BROADWAY, IST-2001-32686

WP2-D17

Date 31-01-2005 Programme: IST

IST-2001-32686 BROADWAY

WP 2, D17 – DLC- CS layer and MAC/PHY final implementation and integration of key components

Contractual Date of Delivery to the CEC: 31/12/2004 Actual Date of Delivery to the CEC: 31/01/2005 Work Package Leader: Intracom Editors: Athanasios Vaios, Nikolaos Zinelis, Konstantinos Ntagkounakis, Konstantinos Oikonomou, Pietro Pellati, Gregory Yovanof, Sebastien Simoens Participating partners: Intracom, Motorola, UoA Security: PU Nature: Deliverable Pages: 101

Abstract:

The scope of this document is to provide the description, evaluation and validation of the integrated parts of the BROADWAY system and their functionalities. This consists of several components like the DLC and CS layer that combined and interfaced with the Physical layer provide for the fine operation of a dual 5/60 GHz HIPERLAN/2 system with all the required enhanced functionalities. The document focuses on the NS2 implementation of the proposed architecture and the evaluation through simulations of the developed code related to its specification. Extensions of the specified work concerning security aspects during the Neighborhood Discovery process and connectivity studies in BROADWAY environment are also included.

Keyword list: Implementation, Simulations, NS-2, HiperLAN/2, Ad-hoc, Neighborhood Discovery, Routing and Clustering, Security.

BROADWAY, IST-2001-32686

WP2-D17

Date 31-01-2005

Programme: IST

Contents Abbreviations .....................................................................................................................................................4 Scope ..................................................................................................................................................................6 1 The Physical Layer Abstraction .................................................................................................................7 1.1 The Propagation Modeling .................................................................................................................7 1.1.1 Path Loss ....................................................................................................................................7 1.1.2 Shadowing ..................................................................................................................................9 1.1.3 Fast Fading ...............................................................................................................................10 1.2 The Interference Modeling ...............................................................................................................12 1.3 Validation of the Physical layer abstraction .....................................................................................15 2 Dual Medium Access Control (MAC) Layer (5/60 GHz) ........................................................................18 2.1 Scheduling the Dual Frame ..............................................................................................................18 2.2 Allocating the Resources ..................................................................................................................19 2.3 Scheduling Traffic in Dual MAC .....................................................................................................21 2.4 Interfaces with Physical Layer - BWSSCS ......................................................................................21 2.5 Neighbourhood Discovery (ND) ......................................................................................................22 2.5.1 Phase 1: Triggering ND............................................................................................................23 2.5.2 Phase 2: Discovering one-hop away neighbors ........................................................................24 2.5.2.1 Exchanging Hello Messages (NCH).....................................................................................25 2.5.2.2 Link State Tables ..................................................................................................................26 2.5.3 Phase 3: Exiting ND .................................................................................................................26 2.5.4 Evaluating ND Performance.....................................................................................................26 3 BroadWay Service Specific Convergence Sublayer.................................................................................31 3.1 Overview of BWSSCS Structure and Functionality.........................................................................31 3.2 BroadWay Simulator Protocol Stack................................................................................................32 3.3 BWSSCS Class in the BroadWay Simulator....................................................................................32 3.3.1 Variables...................................................................................................................................33 3.3.2 Data Tables ...............................................................................................................................33 3.4 Interfaces and Management Functions .............................................................................................36 3.4.1 Interface with MAC-Control Plane (DLC)...............................................................................36 3.4.2 Peer-to-Peer Message Exchange ..............................................................................................37 3.4.2.1 Peer-to-Peer Messages..........................................................................................................37 3.4.2.2 Internal PtP Message Queue.................................................................................................39 3.4.2.3 Message Handler ..................................................................................................................39 3.4.3 Interface with ESSCS - User Plane (MFL)...............................................................................41 3.5 Algorithms........................................................................................................................................41 3.5.1 Neighborhood Discovery Initiator Algorithm ..........................................................................41 3.5.2 BroadWay Routing Algorithm .................................................................................................42 3.5.3 Practical Aspects of NDI, BWR Algorithms............................................................................44 3.5.4 Functionality Overview State-Diagram....................................................................................45 4 Performance of CANA (based on the integrated NS-2 code)...................................................................46 4.1 Parameters used and quantities measured.........................................................................................46 4.2 Multiple MTs....................................................................................................................................49 4.3 Offloading Capability – Throughput Increase..................................................................................50 4.4 Repetition Frequency of ND Execution ...........................................................................................52 4.5 Induced overhead..............................................................................................................................53 4.6 Mobility ............................................................................................................................................55 5 Security and Cooperation Reinforcement in the CANA Architecture .....................................................58 5.1 HIPERLAN/2 Security .....................................................................................................................58 5.2 CANA Specific Security Requirements ...........................................................................................59 5.3 Reinforcing Cooperation during the ND execution..........................................................................61 5.3.1 The general framework.............................................................................................................61 5.3.2 The need for a reputation mechanism.......................................................................................62 5.3.3 The reputation metric as a random walk process......................................................................63 5.3.4 Measuring empirical frequencies of suspicious events.............................................................64 5.3.5 Dealing with time dependent suspicious events .......................................................................67 2

