Advanced Topics in Distributed Systems Mobile Ad hoc ...

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de

Advanced Topics in Distributed Systems

(Spring/Summer 2005)

Mobile Ad hoc Networking

Prof. Dr.-Ing. Ralf Steinmetz, Dr.-Ing. Matthias Hollick

Technische Universität Darmstadt,

Fachbereich Elektrotechnik und Informationstechnik, Fachbereich Informatik (Zweitmitglied),

Fachgebiet Multimedia Kommunikation – KOM, Tel.+49 6151 166158, Fax. +49 6151 166152

Merckstr. 25, D-64283 Darmstadt, [email protected]

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de

Slide 8.2Matthias Hollick, Ralf Steinmetz, Mobile Networking (WS2004/2005)05. Jan. 2005

??

?

?

?

?

Outline (1)

Introduction and Motivation

• Terminology and Basics

• Applications for Ad hoc Networks• Ad hoc vs. the Internet Model

Routing in Mobile Ad hoc Networks

• Characteristics of Ad hoc Networks• Ad Hoc Routing Paradigms

Selected Routing Protocols (1)

• Ad Hoc On-demand Distance Vector Routing (AODV)

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


??

?

?

?

?

Outline (2)


• Dynamic Source Routing (DSR)

• Location Aided Routing (LAR)

Performance Evaluation of Ad hoc Networks

• The Art of Performance Evaluation

• Analyzing Ad hoc Network Performance

Summary and Conclusion

Appendix

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Motivation for Mobile Ad hoc Networks

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Terminology and Paradigms

“Ad hoc”

• often improvised or impromptu; „an ad hoc committee meeting“

Wordnet

• formed or used for specific or immediate problems or needs; „ad hoc solutions“

• fashioned from whatever is immediately available: improvised;„large ad hoc parades and demonstrations“

Encyclopædia Britannica

“Spontaneous”

• arising from a momentary impulse

• controlled and directed internally; „self-acting“

• produced without being planted or without human labor;„indigenous“

• developing without apparent external influence, force, cause, ortreatment

Encyclopædia Britannica

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Basics

(Mobile) Ad Hoc Communication Networks - MANET

• Historical successor of packet radio networks

• Self-organizing, mobile and wireless nodes • Absence of infrastructure, multi-hop routing necessary• Systems are both, terminals (end-systems) and routers (nodes)

• Constraints (dynamics, energy, bandwidth, link asymmetry)

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Applications

Military applications

• Battlefield communication (soldiers, tanks, planes, …)

• Smart dust (sensor networks to detect chemical, biological threats)

Civilian applications

• Vehicular environment (telematics, car to car communication, taxi cab network, …)

• Entertainment (filesharing, gaming, ... in train, car, plane, school, …)

• Event support (conferences, sport-events, exhibitions, meetings, lectures)

• Home networking / Personal Area Networking (VCR, DVD, home entertainment, remote control, cell phone, laptop, watch, …)

• Disaster recovery (emergency services, ambulance, police, …)

• Smart dust (sensor networks for civilian applications)

• Ubiquitous computers with short-range interactions (embeddedsystems, smart buildings/artefacts, …)

• Cellular range extension, moveable base stations (UMTS, WLAN, WMAN, …)

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Ad Hoc vs. The Internet Model

email WWW IPtel IMS...

SMTP HTTP RTP...

TCP UDP…

IP

ethernet

CSMA sonet...

copper fiber radio...

Spontaneous App. Behavior

Disconnected Operations

P2P App. Characteristics

User / Device Mobility

Wireless Communication

Distributed Medium Access

Service Discovery

Addressing

Name Resolution

Power Control

Gateway Discovery

New Apps.

