A-MAC: Adaptive Medium Access Control for Next Generation Wireless Terminals

R96725018 陳品宏R96725044 曾有德

Mehmet C. Vuran, Member, IEEE, and Ian F. Akyildiz, Fellow, IEEESchool of Electrical and Computer Engineering, Georgia Institute of

TechnologyIEEE/ACM TRANSACTIONS ON NETWORKING, VOL. 15, NO. 3, JUNE 2007

2

Outline

I. Introduction

II. Related Work

III. The Virtual Cube Concept

IV. Network Modeling

V. A-MAC: Adaptive Medium Access Control

VI. Performance Evaluation

VII. Conclusion

3

I. Introduction

4

Next Generation Wireless Networks

Next Generation (NG) wireless networks are envisioned to provide high bandwidth to mobile users via bandwidth aggregation over heterogeneous wireless architectures

The various types of MAC protocols: TDMA, CDMA, WCDMA and CSMA/CA

QoS classes: Conversational, streaming, interactive and background Fig. 1. Next generation wireless networks.

5

Challenges

Heterogeneity in Access Schemes: Different access schemes in different wireless networks

Heterogeneity in Resource Allocation: No unified metric for comparison of the allocated resources

Heterogeneity in QoS Requirements: The MAC layer must efficiently evaluate the available resources in different networks to satisfy the QoS requirements

6

The Structure and The Components of A-MAC

Two-layer Adaptive Medium Access Control (A-MAC)

Access sub-layer: Specialized for accessing the network

Master sub-layer: Perform decision and scheduling requests for the most efficient network

A novel Virtual Cube concept is used to model the networks

Fig. 2. Main Components of A-MAC.

7

Main Object

Incompatibility among medium access and resource allocation techniques are melted into a unified medium access control framework, providing self-contained decision flexibility as well as capability to access various networks.

8

II. Related Works

9

Related Works

An Ad-hoc CEllular NETwork (ACENET) TCDMA protocol Hybrid TD/CDMA systems A multiple access protocol of cellular Internet and

satellite-based networks The proposed methods requires either a modification in

the existing infrastructure and base stations or a completely new architecture

Integration problems: Implementation costs, scalability and backward compatibility

Assumption

NG wireless terminals are capable of receiving signals from multiple network access points and transmitting signals to different access schemes simultaneously.

10

11

III. The Virtual Cube Concept

A. Resource-Space

B. Virtual Cube Structure

12

Resource-Space

Time Dimension: The time required to transfer information

Rate Dimension: The data rate of the network

Power Dimension:

The energy consumed

for transmitting information

through the network

13

Virtual Cube Structure

Three parameters: Cube Capacity (M bits/cube) Cube Power (P Watts/cube) Cube Duration (T sec/cube)

Two types of virtual cubes are used in the A-MAC: Virtual Information Cube

(VIC): The information sent through the network

Virtual Power Cube (VPC): The additional power needed to transmit M bits of information

Fig.3. Virtual cube model.

M

VIC

VPC

14

IV. Network Modeling

A. TDMA Modeling

B. CDMA Modeling

C. CSMA Modeling

D. Multi-Rate Networks

15

TDMA Modeling

A TDMA slot is characterized by slot duration and transmission rate

Each interface is characterized by an average energy per bit The number of virtual cubes that can be filled in three dimensions

of a resource bin:

Fig. 4. (a) TDMA and CDMA modeling.

(1)

16

CDMA Modeling

Direct Sequence-CDMA (DS-CDMA) each bit of duration is coded into a pseudo-noise code of chips of

duration the spreading gain the transmitted energy per bit is the bandwidth , is the data rate and is the transmitted signal

power

TDD CDMA the time is divided into radio frames, slots and sub-frames is determined by the length of the allocated slot

FDD CDMA is determined by the duration of the connection

17

CSMA Modeling

IEEE 802.11 It is impossible to deterministically calculate the

transmission time in CSMA based systems Using the last transmission information to model the

resource bin

Fig. 4. (b) CSMA Modeling.