BROADWAY, IST-2001-32686

WP2-D17

Date 31-01-2005

Programme: IST

Reinforcing cooperation during the Data Transmission Phase.........................................................70 5.4 5.4.1 Detecting misbehaving nodes in various types of networks.....................................................70 5.4.2 End-to-end authenticated acknowledgments in CANA............................................................72 5.4.3 Employing reputation ratings to reinforce cooperation ............................................................76 APPENDIX I: Implementation Details on Security Enhancement ..................................................................78 APPENDIX II: Implemented NS-2 Files and Functions/Handlers for the CANA MAC Operation................80 APPENDIX III: Minimizing Power Consumption based on Topology and Load Constraints ........................87 Description of the Scenarios.........................................................................................................................87 Study of Communication Distances .............................................................................................................88 Studying more Topology Metrics.................................................................................................................89 Constructed Graphs ......................................................................................................................................91 Studying load distribution ............................................................................................................................92 Studying Interference ...................................................................................................................................93 Timeslot Resource Allocation (TDMA Structure) ...........................................................................................95 References ........................................................................................................................................................97 Glossary..........................................................................................................................................................100

3

BROADWAY, IST-2001-32686

WP2-D17

Date 31-01-2005

Abbreviations ACH

Association CHannel

AP

Access Point

AsF

Association Frequency

BCH

Broadcast CHannel

BWCL

BroadWay Convergence Layer

BWDLC

BroadWay Data Link Control Layer

BWNA

BroadWay Network Architecture

BWR

BroadWay Routing

BWSSCS

BroadWay Service Specific Convergence Sublayer

CANA

Centralized Ad-Hoc Network Architecture

CH

Cluster Head

CL

Convergence Layer

CP

Control Plane

CPCS

Common Part Convergence Sublayer

DL

Down Link

DLC

Data Link Control

ESSCS

Ethernet SSCS

FCH

Frame CHannel

FIU

Flow Information Update

FN

Forwarder Node

HIPERLAN

HIgh PErformance Radio Local Area Network

HL/2

HiperLAN Type 2

ID

Identity

LCH

Long transport Channel

MAC

Medium Access Control

MCH

Multi-hop Cellular Network

MFQ

Monitor Flow Queues

MH

Message Handler

MT

Mobile Terminal

NBCH

Neighbourhood discovery Broadcast CHannel

NCH

Neighbourhood discovery CHannel

ND

Neighborhood Discovery

NDCH

Neighbourhood Discovery CHannel

NDF

Neighborhood Discovery Frequency

NDI

Neighborhood Discovery Initiator

NDM

Neighborhood Discovery Message

NDT

Neighborhood Discovery Table 4

Programme: IST

BROADWAY, IST-2001-32686

WP2-D17

Date 31-01-2005

NFCH

Neighbourhood discovery Feedback CHannel

NS-2

Network Simulator 2

PDU

Protocol Data Unit

PHY

PHYsical layer

RCH

Random CHannel

RN

Resource Needs

RoF

Routing Frequency

RSS

Received Signal Strength

SAR

Segmentation, Re-assembly

SCD

Scheduler

SSCS

Service Specific Convergence Sublayer

UL

Up Link

WLAN

Wireless Local Area Network

WPAN

Wireless Personal Area Network

5

Programme: IST

BROADWAY, IST-2001-32686

WP2-D17

Date 31-01-2005

Programme: IST

