FERMI: A Femtocell Resource Management System for Interference Mitigation in OFDMA Femtocell Networks

Mustafa Y. Arslan Jongwon Yoon Karthikeyan Sundaresan UC Riverside U Wisconsin Madison NEC Laboratories America Inc.

Srikanth V. Krishnamurthy Suman Banerjee UC Riverside U Wisconsin Madison

ACM Mobicom 2011

Femtocells

• Small cellular base stations deployed indoors.

✓ Use existing cable backhaul and cellular access technology

✓ Short range, high throughput✓ Clients save power on the uplink

• Interference is inevitable among collocated femtocells.

✓ different problem than interference in WiFi

•What can we do?

Contributions

• FERMI - mitigates interference among OFDMA femtocells deployed in an enterprise.

✓ Centralized algorithms to assign orthogonal frequencies to interfering femtocells.

✓ F l e x i b l e f r a m e f o r m a t t h a t s u p p o r t s heterogeneous client requirements for better spatial reuse.

• First solution implemented on an actual OFDMA femtocell testbed with off-the-shelf clients!

Roadmap

• WiMAX preliminaries

• Interference among femtocells✓ Can we leverage existing WiFi solutions?✓ If not, how should the solution look like?

• Algorithms for interference management

• Evaluation

WiMAX Preliminaries

• Multiple users scheduled in the same frame.

• BS schedules tiles for both downlink and uplink.

• Synchronous MAC (no carrier sensing).✓ frames sent every 5 ms (1 ms for LTE)

PREA

MBL

E

Sub-

chan

nels

FCH

DL-M

AP

UL-M

AP

User 1 DL Burst

User 2 DL Burst

User 3 DL Burst

Symbol Duration

User 1 UL Burst

User 2 UL Burst

User 3 UL Burst

Transition Gap

MCS 1 : QPSK 3/4 MCS 2 : 16QAM 1/2 MCS 3 : 16QAM 3/4 MCS 4 : 64QAM 1/2 MCS 5 : 64QAM 2/3 MCS 6 : 64QAM 3/4 MCS 7 : 64QAM 5/6

DOWNLINK UPLINK

Tile

OFDMA vs OFDM

• WiMAX uses OFDMA technology at the PHY.

WiFi (OFDM)

Channel

WiMAX (OFDMA)

Sub-channels

0

1

2

3

Roadmap

• WiMAX preliminaries

• Interference among femtocells✓ Can we leverage existing solutions?✓ If not, how should the solution look like?

• Algorithms for resource management

• Evaluation

Existing Solutions for WiFi

• Tune interfering WiFi APs to orthogonal channels.

• Licensed spectrum

• Orthogonal sub-channels to interfering femtocells.

• Under-utilization for clients who are not subject to interference.

• Multiple clients should coexist.

Equivalent Solution for Femtocells0

1

2

3

How do we define interference?

• Degradation of decoding at the clients (need isolation).

Su

b-c

han

ne

ls

Time

How do we define interference?

• Degradation of decoding at the clients (need isolation).

Su

b-c

han

ne

ls

Time

How do we define interference?

• Degradation of decoding at the clients (need isolation).

Su

b-c

han

ne

ls

Time

• GOAL: Intelligent resource management to improve network utilization (taking into account both clients.)

What should the solution look like?

REUSEZONE

ISOLATIONZONE

Su

b-c

han

ne

ls

Time (Symbols)

USED BY OTHER CELLS

What should the solution look like?

REUSEZONE

ISOLATIONZONE

Su

b-c

han

ne

ls

Time (Symbols)

USED BY OTHER CELLS

What should the solution look like?

REUSEZONE

ISOLATIONZONE

Su

b-c

han

ne

ls

Time (Symbols)

USED BY OTHER CELLS

What should the solution look like?

REUSEZONE

ISOLATIONZONE

Su

b-c

han

ne

ls

Time (Symbols)

USED BY OTHER CELLS

✓ Load-based adjustment of zones.

Roadmap

• WiMAX preliminaries

• Interference among femtocells✓ Can we leverage existing solutions?✓ If not, how should the solution look like?

• Algorithms for resource management

• Evaluation

Algorithms (Overview)

REUSEISOLATION

REUSEISOLATION

Algorithms (Overview)

REUSEISOLATION

REUSEISOLATION

allocate &assign (coloring)

Algorithms (Overview)

REUSEISOLATION

REUSEISOLATION

determine commonreuse zone size

Sub-channel Allocation

• Weighted max-min fair allocation

• Need to list all maximal cliques: NP-hard

Sub-channel Allocation

• Weighted max-min fair allocation

• Need to list all maximal cliques: NP-hard

D

C

G

E

FA

B

10

10

10 10

10 20

20

30 sub-channelswith equal load

Sub-channel Allocation

• Chordal graphs: no cycles of more than 3.