18

Multi-Rate Networks

Multi-code CDMA, e.g., IS-95-B and cdma2000

The number of virtual cubes in the rate dimension, ,of the CDMA frame model increases with the data rate

The power level, ,for a CDMA signal with a spreading gain , where is the power level of the fundamental spreading gain and is the fundamental spreading gain

Fig. 4. (c) Multi-code CDMA modeling.

19

Multi-Rate Networks (cont.)

Multi-channel communication, e.g., cellular networks and IEEE 802.11 standard

The number of virtual cubes in the rate dimension, , depends on many channel and transceiver dependent parameters such as the channel bandwidth, modulation and channel coding schemes

Fig. 4. (c) Multi-channel modeling.

20

V. A-MAC: Adaptive Medium Access Control

A. Decision

B. Scheduling

C. Adaptive Network Interfaces (ANIs)

21

The Structure and The Components of A-MAC

Fig. 2. Main Components of A-MAC.

SCHEDULING

3G GSMIEEE 802.11

DECISION

ADAPTIVE NETWORK INTERFACES

MASTER SUB-LAYER

ACCESS SUB-LAYER

NETWORK MODELS

22

Decision

A-MAC performs decision in each decision interval For a specific traffic flow , the decision block chooses the inte

rface such that the utility function

Aims to find the interface with maximum throughput capability for the minimum transmission power

(2)

23

Decision (cont.)

Constraints:

is the bandwidth share of the traffic in interface is the maximum power dissipation allowed for the decision inte

rval is the minimum required bandwidth in terms of VICs for the tra

ffic type in order to guarantee its QoS requirements is the set of active interfaces that are chosen by the decision block

(3)

(4)

24

Decision (cont.)

1) Single Type Traffic—Single Interface (STSI)

2) Single Type Traffic—Multiple Interfaces (STMI) If the (3) and (4) are met by one or more interfaces, the one that

maximizes (2) is chosen else…

the interfaces are sorted in a decreasing order of their utility functions

the constraint (4) becomes

A weight is associated with interface

(5)

(6)

25

Decision (cont.)

3) Multiple Type Traffic—Single Interface (MTSI) checks if the decision constraints hold for each traffic type each eligible traffic type is assigned a bandwidth share

such that checks if each flow satisfies (3) and (4)

the scheduler is informed of the bandwidth shares and MAC frame size of the interfaces

or the traffic type with the lowest priority is rejected and the bandwidth shares are updated

4) Multiple Type Traffic—Multiple Interfaces (MTMI)

26

Scheduling

In designing a scheduler, the requirements are considered:QoS GuaranteeChannel Dependent SchedulingDynamic BehaviorImplementation Complexity

Bin Sort Fair Queuing (BSFQ) schedulerFrame-based methoduses virtual time stamps to determine the scheduling orderBuilt-in buffer management component

BSFQ method

it

[ , + )it

NtNt

[ , + )

Fig. 1. Bin Sort Fair Queuing.27

virtual system clock

The output buffer is organized into N bins

Each bin is implicitly labeled with a virtual time interval and

each interval has length Δ

The current bin has the label

The i-th bin has the label

Δ

Packets from the current bin are transmitted, and updatesτ( t ) = t2

BSFQ method (cont.)

The output buffer is organized into N bins Each bin is implicitly labeled with a virtual time interval and

each interval has length Δ The current bin has the label

The ith bin has the labelOnly packets from the current bin are transmittedBin list is cyclic-list

The output link rate may change during the connection time due to wireless channel conditions

In order to prevent fluctuations in the bandwidth share of flows, the scheduler updates each bandwidth share in each decision interval

28

BSFQ method (cont.)

The jth packet of flow f is assigned with the virtual time stamp

is the arrival time of packet is the length of is the reserved data rate of f

29

BSFQ method (cont.)