Scope This document describes and illustrates the results of the integrated BROADWAY system composed of the software modules for the Data Link Control and Convergence Sublayer architecture implementation (also by taking advantage of the enhanced capabilities of the Physical layer at 60 GHz) as specified in previous documents. The software platform used for the implementation of the proposed architecture is NS-2 [1]. In the already implemented modules regarding the traditional operation of the 5GHz HL/2, [2], [4], [6], modifications and enhancements have been made to cater to the BROADWAY requirements [3], [5], [7], [8], [9], [10], [11], [12]. In order for the BROADWAY system (or else the term CANA will be used in this document to indicate the architecture proposed for the BROADWAY system) to accommodate the enhanced new features, several parts (e.g. radio link management function, dual frame operation, scheduling of resources at both frequencies 5/60GHz) have been developed in software. The document sheds light on the basic ideas behind the implemented modules and the way they function to provide the desired processes. Furthermore, it provides the results of the validated and tested integrated system in terms of the enhanced capabilities of CANA to cater to greater data rate demands and increased total throughput. The document is organized as follows: Section 1 presents the key aspects of the Physical layer that interfaces with the DLC layer and is closely related to the performance of both the MAC and the BWSSCS modules. Section 2 describes the main NS-2 modules that have been implemented in the MAC, including the ND algorithm that is separately evaluated, to cater to the requirements of the BROADWAY system. Section 3 provides a thorough description of the basic functionalities of BWSSCS. Section 4 illustrates the results of the testing phase for the integrated system showing the capabilities of CANA in increasing the total throughput with no prohibitive additional overhead cost. In Section 5, research results on security aspects concerning the proposed system are discussed. In the Appendices, further details of the code itself and extensions of the work are included.

6

BROADWAY, IST-2001-32686

WP2-D17

Date 31-01-2005

Programme: IST

1 The Physical Layer Abstraction In order to test and validate the architecture designed at the DLC layer for the BROADWAY system, it has been decided to adopt a common software platform allowing to perform simulations at system level. The adoption of such a tool required, on one hand, the accurate implementation of the algorithms conceived and, on the other hand, the application of a fine abstraction of the physical layer. The framework on which the simulation platform has been based, is made up by the well known Network Simulator version 2 (NS-2), but its original distribution presented an evident lack of accuracy in the addressing the physical layer modeling and one of the goal of the project was to fill this gap for both the modes of operation of the BROADWAY system. In the following, the enhancements implemented for the physical layer abstraction in the simulator both at 5 and 60GHz will be briefly described. The modification of the code required a careful analysis of each file involved in the physical layer modeling and we provide hereafter a detailed description of the variables and methods that have been introduced with respect to the characteristics of the physical layer modeling that they represent. The main difficulty that the BROADWAY raised was the design for each node of a double interface to be used in accordance with the mode adopted by the nodes to transmit/receive a packet. This meant that, along with the modifications needed to model the physical phenomena that affect the transmission in a wireless system, each packet needed to be “stamped” with a value indicating the frequency at which it had been transmitted and that the node had to call the methods concerning the physical interface actually switched on. Practically, this constraint was brought into effect by the transformation of many simple variables into array (of variables) to take into account the interface they were referred to. The physical layer abstraction concerns basically three phenomena that affect the signal through its transmission from a given transmitter to a given receiver: the loss of power due to the signal’s path (path loss) and fading. The fading is the change in signal strength due to direct and reflected signals (multi-path) interfering with each other (fast fading) and due to distance and terrain effects (slow fading or shadowing). Besides, in order to provide as much as possible reliable simulation results, the impact of the interference had to be taken into account. In the currently wireless systems accessing the medium with a TDD-TDMA scheme the signal strength is affected by the interference especially in multi-cell scenarios where more than one AP operates at several different channels, whereas in the BROADWAY system this problem arises as well within one cell due to the clustering architecture for the 60GHz mode of operation. Thus, the interference modeling has been extended for that mode.

