End-to-End QoS Provisioning in UMTS networks

-Midterm Presentation

Group 995
Haibo Wang
Devendra Prasad
[email protected]

Supervisors:
Hans-Peter Schwefel
Oumer Teyeb

2005-2-22

1
Presentation outline
1. Motivation
2. Background
3. Problem Delimitation
4. Preliminary Simulation
5. Proposed QoS algorithms
6. Future work

2
1. Motivation
What is End2End QoS?
• Quality of Service
is a set of requirements to be met by the network while transporting
a traffic flow.

• Only the QoS perceived by end-user matters [1]

3
 End2End QoS Parameters
• Delay
• Jitter
• Loss Rate
• Throughput
These parameters reflect the traffic flow through the network.

QoS mechanisms
QoS provision mechanisms and QoS control mechanisms [2]:
• QoS provision mechanisms include parameters mapping, admission and res
ource reservations schemes.
• QoS control mechanisms consist of traffic shaping, scheduling, policing and
control.

4
Offered QoS through SLA

• QoS can be offered by network service provider in terms of Service Level Ag

reement (SLAs)
i) Network Availability
ii) Guaranteed bit-rate
iii) Payment model
iv) Other legal necessities

5
Why is QoS needed in UMTS?
UMTS Release 4
A main challenge for UMTS is t
o convey various types of traffi
cs on the same medium while
meeting their different QoS req
uirements, especially for real-ti
me applications.

UMTS Release 5

Note: UMTS core

networks evolution to all
IP backbone.

6
2. Background
The E2E UMTS QoS case we are investigating

Application
Servers

UEs UTRAN SGSN GGSN Internet

Two domain involved form end-2-End QoS point of view:

UMTS domain QoS and IP domain QoS

7
2.1 UMTS QoS Domain:
End2End QoS Architecture

UMTS

TE MT RAN CN CN TE
EDGE Gateway
NODE
End-to-End Service

TE/MT Local UMTS Bearer Service External Bearer

Bearer Service Service

Radio Access Bearer Service CN Bearer

Service

Radio Bearer RAN Access Backbone

Service Bearer Service Bearer Service

Physical Radio Physical

Bearer Service Bearer Service

• 3GPP layered structure [1]

8
UMTS QoS - UMTS QoS classes

Traffic class Conversational Streaming class Interactive class Background

class streaming RT Interactive best Background
conversational RT effort best effort

Fundamental -Preserve time -Preserve time -Request response -Destination is

characteristics relation (variation) relation (variation) pattern not expecting
between information between -Preserve payload the data within
entities of the information entities content a certain time
stream of the stream
-Preserve
Conversational payload
pattern (stringent content
and low delay )
Example of the -voice -streaming video -Web browsing -background
application download of
emails

9
UMTS QoS provisioning

• Existing UMTS QoS mechanisms:

- QoS profile
- Traffic Flow Template (TFT)
- PDP context

• What is missing in 3GPP standardization?

- UTRAN part: QoS parameters mapping to Radio Resource Mana
gement strategies
- UMTS core network part: IP layer transport mechanisms on the Gn
and Gi interface

10
 2.2 IP QoS
DiffServ and InterServ comparison
Integrated services Differentiated services
State in routers (e.g., scheduling Per-flow Per- Aggregate
buffer management)
Traffic classification basis Several header fields The DS field (6 bits) of the
IP header
Admission control Required Required for absolute
differentiation only
Signaling protocol Required (RSVP) Not required for relative
schemes; absolute schemes
need semi-static
reservations or broker
agents
Coordination for service End-to-end Local (per-hop)
differentiation
Scalability Limited by the number Limited by the number of cl
of flows asses of service

Network management Similar to circuit- Similar to existing IP

switched networks networks 11
3. Problem Delimitation
UE GGSN Remote Remote
AP Host

Local IP Bearer Service Remote IP Bearer

UE Host Layer
Remote
GGSN Backbone IP Access
Network Point
Access
SGSN Bearer
Layer
Gn/Gp (eg. UMTS
Bearer)
Scope of PDP Context

Network Architecture for QoS Conceptual Models

12
 Problem Statement
1. External IP network: IETF standard QoS mechanism – DiffServ or InterServ
2. UMTS network domain:
1) QoS classes mapping between External IP QoS classes and UMTS QoS classes, and

UMTS Bear Service (BS) attributes mapping to Core Network (CN) BS and Radio
Access Bearer (RAB) and further to Radio BS and Iu BS.
2) Call Admission Control in every multiplexing point (GGSN, SGSN, etc)
3) GGSN policing: check downlink flow and filter it according to TFT
4) Scheduling and congestion mechanisms in CN transmission (I,e, WFQ, WRED, etc)
5) QoS differentiation implementation in RRM
3) Other open issues: Application level End2End QoS provisioning, i.e, SIP signalling

13
 State of the Arts
1. Advanced Radio Resource Management for Wireless Services(ARROW) :
www.arrow-ist.es
This project aims at providing advanced Radio Resource Management
(RRM) and Quality of Service (QoS) management solutions, for both
UTRA-TDD and UTRA-FDD modes. It includes packet access,
Asymmetrical traffic and high bit rate (2 Mbit/s) services for multimedia IP
based applications.

Although ARROWS concentrates on the QoS aspects of UTRAN, a

global QoS framework is proposed, which is relevant with our work.

2. Advanced Services by Mastering UMTS (SAMU) :

www.samu.crm-paris.com
The innovative work in QoS from SAMU includes UMTS/IP QoS mapping
architecture and UMTS link layer optimization for TCP. The architecture
includes both mapping between IP DiffServ and UMTS bearer and mapping
from UMTS bearer to lower layer.