Absence of Infrastructure

Ad Hoc Routing

User Demands

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Peer to Peer vs. Ad Hoc

P2P

• Relevant resources located at nodes at the edges ("peers")

• Variable connectivity is the norm• e.g. does it support dial-up users with variable IP addresses

• Combined Client and Server functionality• for all end system nodes • also for intermediate nodes

• Peers with significant autonomy• e.g. storage / processing done by autonomous end-systems

• Direct data transfer between peers• e.g. more-or-less no central control

• Content locations widely distributed and most often replicated

Ad Hoc and P2P share paradigms

• Ad Hoc focuses on network level and below• P2P focuses on application level• Coexistence of Ad Hoc and P2P is possible / synergetic

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Vendor Application Demo

Demo seen at www.meshnetworks.com (now Motorola)

Other companies include

• “Flarion”, “Radiant Networks”, “Invisible Networks”, “Green Packet Inc.”, “LocustWorld”, and some other companies …

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


??

?

?

?

?

Outline





• Characteristics of Ad hoc Networks

• Ad Hoc Routing Paradigms



htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Characteristics of Ad Hoc Communications

Characteristics are dominated by heterogeneity and variability

• Mobility characteristics • Speed, uniformity of movement, …

• Wireless characteristics • Broadcast nature, transmission errors, limited range, partitioning,

hidden and exposed terminals, …

• System characteristics • Distributed system, absence of infrastructure, high topology

dynamics, etc.

Inherent heterogeneity

• Do nodes have identical capabilities, responsibilities, and constraints?

• Transmission ranges and radios may differ, battery life may differ, processing capacity may differ, … (asymmetry)

• Only some nodes may route packets, some nodes may act as leaders of nodes, e.g. cluster head (asymmetric responsibilities)

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Example User Mobility: What is Realistic?

4000m x 3000m

10 nodes?100 nodes?1000 nodes?10000 nodes?

1000m x 1000m, 333 nodes

1 m/s 1000s5 m/s 200s20m/s 50s

1000m

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Node Density Quiz: Street Lamps

Moving from mobile to stationary, more mesh like multihop scenarios:

What do you think:

How many street lamps are operated in Frankfurt (am Main)

(1) Please write your first guess on the provided post-it

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Node Density Quiz: Street Lamps

Moving from mobile to stationary, more mesh like multihop scenarios:

What do you think:

How many street lamps are operated in Frankfurt (am Main)

(1) Please write your first guess on the provided post-it

Consider the following Information:~ 650.000 Residents (with ~330.000 cars)~ 248 square-km~ 1000 km streets

(2) Please write your second guess on the provided post-it

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Macroscopic Workload / Mobility Model

Synthetic mobility models (see demo)

• Easy to use

• Strict separation from traffic models• Unrealistic for large scenarios (e.g. random waypoint)

Empirical workload / mobility models

• Data is hard to obtain• Can often not be separated in mobility vs. traffic

• Available for past scenarios (may not be generalized easily)

Hybrid workload / mobility model (synthetic traffic, empirical mobility) (see demo)

• Pros: flexibility, realism• Cons: lots of parameters, data is hard to obtain

• Trade-off

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Demo of Mobility Models

Synthetic mobility model

• e.g. random walk, random waypoint

• Demo: ANSim, see http://www.i-u.de/schools/hellbrueck/ansim

Hybrid workload / mobility model

• Demo: MobQoS Model

• Seehttp://www.kom.tu-darmstadt.de/Research/MobQoS

• Developed together withSiemens Corporate Technology, Munich

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Quiz on Ad Hoc Routing

Why do we need specialized ad hoc routing?

(A) To deal with topology dynamics induced by mobility

(B) To reach nodes that are no direct neighbors

(C) To match the characteristics of wireless communication

(D) To support spontaneous formation of the network

(E) To operate without fixed infrastructure

(F) Because all end-systems are also acting as routers

Correct answers are A, B, C, D, E, and F!

We discuss these issues with selected protocols in a few minutes

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Requirements for Ad Hoc Routing

The routing protocol needs to

• Converge fast

• Minimize signaling overhead

The routing strategy (algorithm) may include

• Shortest distance, Minimum delay, Minimum loss

• Minimum congestion (load-balancing), Minimal interference• Maximum stability of routes or maximal signal strength

• Minimum energy (power aware routing)

Standard Internet routing cannot fulfill these requirements

• Assumes infrastructure, assumes symmetrical conditions, assumes plenty of resources, converges slow, typically supports hop-count metric only, …

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


??