1.1 The Propagation Modeling 1.1.1 Path Loss It is well known that the path loss is the attenuation that a signal suffers along the path from a transmitter to a receiver. The simplest model is the free-space one that is described by the Friis’ law:

⎛ λ ⎞ Pr = P ⋅ Gt ⋅ G r ⋅ ⎜ ⎟ ⎝ 4πd ⎠

2

A combination of this model with the Two Ray Ground one was chosen for the implementation of the original NS-2, but it clearly cannot be assumed as suitable for the types of environments encountered in WLANs deployment. In order to fill this gap, the more realistic Log-Distance path loss model has been implemented in the simulator. It assumes that the average signal power decreases with distance raised to some exponent n, whose value depends on the environments. The general formula that applies for the path loss (PL) is the following:

7

BROADWAY, IST-2001-32686

WP2-D17

Date 31-01-2005

⎛ d PL(d )[dB] = PL0 (d 0 ) + 10 ⋅ n ⋅ log⎜⎜ ⎝ d0

Programme: IST

⎞ ⎟⎟ ⎠

Where n is the path loss exponent that indicates the rate at which the path loss increases with distance. d0 is the free space close-in reference distance and PL0(d0) is the free space path loss at the distance d0 from the transmitter calculated with the Friis equation. Then for d0=1m, Gt=1dBi, Gr=1dBi and f=5.25GHz, the path loss is PL0(d0)=44.85dB in the 5GHz band, whereas in the 60GHz one the same expression gives PL0(d0)=69.52 dB. In the special case where n=2, the model amounts to free-space propagation. As observed in Table 1, the path loss exponent typically varies between 2 and 4. In a large open space with Line Of Sight conditions, n≈2 is a good choice, but in a home environment with NLOS conditions n=3 or higher is likely. Name

Path loss Exponent

A-LOS-IF-Iso

2.2

A-NLOS-IF-Iso

3.6

Ab-LOS-IF-Iso

3.2

Ab-NLOS-1F-Iso

4.6

Ab-NLOS-1F-Dense

3

Ab-NLOS-MF-Iso

6.5

Ab-NLOS-MF-Dense 3.7 C-LOS-1F-Iso

2.05

C-NLOS-1F-Iso

3.3

Table 1: Path loss exponents for the typical environments implemented in the simulator

This model has been implemented in the simulator as a derivate class of the mother class Propagation that is the original class employed. The class is named LogDistance and an instantiation of it is contained in the class WirelessPhy that represents the lowest layer of the protocol stack for a wireless node. Class WirelessPhy : public Phy { … Propagation *propagation_ … } By means of the previous pointer the Log Distance model is applied for the evaluation of the path loss. In fact, at the reception of a packet the method sendUp of the class WirelessPhy is called and the method Pr of the class LogDistance is executed. In the files logdistance.h,cc the class and the method that apply for the Log-Distance path loss model have been implemented. In file logdistance.h, the static "factor" array is suppressed and the notion of scenario is removed. Instead, an array of propagation exponents log_dist_exponents of size nb_phy_interfaces is created in C++ along with a TCL command "set-logdist-exponents" which creates and initializes the log_dist_exponents array for each node as well as for each interface. To lighten the complexity of the implementation, it has been decided to hard code the number of possible interfaces in the LogDistance class: [ $ns_ set propInstance_ ] set-logdist-exponents $interface_index $value $value This command is called in the TCL script scenario, assuming that some parameters may change from one simulation set to another. In the above expression can be noticed that two values must be specified. This, because it has been assumed that in each scenario a break distance exists from which is likely to consider the MTs in NLOS conditions with respect to the AP. It means that two different propagation conditions are considered per scenarios and the consequent exponents must be set. Further, the break distance has to be specified for each interface: 8