We refer to the mapping between IP DiffServ and UMTS Bearer from this
project.
14
3. Simulation of Enhanced UMTS Access and Core Networks (SEACORN) :
www.seacorn.ptinovacao.pt
Development, and implementation of resource management algorithms enabling QoS

provisioning and differentiation while optimizing resource efficiency

This project contribute a system level UMTS network simulator based on NS-2,
named Enhanced UMTS Radio Access Network Extensions (EURANE), which
We chose as the base of our simulator.

• Contributions and drawbacks of these related works:

a). Many E2E QoS frameworks were proposed but solid implementation and simulatio
n output are still missing.
b). QoS attributes mapping are designed but not implemented.
c). These work more focused on the RRM implementation, no core network QoS mec
hanisms are considered.
d). None of the framework gives a clear picture about the impact of mixing real-time
and non-realtime traffics on the QoS aspects.

15
Problem Delimitation
• Traffic: Mixed types of user data (real-time and non-real-time) flow from ex
ternal application servers to UMTS User Equipments (UEs). And we only foc
us on downlink traffic flow.

• For Internet/External network QoS -> IETF DiffServ approach was chosen.
Reason: more scalable and easy to manage when more than one network o
perator involved.

• In UMTS Core Network side:

1) Mapping mechanism from DiffServ QoS classes to UMTS QoS classes in G
GSN
2) Call Admission Control in GGSN considering required QoS profile and avai
lable equivalent bandwidth
3) Scheduling and queuing mechanisms to differentiate different UEs accord
ing to their UMTS service classes (Conversational, Streaming, Interactive, B
ackground)

16
 Simulation Tool – Network Simulator
version 2 (NS-2) and its UMTS extension
•No PDP Context Functionality.
•SGSN and GGSN are “Routers” from standard N
S-2.
•DiffServ functionalities are available
•Some buffer management algorithms are availabl
e for congestion control, i.e. Random Early Discar
d (RED)

17
4. Preliminary Simulation
Simulation Scenarios:
1. Network Topology

Appl. Server1
UE1
2Mb
10Mb

2Mb 10Mb 10Mb 1Mb 10Mb

UE2 Appl. Server2

Node B RNC SGSN GGSN 10Mb

2Mb

UE3 Appl. Server3

18
2. Traffic Model
• Constant Bit Rate (CBR) source for real-time applications
• Exponential traffic source for non-realtime applications
• All traffic is working on UDP protocol

3. Mobility Model
none

4. Propagation Model
Ideal,
Standard NS-2 error model to be set in future

5. QoS mechanism
Best Effort

19
 Simulation Result 1

User throughput when the total traffic load is lower then the
bandwidth in Core network (SGSN-GGSN)

20
 Simulation Result 2

Slightly overload Heavily overload

User throughput when the total traffic load is higher then the
bandwidth in Core network (SGSN-GGSN)

21
 Conclusion
In case of Best Effort transmission, the real-time traffics will
suffered from the mixing with non-realtime
Traffics in case of network congestion.

22
 5. Proposed End2End QoS algorithms

PDP Context generation through event based

Database (At the time of event

UE1
Generator and use it for CAC DS 1
Core Router
Edge Router
SGSN GGSN

UE1
UE2
CAC
Mapping
UE3

23
 PDP Context generation through event based

•Event based PDP generation,

•Acknowledgement based database for PDP Context
•Event based Traffic generation,
•1,2,3,4 is Available Bandwidth and 6 is an acknowledgement

Database: UE id,APN,BW Available,

1 2 3 4 5 6
UE Node B RNC SGSN GGSN

AS
Duration : 50ms

Acknowledgement Duration : 55ms

7
Total Duration : 60ms

24
 Mapping

•UMTS QoS Attributes are Guaranteed Bit Rate and SDU size.

•IP QoS Attributes are burst size and arrival rate stored in traffic profil
e of the SLA. [2]

•Differentiated Services considered.

•IPv4 header field TOS is used for DSCP.

•Assured Forwarding is considered for the traffic class buffering.

25
 Mapping
-continue

•Assumptions are IPv4, AF for PHB

Core Router: PHB
Scheduling,Buffering
Traffic Profile:
Burst Size,
Arrival Rtes
DS1

GGSN
Guaranteed Bit rate, SDU Size

26
 Call Admission Control
database

2 External Net

ER
4 3 1
SGSN GGSN

1. Incoming Traffic class,

2. Check with the Database for the Resource reservation,
3. Set priority based on the Available BW and Traffic class,
I.e Real-time: High priority, Non Real-time: Low Priority,
4. Scheduling, policy setting

27
 Conclusion
•The most of technical paper or EU funded project proposed the the concept
but none with the E2E aspect.
•As the emphasis is on the Multimedia Services , Hence the mapping between
the UMTS and the IP Network is the highest priority.
•The proposed algorithm will offset the limitation of an available simulator wit
h the implementation of Event based PDP Context and traffic generation.
•The purpose of the proposed topology is to design the algorithm which deals
the End to End QoS aspect including mapping,Scheduling and resource reserv
ation.
•The proposed algorithm will reduce the end2end delay and packet loss prob
ability

28
 Future Work

•Implementation of proposed topology and validation of proposed

algorithm.
•To show that the mix of non real time application makes the real
time application to suffer.
•To show the proposed queuing and scheduling technique will
improve the real time application When it mix with non real time
application.
•To show the Guaranteed bit rate per user can be achieved even
during the congestion .

29
References
[1] 3GPP TS23.107
[2] SAMU project, QOS Deliverable SP2-D1,
http://samu.crm-paris.com/, 2001.

30
Backup

Proposed Mapping in ASMU project [2]

31