evaluation of euqos system by simulation: tools and resultsaai.tele.pw.edu.pl/artikuly/evaluation of...

/27© EuQoS consortium University of Pisa/CPR

Evaluation of EuQoS System by Simulation: Tools and Results

Giovanni Stea, University of Pisa/CPRAndrzej Beben, Warsaw University of Technology


D2.1.2: Validation of the EuQoSsystem by simulation

• Editor: Wojciech Burakowski.• Contributors:

– Gerardo García de Blas, Maria L. García Osma, S. Pérez, R. Trueba (TID)– Jorge Sá Silva, Maxweel Carmo, Rómulo Ribeiro (UoC);– Giovanni Stea, Enzo Mingozzi, Claudio Cicconetti (UoPisa/CPR);– Gianfranco Santoro, Andrea Angelini, Marco Cavicchioni, Vito D’Eri,

Domenico Fiorentini, Riccardo Proietti (UoR/CRMPA);– Xavi Masip, Sergi Sánchez, Jordi Domingo (UPC);– Robert Janowski, Halina Tarasiuk, Jaroslaw Sliwinski, Wojciech

Burakowski, Damian Duda, Jordi Mongay Batalla, Andrzej Beben (WUT).


Simulation modelfor the EuQoS system

• Motivation:– validate & verify the EuQoS architecture;– provide guidelines based on simulation results.

• Simulation tools:– SiM-EuQoS-PTL: Packet transmission level;– SiM-EuQoS-CIL: Call invocation level;– SiM-EuQoS-PL: Provisioning level.


Overview of SiM-EuQoS-PTLPacket Transmission Level Simulator

• Access networks connected to the core through a single link

• No. of flows is fixed• Routing does not change

• Traffic distinguished into – Foreground, i.e. applications

which are actually measured– Background, i.e. aggregated

interference traffic

Hypotheses


Access networks

• Four different access networks are considered:

UMTS

xDSL

WiFi

Ethernet


Access Networks: UMTS

• Urban and rural scenarios considered– Different propagation and

mobility models

• Single Node B– Intercell interference accounted

for by using interference models

• DSCH and DCH implemented


Access Networks: WiFi

• Nodes using DCF in an 802.11 infrastructure mode network (802.11e left for the 2nd phase)

• RTS/CTS handshake• Fragmentation

• No mobility simulated• Channel error models to be

added• No power saving nor rate

switching


Access Networks: Ethernet

• Switched Ethernet:– Small Office Home Office

(SOHO)– Campus/LAN scenario– MAN organization scenario– MAN residential scenario

(EPON)

• 802.1p and 802.1q standards implemented


Access Networks: xDSL


Core Network

• DiffServ-capable routers with (at least) 2 PHBs

• QoS-oriented classification, policing and scheduling mechanisms enforced

• Interfering traffic represents aggregated traffic from access networks and core domains

• Bottlenecks are at the inter-domain links


Traffic models

Foreground traffic• VoIP

– G.711, G.729 codecs with VAD• Video Conference

– MPEG4 and H263 VC traces• Video streaming

– MPEG 4 encoded movies• Input expected from

application developers in EuQoS

Background traffic• Depends on the network

(access tech. and core)• Depends on what applications

are actually used• Many models available in the

literature• Input expected from

measurement activities in EuQoS


Implementation of the packet level simulator

• Ns-2 has been used• Devised a framework that allows the seamless addition

of new modules• Simulation scenarios can be built incrementally


First simulation results

• Objectives:– Validate the PTL simulator– Assess the level of QoS provided by the existing

“bare” access and core technology• What do we get without EuQoS• What do we need from EuQoS

– Test QoS in heterogeneous access networks• Do different QoS mechanisms interoperate correctly?

– “lay the ground” for simulating the effectiveness of EuQoS mechanisms at all relevant layers


Example: impact of non QoStraffic on QoS traffic

ATU-R

ATU-R DSLAM

DSLAM

Switch

Switch

BRAS

Background traffic node 1





User S

User T

User T

Server S

Server T

User SATU-R

ATU-R DSLAM

DSLAM

Switch

Switch

BRAS






User S

User T

User T

Server S

Server T

User S

Negligible where native QoSdifferentiation exists

Arbitrarily large where no QoS differentiation exists


SiM-EuQoS-PTL: xDSLResults

• The performance of CBR and VBR-nrt traffic depends on the amount of CBR traffic.

• Thus, AC algorithms for guaranteeing QoS are needed.

0,00%

0,20%

0,40%

0,60%

0,80%

1,00%

1,20%

0% 5% 10% 15% 20% 25% 30% 35% 40%

Demand / Capacity (%)

Pac

ket L

oss

Rat

io (%

)

0

50

100

150

200

250

300

0 20 40 60 80 100 120 140 160 180

File size (KB)

Thro

ughp

ut (K

B /

s)

20%30%40%Exponencial (20%)Exponencial (30%)Exponencial (40%)

Exponential (20%)Exponential (30%)Exponential (40%)


SiM-EuQoS-PTL: LAN

• SOHO scenario with full-duplex 10 Mbps links.• QoS is supported by means of strict priority queueing.• It is possible to provide EuQoS applications with full

isolation from background traffic.


SiM-EuQoS-PTL: WiFiDescription

• Two 802.11g @ 24 Mbps hot spots.• No specific QoS mechanism has been simulated.