The index of the bin used to store packet is equal to:

If = 0 then is stored in the current bin.If < N then is stored in the -th bin following the

current bin.If > N then the packet is discarded.

30

31

Adaptive Network Interfaces (ANIs)

1) Network Structure Awareness gathering information about the underlying network structures achieved in different wireless systems, i.e., GSM, UMTS, cdma2000,

and WLAN

2) Network Modeling ANIs model or update the network MAC structure (resource bin) and

inform the master sub-layer

3) Access and Communication An ANI performs access to the network if it is selected for a transmissi

on by the master sub-layer The communication with the AP is performed according to the network

specific procedures

32

VI. Performance Evaluation

A. Traffic Models

B. Network Models

C. Simulation Results

33

Traffic Models

Modeling based on a three-step Morkov model.

Modeling based on a three-step Morkov model.

Exponential distribution and statistically independent

Exponential distribution and statistically independentExponential distributionExponential distribution

The stations generate IP packets with length of 260 bits in every 20 seconds

The stations generate IP packets with length of 260 bits in every 20 secondsExponential distributionExponential distribution

Modeling with a multi-state model.

Modeling with a multi-state model.

Exponential distributionExponential distributionThe bit rate in each state is determined independently using a truncated exponential distribution

The bit rate in each state is determined independently using a truncated exponential distribution

The length of each packet is geometrically distributed and independent of each other

The length of each packet is geometrically distributed and independent of each other

34

Network Models

35

Simulation Results

In a 200 m × 200 m grid, nodes are placed with uniform distribution

Adaptive Node refer to the node equipped with A-MAC Analyze the performance of A-MAC with different number of

nodes and different percentage of traffic distribution in the heterogeneous network structure

Each simulation lasts 230 s and the results are average of 5 trials for each 5 random topologies

Two sets of simulations: Fixed Topology and Dynamic Topology

36

Fixed Topology

All nodes are stationary

The adaptive node is assumed to be in the coverage area of three of the network structures

The traffic type distribution (Voice, ABR, CBR, VBR) is chosen as (65%, 15%, 10%, 10%)

, and denote the percentage of nodes in TDMA, CDMA and WLAN networks

37

Fixed Topology (cont.)

Fig. 5. Average throughput of voice traffic with = (a) (10%, 10%, 80%), (b) (50%, 30%, 20%) .

38

Fixed Topology (cont.)

Fig. 5. Average throughput of CBR traffic with = (c) (10%, 10%, 80%), (d) (50%, 30%, 20%) .

39

Fixed Topology (cont.)

Fig. 6. Average throughput of best-effort traffic with = (a) (10%, 10%, 80%), (b) (50%, 30%, 20%).

40

Dynamic Topology

100 mobile nodes

The adaptive node roams through TDMA to the CDMA network, passing through the coverage area of WLAN APs

The simulation lasts 230 s

Averaging the throughput for intervals of 5

Fig. 7. Sample topology used in the mobility simulations. ●: TDMA BS; : ◆CDMA BS; ■: WLAN AP.

41

Dynamic Topology (cont.)

Fig. 8. (a) Throughput of voice traffic in dynamic topology.

42

Dynamic Topology (cont.)

Fig. 8. (b) Throughput of CBR traffic in dynamic topology.

43

Dynamic Topology (cont.)

Fig. 8. (c) Throughput of VBR traffic in dynamic topology.

44

Dynamic Topology (cont.)

Fig. 8. (d) Throughput of ABR traffic in dynamic topology.

45

VII. Conclusion

46

Conclusion

Adaptive medium access control (A-MAC) for NG wireless terminals Novel virtual cube model Network modeling QoS-based decision QoS-based scheduling

The first effort on designing a MAC protocol that achieves adaptation to multiple network structures with QoS aware procedures

Doesn’t require any modifications to the existing network structures

A cost function can be incorporated into the network modeling framework as a fourth dimension in the virtual cube concept such that the decision is performed accordingly

Thanks for your listening.