently than non-conforming one

Lund University / Presentation 2012


Classification and Conditioning

Packet is marked in the DS field


6 bits used for Differentiated Service Code Point (DSCP) and
determine PHB that the packet will receive
2 bits are currently unused

Lund University / Presentation 2012


Classification and Conditioning

Traffic conditioning to provide desired service


Classifier
Separate packets into classes
Meter/Police
Measure traffic for conformance to profile
Marker
Policing by remarking codepoints if required
Shaper
Dropper

Lund University / Presentation 2012


Classification and Conditioning

may be desirable to limit traffic injection rate of some


class:
user declares traffic profile (e.g., rate, burst size)
traffic metered, shaped or dropped if non-conforming

J.F Kurose and K.W. Ross

Lund University / Presentation 2012


Forwarding (PHB) I
Per Hop Behaviour results in a different
observable (measurable) forwarding
performance behaviour
PHB does not specify what mechanisms to
use to ensure required behavior
Examples:
Class A gets x% of outgoing link bandwidth over
time intervals
Class A packets leave first before packets from
class B

Lund University / Presentation 2012


Forwarding (PHB) II

Defined PHBs:
Expedited Forwarding: packet departure rate of a class equals or
exceeds specified rate
c.f. logical link with a minimum guaranteed rate
Assured Forwarding: 4 classes of traffic
each guaranteed minimum amount of bandwidth
each class with three drop preference partitions

Lund University / Presentation 2012


Summary
IntServ and DiffServ are QoS frameworks for IP internets
IntServ requires additional signalling protocol to reserve resource and
need to keep state per flow => not scalable
DiffServ works based on aggregate classes and has minimal impact
on the end-systems => scalable & a more popular alternative

Lund University / Presentation 2012

You might also like