?

?

?

?

Outline





• Characteristics of Ad hoc Networks

• Ad Hoc Routing Paradigms



htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Features of the AODV protocol

Protocol operation

• Reactive routing protocol

• Route discovery cycle for route finding• Flooded / Broadcasted Route Request (RREQ)• Unicast Route Reply (RREP) along reverse path of RREQ• Unicast Route Error (RERR)

• No overhead on data packets

Protocol principles

• Based on distance vector routing

• Loop freedom is achieved through sequence numbers, also solves “count to infinity” problem

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


• Broadcast flood acquisition using Route Request (RREQ)• Nodes set up reverse path pointing towards the source

Route acquisition process

• RREQ must never be broadcasted more than once by any node• Unicast Route Reply (RREP) propagation

AODV – Route Acquisition

S

D

Y

X

Z

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


AODV – Route Error

Link failure reporting / repairing routes

• If X is unable to forward packets from S to D on link (X,Y)

X increments its Destination Sequence Number (N) for D

A Route Error (RERR) message is generated at X

The RERR includes N and is sent out based on precursor lists

Upon receipt of RERR, S initiates new route discovery for D using destination sequence number at least as large as N

SYX

D

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


AODV – Route Maintenance

Timers to keep route alive

• A routing table entry maintaining a reverse path is purged after a timeout interval (Discuss: How long should interval be at least?)

• A routing table entry maintaining a forward path is purged if not used for the active_route_timeout interval

DB

E

CA

Here: Destination Sequence Number to determine freshness

• To avoid using old/broken routes

• To prevent formation of routing loops

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Example

Parameters

AODV

25 nodesPure Flooding

Area = 836m*836m

Radio Range = 250mNode Density = 7

RREQ from #13 to #2

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Other Features of AODV

Status

• Implementations available (for IPv4, IPv6)

• IETF Experimental RFC status issued (July 2003)

Target networks

• Where routing churn is high enough that proactively maintaining routes is unproductive & can absorb network wide broadcast rate

Multiple optimizations

• AODV-LR – Local Repair

• AODV-ESP – Expanding-Ring Search

• Multipath & Multicast extensions (AODVM, AOMDV, MAODV)

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


AODV Optimizations – Gossiping (1)

Flooding is very inefficient, esp. if node density is high

• Probabilistic techniques are expected to better deal with the

highly dynamic and mobile characteristics of the network

• Probabilistic techniques may easily be adapted to network

density without breaking symmetric conditions

Gossiping as example of epidemiological algorithm

• Assume a large population of n people

• A rumor is initially transmitted to one member of the population

• This person passes (forwards) the rumor to a fixed number of

confidants with probability p, the rumor is kept secret with

probability 1- p (other modes possible)

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


AODV Optimizations – Gossiping (2)

Gossip-variants

• Gossip (p1), Gossip (p1, k), Gossip (p1,k,m)

• General forwarding probability p1

• Number of neighbors n; probability p2 for n < n0 , p2 > p1

• Hop count k; forwarding probability p = 1 for k k0

• Number of overheard messaged m; p = 1 for m m0

Pseudo-Algorithm for Gossip (p1, k)upon reception of message m at node nif message m received for the first time then

if hop count k less or equal k0 then

broadcast(m) with probability 1else

broadcast(m) with probability p1

end if

end if

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


AODV Extension – Multipath (AODVM)

Multipath variants for multiple protocols

• Easy for source routing algorithms

• Not trivial for optimized protocols like AODV• “Multipath” should NOT be mixed up with “Multicast”

AODVM

• Ad hoc On-demand Distance Vector Multipath (AODVM) Routing

• Instead of dropping duplicate RREQ packets, AODVM uses an RREQ table to store the redundant RREQ information.