SiM-EuQoS-PTL: WiFiResults

• Mixing QoS and non-QoS traffic leads to inacceptableperformance degradation, even with low traffic.

• Special AC algorithms are needed.

0.001

0.01

0.1

1

10

100

0 10 20 30 40 50 60

dela

y (s

), 95

per

cent

ile#flows

conf c dl perdlconf c ul perdlconf v dl perdlconf v ul perdl

0.001

0.01

0.1

1

0 10 20 30 40 50 60

dela

y (s

), 95

per

cent

ile

#flows

video dl perdl


Test specific QoS mechanisms: admission control in UMTS

• Suburban environment mobility model for UEs.• Node-Bs placed at 800m from each other.• The bottleneck for providing QoS in UMTS is the

UTRAN, for which special AC algorithms are needed.

Throughput

0

500

1000

1500

2000

2500

3000

3500

4000

0,20

1,80

3,40

5,20

6,80

8,40

10,2

0

11,8

0

13,4

0

15,2

0

16,8

0

18,4

0

20,2

0

21,8

0

23,4

0

25,2

0

26,8

0

28,4

0

30,2

0

31,8

0

33,4

0

35,2

0

36,8

0

38,4

0

time (sec)

Arrows CAC Interference CAC Joint Scheduler CAC


SiM-EuQoS-PTL: IP Core

• Admission control performed online per flow.

• Router configuration on provisioning (long) time scale.

• No per-flow policing.• AC algorithms based on

enhanced rules for calculatingadmissible traffic load are needed.

• It is recommended for end-to-end simulations to test telephony CoS and video conference CoS as separated.

Inter-domain service AS1-AS2

AS1

output port

Egress BR

AS2

Intra-domain service in AS2

input ports Router input port

Ingress BR

Ingress BR

Ingress BR


output port

AC AC

AC

AC

Inter-domain service AS1-AS2

AS1

output port

Egress BR

AS2


input ports Router input port

Ingress BR

Ingress BR

Ingress BR


output port

AC AC

AC

AC


SiM-EuQoS-CIL

• Implementation in the Network Simulator 2, based on the RSVP module.

• Current version:– UDP only;– simplified NSLP.


SiM-EuQoS-CIL: Example of Results

Exponential ON/OFFTraffic

VoIP Traffic

Without NSIS With NSIS

•SOHO scenario, with VoIP traffic.•NSIS agents configure 802.11p priorities along the path.•Advisable to carry NSIS messages asQoS traffic.


SiM-EuQoS-PL (1): Objectives

• To evaluate the performance of EQ-BGP (convergence time, scalability), under:

• announcement of previously unavailable network (route)

• withdrawing existing network (route)• link or node failure (reliability)• occurrence of route flapping • degradation or improvement of QoS level offered by a

given domain • iBGP processing


SiM-EuQoS-PL (2): Assumptions

• fixed values of QoS offered by ASs• different „QoS assembling” functions (delay losses, jitter)• different decision algorithms• different network topologies and size (4, 11, 20, 29, ASs)

Internet driven topology

ring

full mesh

4 5 6 7

B-clique


SiM-EuQoS-PL (3): Simulation tool

• NS2 with enhanced ns-BGP module:– QoS-NLRI attribute– new decision algorithm replacing AS path length by

„cumulative” value of QoS– an algorithm for assembling QoS (calculating

„cumulative” value of QoSadvertise start node 1 route 10.0.1.0/24 1.78075 send 1 -> 2 wds: - ads: 10.0.1.0/24 (1) 3 1.78075 send 1 -> 0 wds: - ads: 10.0.1.0/24 (1) 3 1.78252 rcv 2 <- 1 wds: - ads: 10.0.1.0/24 (1) 3 1.78252 send 2 -> 3 wds: - ads: 10.0.1.0/24 (2 1) 13 1.78252 rcv 0 <- 1 wds: - ads: 10.0.1.0/24 (1) 3 1.78252 send 0 -> 3 wds: - ads: 10.0.1.0/24 (0 1) 12 1.78431 rcv 3 <- 2 wds: - ads: 10.0.1.0/24 (2 1) 13 1.78431 send 3 -> 0 wds: - ads: 10.0.1.0/24 (3 2 1) 20 1.78431 rcv 3 <- 0 wds: - ads: 10.0.1.0/24 (0 1) 12 1.78431 send 3 -> 0 wds: 10.0.1.0/24 ads: -


SiM-EuQoS-PL (4): Summary and conclusions

• EQ-BGP gives a stable routing establishing the e2e QoS paths

• The convergence time is at the similar level as in case of BGP-4, however:– It is longer after route advertisement– It is shorter after route withdrawal

• EQ-BGP protocol needs to exchange a large number of messages

• Further studies are required!


Conclusions and future work

• For the time being:– The software tools (with full documentation) for

evaluating different aspects related to EuQoSarchitecture and QoS framework were developed.

– Preliminary simulation results were collected and guidelines for other WPs have been devised.

• Experimental limits to CAC algorithms evaluated

• Most exhaustive simulation studies will be performed, to be included in the next version of D2.1.2.– Take into account newly defined EuQoS CoS and

apply QoS mechanisms


Thanks for your attention.

Comments or questions?

evaluation of euqos system by simulation: tools and resultsaai.tele.pw.edu.pl/artikuly/evaluation of...

Documents