BROADWAY, IST-2001-32686

WP2-D17

Date 31-01-2005

Programme: IST

[ $ns_ set propInstance_ ] set-dbreak $interface_index $topo(r60GHz) In logdistance.cc, the method Pr(PacketStamp *t, PacketStamp *r, int) was modified to use the correct value of the propagation exponent with respect to the mutual distance between the nodes. The characteristics of the propagation at 60GHz outlined in the D2 (Error! Reference source not found.) make the LogDistance model not fully appropriated. Aiming to evaluate the systems in a typical indoor environment, the model that would have fit better is the Keenan-Motley’s one. It suggests that the average path loss can be estimated from the free-space path loss PL0(d) and from the number n of walls and floors between MT and AP: NW

NF

i =1

j =1

PL(d ) = PL0 (d ) + ∑ LWi + ∑ LFj where LWi is the attenuation of wall i and LFj is the attenuation of floor j. Yet, the above model necessitates of simulation scenarios quite complex to prove its effectiveness as well as considerably large simulation resources in terms of both time and computation power. In fact, being based on a detailed description of the environments surrounding the MTs and on their reciprocal position, it would have required the implementation of structures that are memory consuming and that would have slowed the results acquisition and analysis. Moreover, since the attenuation of the 60 GHz signal by a wall is of the order of 80 dB, it results that the simulation of dual-band 5/60 connection only makes sense inside a room. Therefore, we decided to keep the LogDistance model for the 60GHz mode of operation, and use a trick which consists in using a very high (e.g. 1000) log-distance path-loss exponent at 60 GHz when the signal crosses a wall.

1.1.2 Shadowing The above path-loss models only give the average path loss, however, two different locations that have the same distance to the base station could receive the same signal at quite different strength. This is due to random effects caused by obstacles in the propagation path between the emitter and the receiver. They are typically dependent on the specific topography of the surrounding environment. Generally, the shadowing is assumed to have a log-normal distribution. Therefore, the power loss in dB is affected by a χσ factor which is a Gaussian distributed random variable with zero mean and standard deviation σ. The variance depends naturally on the specific propagation environment. Actually, the current simulator’s implementation adds a log-normal random variable each time a packet is received. More precisely, for a given set of simulations, a value of standard deviation is chosen depending on the environment we are in. Then, for each of the simulations, a value of shadowing (χ) is drawn: this number remains constant throughout the simulation and will be removed (in dBm) from the received power of a packet. The variance depends naturally on the specific propagation environment. As for the pathloss exponent, a set of realistic values have been drawn. Name

Standard deviation (dB)

A-LOS-IF-Iso

6

A-NLOS-IF-Iso

5

Ab-LOS-IF-Iso

6

Ab-NLOS-1F-Iso

5

Ab-NLOS-1F-Dense

3

Ab-NLOS-MF-Iso

6

Ab-NLOS-MF-Dense 4 C-LOS-1F-Iso

7 9

BROADWAY, IST-2001-32686

WP2-D17

Date 31-01-2005

C-NLOS-1F-Iso

Programme: IST

6

Table 2: Shadowing standard deviation for possible simulation scenarios

Previously, in the class Shadow the shadowing values were stored in a symmetric matrix of range N×N where N is the number of nodes involved in the simulation. Each entry of the matrix was evaluated at the beginning of the simulation accordingly to a value of σ dependent on the environment. In BROADWAY, it was necessary to add another dimension, which is the number of physical interfaces. More precisely, in file shadow.h, the static variable sigma and the static array possible_sigma were deleted. Instead, an array shad_val and an array of sigma values sigma_values of size nb_phy_interfaces is created in C++ along with a TCL command "set-shad-sigma" which creates and initializes the sigma_values array for each node: $node set-shad-sigma $interface_index $value This command is called just after the node creation. Then in ns-mobilenode.tcl, this command is translated to a command to the wireless-phy. The command in wireless-phy.cc allocates space equal to nb_phy_interfaces if the space had not yet been allocated and invokes shad->set_sigma(interface_index,atof(argv[3])). The shadowing may be positive or negative, but overall it will degrade the performance of the system since more trains will be detected as erroneous. And naturally, the larger the standard deviation, the worse the performance. When running the simulations, it should not be forgotten to take a number of draws in the Gaussian distribution of the shadowing values large enough to cover the whole distribution. The experience shows that a number of draws taken in the range [300 500] is reasonable. The 60GHz mode of operation of the system will affect basically the value of the standard deviation σ. Practically, a second matrix of shadowing values associated to the 60GHz mode has been implemented and, for each packet at 60GHz, the shadowing value will be taken from it.