Route Request (RREQ)

B

A

SE F

H

JC

G

I

K

Z

M

N

D

Node M’s RREQ table

3ZS

J

Neighbor

3S

DistanceSource

Source: Zhenqiang Ye

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


AODVM Explained

Path Discovery Procedure

• Destination initiates an RREP for each RREQ that is received (from different neighbors).

• Nodes overhear the RREP packets• A node that is assigned to a route is deleted from its neighbors’

RREQ tables

3ZS

J

Neighbor

3S

DistanceSource

B

A

SE F

H

J

I

K

Z

M

N

D

Node M’s RREQ table

CG

Route Request (RREQ)

Route Reply (RREP)

Source: Zhenqiang Ye

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


??

?

?

?

?

Outline



• Location Aided Routing (LAR)• Optimized Links-State Routing (OLSR)

Current Research Challenges in Mobile Ad hoc Networking

• Dependable Routing in Ad hoc Networks• Modeling the AODV Route Acquisition Process


Appendix

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Dynamic Source Routing (DSR)

Shares some principles with AODV

• When node S wants to send a packet to node D, but does not know a route to D, node S initiates a route discovery

• Source node S floods Route Request (RREQ)

• Each node appends own identifier when forwarding RREQ

• Following an example for a route discovery from source S to destination D

B

A

S EF

H

J

D

C

G

IK

Z

Y

M

N

L

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Route Discovery in DSR

B

A

S E

F

H

J

D

C

G

I

K

Z

Y

M

N

L

Represents a node that has received RREQ for D from S

Source: Nitin Vaidya

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de



B

A

S E

F

H

J

D

C

G

I

K

Z

YBroadcast transmission

M

N

L

[S]

Represents transmission of RREQ

[X,Y] Represents list of identifiers appended to RREQ


htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de



B

A

S E

F

H

J

D

C

G

I

K

Z

Y

M

N

L

[S,E]

[S,C]

Node H receives packet RREQ from two neighbors:

potential for collision


htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de



B

A

S E

F

H

J

D

C

G

I

K

Z

Y

M

N

L

[S,C,G]

[S,E,F]

Node C receives RREQ from G and H, but does not forward

it again, because node C has already forwarded RREQ once


htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de



B

A

S E

F

H

J

D

C

G

I

K

Z

Y

M

N

L

[S,C,G,K]

[S,E,F,J]

Nodes J and K both broadcast RREQ to node D

Since nodes J and K are hidden from each other, their

transmissions may collideSource: Nitin Vaidya

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de



B

A

S E

F

H

J

D

C

G

I

K

Z

Y

M

N

L

[S,E,F,J,M]

Node D does not forward RREQ, because node D

is the intended target of the route discovery


htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de



Route Reply (RREP)

• Destination D on receiving the first RREQ, sends a (RREP)

• RREP is sent on a route obtained by reversing the route appended to received RREQ

• RREP includes the route from S to D on which RREQ wasreceived by node D

• Route Reply can be sent by reversing the route in Route Request (RREQ) only if links are guaranteed to be bi-directional

• If unidirectional (asymmetric) links are allowed, then RREP may need a route discovery for S from node D

• Unless node D already knows a route to node S

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Route Reply in DSR

B

A

S E

F

H

J

D

C

G

I

K

Z

Y

M

N

L

RREP [S,E,F,J,D]


Represents RREP control message

Node S on receiving RREP, caches the route included in the RREP

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Data Delivery in DSR

B

A

S E

F

H

J

D

C

G

I

K

Z

Y

M

N

L

DATA [S,E,F,J,D]


Packet header size grows with route length

When node S sends a data packet to D, the entire route is included in the packet header (hence the name source routing)

Intermediate nodes forward according this route

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Location-Aided Routing (LAR)

Exploits location information to limit scope of flooding for route request

• Location information may be obtained using GPS

Expected Zone is determined as a region that is expected to hold the current location of the destination node (D)

• Expected region determined based on potentially old location information, and knowledge of the destination’s speed

