Network Capacity Enhancement of OFDMA System Using Self-organized Femtocell Off-load S. Akbarzadeh( 1 ,2 ), R.Combes (1 ,3 ), Z.Altman (1 ) 1 Orange

Labs

2 CNAM 3 University

of Paris VI

WCNC 2012

1 / 12

Scope ◮

Downlink of a wireless network

◮

Macro-cells + indoor femto-cells

◮

Elastic traffic, users arrive and depart dynamically

◮

What is the capacity gain compared with macro-only ?

◮

How to configure the pilot powers of femtos ?

◮

In dense networks, how to manage interference ?

2. Self-configuration of femto pilot power

3. Self-optimization of ICIC

Femtocell

1. eNB Pilot

Macrocell

2 / 12

The model

◮

Distance loss at distance d : A + B log(d)

◮

Indoor-to-outdoor / outdoor-to-indoor: A + B log(d) + C (penetration margin)

◮

All stations (macro+femto) interfere with each other

◮

Users are attached to the station with the strongest received pilot signal

3 / 12

The model (cont’d)

◮

Rs (r ) data rate of a user located at r served by s when all stations transmit at full power (worst case interference)

◮

Round-robin scheduling, user throughput

◮

Users arrive in the network according to a Poisson process of intensity λ

◮

Mean flow size E[σ], users leave after completing their transfer

Rs (r ) ns

4 / 12

Performance evaluation: one station

◮

Each station is modeled as M/G/1/PS (processor sharing) queue (BonaldMobicom2003)

◮

Traffic intensity I = λE[σ]

◮

Capacity of station s: Cs = (

◮

Load of station s: ρs =

◮

Blocking rate of station s : Bs =

◮

Average file transfer time: Ts =

R

1 −1 As Rs (r ) )

I Cs ρNmax 1+···+ρNmax ρs R1 λ A dr 1−ρs s

5 / 12

Performance evaluation: network

◮

Probability for a user to arrive in s: Ps =

◮

Network capacity of station s: minCs s P Network blocking rate: s Ps Bs P File transfer time: s Ps Ts

◮ ◮

R dr RAs A dr

6 / 12

Dynamic vs static evaluation

x

x

x x

x x x

x x x

x x

x x

x x

x

x x

x

x x x x x x x x x x x x x x x x x x x x x x x x x

Figure: Static

x

x

x

x

x x

x x

x

x

x

x x x

x x x x x x x x

x

Figure: Dynamic

7 / 12

Pilot auto-configuration for femto-cells

◮

Femtos provide densification gains

◮

Trade-off: stations absorb less traffic, but receive more interference

◮

Based on the received macro pilot, femtos are divided into “interior” and “exterior” femtos (cell center / cell edge)

◮

Inner femtos use a smaller power (Pint )than outer femtos (Pext )

◮

Off-line approach: store good values of (Pint , Pext ) in a data base

8 / 12

Pilot auto-configuration for femto-cells: results

1.25

2.8

Pint=0 mW

P =0 mW

Network capacity (bits/s per surface)

Network Capacity (bits/s per surface)

int

P =4 dBm

1.2

int

P =10 dBm int

Pint=16 dBm

1.15

Pint=20 dBm Pint=24 dBm

1.1

1.05

1

0.95 0(mW)

5

10 P

ext

15

20

(dBm)

Figure: Sparse femtos

24

P =4 dBm

2.6

int

P =10 dBm int

2.4

Pint=16 dBm

2.2

P =20 dBm int

P =24 dBm int

2 1.8 1.6 1.4 0 (mW)

5

10 15 P (dBm)

20

24

ext

Figure: Dense femtos

9 / 12

Self-organizing ICIC

◮

Femtos increase the capacity but can create outage for streaming and voice traffic (minimal bit-rate)

◮

Problematic for dense deployments

◮

Solution: split the spectrum in sub-bands and adapt the transmitted power on each sub-band

◮

Transmit power is adapted dynamically to maximize a P Ti1−α function of the users throughput (Mo2000ToN): i 1−α

◮

Search method: distributed projected gradient (StolyarInfocom2010, CombesICC2011)

10 / 12

Self-organizing ICIC: results

16

4 3.5

no ICIC FFR

15 14 Mean transfer time (s)

Block Call Rate (%)

3 2.5 2 1.5

13 12 11

1

10

0.5

9

0 11

no ICIC FFR

11.5

12 12.5 13 Arrival Rate (mobiles/s)

13.5

Figure: Block call rate

14

8 11

11.5

12 12.5 13 Arrival Rate (mobiles/s)

13.5

14

Figure: File transfer time

11 / 12

Thank you for your attention, any questions ?

12 / 12