• Triangulation: transform general graph G to a chordal graph G`

• All maximal cliques can be listed in polynomial time!

Sub-channel Allocation

• Chordal graphs: no cycles of more than 3.

• Triangulation: transform general graph G to a chordal graph G`

• All maximal cliques can be listed in polynomial time!

D

C

G

E

FA

B

1010 10

1010

20

20

Sub-channel Allocation

• Chordal graphs: no cycles of more than 3.

• Triangulation: transform general graph G to a chordal graph G`

• All maximal cliques can be listed in polynomial time!

D

C

G

E

FA

B

1010 10

1010

20

20

Sub-channel Allocation

• Chordal graphs: no cycles of more than 3.

• Triangulation: transform general graph G to a chordal graph G`

• All maximal cliques can be listed in polynomial time!

D

C

G

E

FA

B

1010 10

1010

20

20

Sub-channel Allocation

• Chordal graphs: no cycles of more than 3.

• Triangulation: transform general graph G to a chordal graph G`

• All maximal cliques can be listed in polynomial time!

D

C

G

E

FA

B

1010 10

1010

20

20

10

Sub-channel Assignment

• Coloring with multiple colors (sub-channels).

• Construct a clique tree for chordal graph G`

Sub-channel Assignment

• Coloring with multiple colors (sub-channels).

• Construct a clique tree for chordal graph G`

D

C

G

E

FA

B

Chordal graph

Sub-channel Assignment

• Coloring with multiple colors (sub-channels).

• Construct a clique tree for chordal graph G`

D

C

G

E

FA

B

CBD

BED ACB CFD

GF

Chordal graph Clique tree

Sub-channel Assignment

• Color each level starting from the root.

Sub-channel Assignment

CBD

BED ACB CFD

GF

• Color each level starting from the root.

Sub-channel Assignment

• Color each level starting from the root.

CBD

E A F

GF

Sub-channel Assignment

• Color each level starting from the root.

CBD

BED ACB CFD

G

Sub-channel Assignment

• Color each level starting from the root.

CBD

BED ACB CFD

G

•FERMI guarantees a feasible coloring!

Zoning

• Common reuse zone size: min or max?

REUSEISOLATION

REUSEISOLATION

REUSEISOLATION

Zoning

• Common reuse zone size: min or max?

REUSEISOLATION

REUSEISOLATION

REUSEISOLATION

Zoning

• Common reuse zone size: min or max?

REUSEISOLATION

REUSEISOLATION

REUSEISOLATION

Zoning

• Common reuse zone size: min or max?

REUSEISOLATION

REUSEISOLATION

REUSEISOLATION

AVOIDCASCADES!

Zoning (avoiding cascades)

51510

BS 1 BS 3BS 2

Zoning (avoiding cascades)

51510

BS 1 BS 3BS 2

5

Zoning (avoiding cascades)

51510

5 15

BS 1 BS 3BS 2

5

Reuse clients(using isolated sub-channels)

Zoning (avoiding cascades)

51510

5 15

BS 1 BS 3BS 2

510

Reuse clients(using isolated sub-channels)

Cascade avoided since no interference to BS2’s clients

Roadmap

• WiMAX preliminaries

• Interference among femtocells✓ Can we leverage existing solutions?✓ If not, how should the solution look like?

• Algorithms for resource management

• Evaluation

Evaluation

• Zoning provides around 50% throughput gain over pure sub-channel isolation.

0

10

20

30

40

1 2 3 4 5

Thr

ough

put

(Mbp

s)

Topology

BaselineFreq. IsolationFreq. Isolation + Zoning

Evaluation

• Avoiding cascades provides 30% gain over cascaded zoning.

100

125

150

175

200

225

250

0.25 0.33 0.5 0.66 0.75

Thr

ough

put

(Mbp

s)

Reuse Load in the Network

Baseline FERMICascadedWithout cascade

Conclusion

• FERMI mitigates interference among femtocells in an enterprise. The distinguishing aspects are:

✓ Identify tolerance of clients to interference.

✓ Flexible Frame structure to support the graceful coexistence of clients (reuse and isolation).

✓ Novel use of chordal graphs to achieve near optimal allocation and feasible assignment.

✓ Intelligent zoning to mitigate interference and leverage reuse at the same time.

✓ Implemented, evaluated on a WiMAX testbed (concepts applicable to LTE as well.).

Thank you!

• Questions?