Route requests limited to a Request Zone that contains the Expected Zone and location of the sender node (S)

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Expected Zone in LAR

S = Source node, D = Destination node

X = last known location of node D, at time t0

Y = location of node D at current time t1, unknown to sender S

r = (t1 - t0) * estimate of D’s speed

X

Y/D

r

SExpected Zone

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Request Zone in LAR

Xr

S

Request Zone

Network Space

BA Y/D

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Operation of LAR (1)

Only nodes within the request zone forward route requests

• Node A does not forward RREQ, but node B does

• Request zone explicitly specified in the route request• Each node must know its physical location to determine whether

it is within the request zone

Xr

S

BA

Request Zone

Y/D

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Operation of LAR (1)

Only nodes within the request zone forward route requests

If route discovery using the smaller request zone fails to find a route, the sender initiates another route discovery (after a timeout) using a larger request zone

• the larger request zone may be the entire network

Rest of route discovery protocol similar to DSR (will be discussed later)

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


LAR Variations: Adaptive Request Zone

Each node may modify the request zone included in the forwarded request

• Modified request zone may be determined using more recent information and may differ from original request zone

S

B

Request zone adapted by B

Request zone defined by sender S

Until now a route request explicitly specified a request zone

• Implicit Request Zone: A node X forwards a route request received from Y if node X is deemed to be closer to the expectedzone as compared to Y

• The motivation is to attempt to bring the route request physically closer to the destination node after each forwarding

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Location-Aided Routing

The basic proposal assumes that, initially, location information for node X becomes known to Y only during a route discovery

• This location information is used for a future route discovery• Each route discovery yields more updated information which is

used for the next discoveryVariations

• Location information can also be piggybacked on any message from Y to X

• Y may also proactively distribute its location information

LAR Summary – Advantages• Reduces the scope of route request flood• Reduces overhead of route discovery

LAR Summary – Disadvantages• Nodes need to know their physical locations• Does not take into account possible existence of obstructions for

radio transmissions

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


LAR-Optimization Query Localization

Limits route request flood without using physical information

• Route requests are propagated only along paths that are close tothe previously known route

• The closeness property is defined without using physical location information

Path locality heuristic: Look for a new path that contains at most k nodes that were not present in the previously known route

• Old route is piggybacked on a Route Request• Route Request is forwarded only if the accumulated route in the

Route Request contains at most k new nodes that were absent in the old route

• This limits propagation of the route request

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


??

?

?

?

?

Outline





• The Art of Performance Evaluation• Analyzing Ad hoc Network Performance


Appendix

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


The Art of Performance Evaluation (1)

“That which is monitored improves.”

—unknown Source

The quotes in this lecture part are taken from [Jain1991]

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de



Basics Terms

• Performance Analysis = Analysis + Computer Systems

• System – any collection of hardware and software• Metrics – the criteria used to evaluate the system performance• Workloads – the requests made by the users of the system

Is performance analysis is an art?

• Successful evaluation cannot be produced mechanically• Every evaluation requires a large set of skills

• Given the same data, two analysts may interpret them differently

Example

• 2 systems, which one is the best?

• Throughputs of two systems A, B measured in transactions/second

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de



Possible Solutions

• (1) Compare the average: both systems are equal

• (2) Compare the ratio with system B as base

• (3) Compare the ratio with system A as base

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Common Mistakes & How-to Avoid Them (1)

1. No Goals

2. Biased Goals

• Any endeavor without goals is bound to fail!

• There is no such thing as a general-purpose model!• Goals correct metrics, workloads, and methodology!

• “To show that OUR system is better than THEIRS” ?

• Analysts = Jury!

(Source www.zeit.de)

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de



3. Unsystematic Approach

4. Analysis without understanding the problem

5. Incorrect Performance Metrics

6. Unrepresentative Workload

… see Appendix for othercommon mistakes


• Arbitrary selection of workload Inaccurate conclusions!


• “A problem well stated is half solved”!


• Chosen to be easily computed rather to be relevant?