1.1.3 Fast Fading The detailed description of the fast fading modeling results to be quite complex and somehow out of the scope of this document where mainly the implementation of the adopted expression is supposed to be commented and justified. Thus, here we consider only that expression:

[

Pburts = E y (n)

2

] = ∑ γ E [ x( n − i ) ] = P ∑ γ 2

J −1 i =0

t

J −1

2

s

i =0

2

t

where the fast fading effect is represented by sum of independent chi-squared variables of variance Qi such that ∑Qi=1. For a given terminal this value of fading is added in (dB) to each received burst power (in dBm) after other propagation effects. The value is correlated in time with previous values and 300000 samples of the process were stored in a file, representing about 10000 coherence periods or a duration of about 100s at a speed v=3m/s. When a burst is received at time burst_start_time, the sample number sample_nb of the pre-loaded fading vector is selected with:

⎢ burst _ start _ time ⎥ sample _ nb = ⎢ − initial _ sample⎥ T fading ⎣⎢ ⎦⎥ In practice, this calculation is performed in two times. The initial sample is randomly drawn at the initialization of the class fast_fading for each pair of terminal and it is stored in a symmetric matrix. Afterward, at the reception of the burst, the method apply is executed and the right sample in the array where all the samples are stored. For the 60GHz mode of operation, a new vector fast_fading_60GHz has been implemented where the fading samples corresponding to the channel model provided by the WP3 have been stored. Further, a new channel type was added in the Channel_Type definition in wireless-phy-const.h: CH_INDOOR_60. While for the 10

BROADWAY, IST-2001-32686

WP2-D17

Date 31-01-2005

Programme: IST

5GHz mode two possible propagation channels could be chosen, for the 60GHz one, only one channel has been considered. The duration of the simulated process is the same for both type of channels (5GHz and 60GHz) so that the variable cst_fading_duration was used also for the fast_fading_60GHz vector. To initialize the fast_fad object declared in the class WirelessPhy, two variables tot_nb_nodes and is_fast_fading_modeled were introduced/used. These variables are bound in the constructor of the WirelessPhy with two tcl variables of the scripts used to run simulations and a new one called use_fading. After the binding and depending on the value of is_fast_fading_modeled, the constructor of the class fastfading is called in order to initialise the static array Array< Array > initial_fading_sample. This array is used to randomly choose the first sample of the fading process as well as to compute the right sample of fading for each received packet. Previously the class fastfading contained the variable channel_category that has now been modified because of the addition of the second mode of operation (60GHz). Thus, an array Array channel_category was declared and its size was set in the constructor depending on the value of nb_phy_interfaces. In order to set the right channel for each interface, a new method set_fading_model was declared and defined in the class fastfading. This method is called by a command of the class WirelessPhy. That command is triggered in the .tcl scripts by the following line: $node setchannel $interface_index $value (called as many times as the number of interfaces used in the simulation) In the files fastfading.h,.cc the method to access the vector fast_fading_60GHz is fast_fading::apply is modified as following: double fast_fading::apply(double Rx_Power, int tx_identifier, int rx_indentifier, double pkt_start_time, double cst_fading_duration, int phy_interface_index=0) { if (use_fast_fading == true) { int fading_index = initial_fading_sample(tx_identifier)(rx_indentifier) + int(rint(pkt_start_time/cst_fading_duration)); fading_index = (fading_index % total_nb_fading_samples); double fading_value; switch(channel_category[phy_interface_index]) { case CH_BRAN_A: fading_value = fast_fading_A[fading_index]; break; case CH_BRAN_C: fading_value = fast_fading_C[fading_index]; break; case CH_60: fading_value = fast_fading_60GHz[fading_index]; break; default: break; } // cout