• Should represent the actualusage of the system!


htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


So Everything Went Wrong …


How-to do it correct from the beginning?

How to Analyze (Mobile Ad hoc) Networks?

Best practice, since there is no perfect solution!

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Systematic Performance Evaluation

Again: systematic approach to performance evaluation

(adopted from [Jain1991])

1. The definition of the system (boundaries), goals, and services2. The selection of the performance metrics

3. The definition of the parameters to study4. The selection of the factors/elements of the parameter set5. The choice of the evaluation technique

6. The development of the model and selection of the workload7. The design and conducting of the individual experiments

8. The analysis and interpretation of the obtained data9. The presentation of the results

“Important Files”

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Example: Ad hoc Network Performance

Scenario

??

?

??

?

Challenge

• Qualify and quantify the

effects of node misbehavior

on the overall performance of

the routing system

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


2. Selection of the Performance Metrics (2)

What do we expect?

• Criteria to compare the performance (so-called metrics)

• Metrics relate to the speed,accuracy, and availability of services

Network size

Mobility

Avg.E2Edelay

RoutingMessages

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


2. Selection of the Performance Metrics (3)

Some “realistic?” results

only useful in combination with

other metricsLeB

eP

e2003

PeR

oD

aM

A2001

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


5. Choice of Evaluation Technique

• Real world observations• no large scale MANET, expensive, only deployed technologies

• Emulation / Testbed experiments • only small scale, expensive, complicated (modeling mobility?)

• Simulation studies • medium scale (~50-1000 nodes), restricted to chosen scenarios

• Analytical models • assume ideal protocols, mainly for capacity / connectivity on L2

Current investigation methods (boundaries not fix)

Abstraction Level vs. Generality of Results

Emulation(Testbed)

Analysis Simulation Real World

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


System and Model

Real World

AnalysisSimulation

Emulation(Testbed)

Various degrees of abstraction

Observation

Workload CalculationsSimulation results

Measurements

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Degrees of Abstraction (1)

Real World

Emulation(Testbed)

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Degrees of Abstraction (2)

Abstraction Level vs. Generality of Results

Analysis Simulation Emulation(Testbed)

Real World

L5Application

Layer

L4Transport

Layer

L3Network

Layer

L2Data Link

Layer

L1Physical

Layer

L5Application

Layer

L4Transport

Layer

L3Network

Layer

L2Data Link

Layer

L1Physical

Layer

L5Application

Layer

L4Transport

Layer

L3Network

Layer

L2Data Link

Layer

L1Physical

Layer

L3Network

Layer

Scope of Modeling

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


6. Workload

7. Design and Conducting of Experiment

“Would you tell me, please, which way I ought to go from here?”

“That depends a good deal on where you want to get to,” said the Cat.

“I don’t much care where —” said Alice.

“Then it doesn’t matter which way you go,” said the Cat.

— Lewis Carroll, Alice’s Adventures in Wonderland

Workload is crucial

• Services exercised, Level of detail• Representativeness, Timeliness

• Also consider load level!

Experimental Design

• Simple Designs (vary one factor at a time)

• Full/ Fractional Factorial Designs (utilizes full/partial combination of levels of all factors)

• How many experiments?• Given k factors,

with the i th factorhaving ni levels

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


7. Design and Conducting of Experiments (2)

Workload inExample

• Moderate

• Artifical

• 8% of nodes are sources (random)

• We use constant bitrate traffic over UDP.

Factors that are kept constant (250 node setup)• Average node density = 9; Area = (2334m)2; Simulation time = 900s• Link layer = IEEE 802.11b with Distributed Coordination Function (DCF)• Network layer = Ad hoc On-demand Distance Vector Routing, IPv4 • Transport/Application layer = Constant Bit-Rate (CBR) flows over UDP

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


8. Analysis and Interpretation of Data (1)

A man with one watch knows what time it is.

A man with two watches is never sure.

— Segal’s Law

Steps contd.

• The analysis only produces results and not conclusions• Based on these results the analysts/decision makers can draw

conclusions

• Outcomes of measurements/simulations are random quantities• They change with each replication

• Take into account the variability of the results

• Simply comparing the means can lead to inaccurate conclusions• Proper statistics are basis for a unbiased analysis

• Each analysts will probably draw different conclusions are giventhe same set of results …

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de



2800500050008000

1000010000200006000075000

100000100000100000100000100000150000180000250000250000300000400000500000500000550000

15000001500000

17000000

“Then there is the man who drowned crossing a stream with an average depth of six inches.”

— W. I. E. Gates

Some statistics …

How many street lamps are operated in Frankfurt (am Main) …

• First guess vs. “educated” guess, N = 26 participants• The correct solution was ~ 64,000 to 65,000 (or 42 ;-)

• First guesses range from 2,800 to 17,000,000Mean = 914,454, 95% Confidence interval = +/- 437,160Median = 100,000, = 3,304,848

• Educated guesses range from 3,500 to 750,000Mean = 113,392, 95% Confidence interval = +/- 20,590Median = 57,500, = 155,656

32007000800015000200002000020000250003000040000400005000050000650008000099999100000100000100000125000200000200000200000300000300000750000

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de



Tell me what you see:

(A)

(B)

(E)

(C)

(D)

(C)

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de



Observations for Selfish Nodes

Packet loss increases …

• … nearly linearly with the number of selfish nodes

• … with node mobility

Routing overhead increases

15%42%

i.e. 40 out of 250 nodes

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de



Observations for Black Holes

Successful communication only in close proximity

Packet loss …

• … is extremely high, even for few black holes• … increases with node mobility

49%78%

i.e. 10 out of 250 nodes

Increase inmalicious nodes

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Closing the Art of Performance Evaluation

“Do not plan a bridge capacity by counting the number

of people who swim across the river today.”

—Heard at a presentation

The quotes in this lecture part are taken from [Jain1991]

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


??

?

?

?

?

Outline







Appendix

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Summary

Mobile ad hoc networks (MANET)

• Self-organizing, mobile and wireless nodes

• Absence of infrastructure, multi-hop routing necessary• Systems are both, terminals (end-systems) and routers (nodes)• Constraints (dynamics, energy, bandwidth, link asymmetry)

Characteristics include

• Mobility, wireless, application / traffic, and system characteristics

Routing is a central problem in ad hoc networks

• A large set of routing protocols exists• Open issues persist: QoS, Security, Scalability, Heterogeneity, …

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Taxonomy of Routing Protocols

• The above mentioned protocols are only a selection!

• AODV, DSR, OLSR, and TBRPF are currently moved towards Experimental RFC (within IETF)

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Some Routing Protocols / Frameworks

AODV - Ad Hoc On Demand Distance Vector (Perkins, NOKIA; Belding-Royer, UCSB; Das, UC)

CEDAR - Core-Extraction Distributed Ad Hoc Routing

DREAM - Distance Routing Effect Algorithm for Mobility

DSDV - Destination-Sequenced Distance Vector

DSR - Dynamic Source Routing (Johnson, CMU)

FSR - Fisheye State Routing

LANMAR - Landmark Ad Hoc Routing

LAR - Location Aided Routing

OLSR - Optimized Link State Routing (Clausen, Jacquet, INRIA)

TBRPF - Topology Broadcast based on Reverse-Path Forwarding (Ogier,Templin, SRI)

Tora / IMEP - Temporally-Ordered Routing Algorithm / InternetManet Encapsulation Protocol

ZRP - Zone Routing Protocol (Haas, Cornell)

… see also http://www.wikipedia.org/wiki/Ad_hoc_protocol_list

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Mobile Ad Hoc Network = MANET = Manet?

There is even more to the name MANET

• Édouard Manet (1832 – 1883)

• “Godfather” of impressionisms

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Acknowledgements

Acknowledgements for inspirations to the ad hoc tutorial go to:

• N. Vaidya @ UIUC

• C. Perkins @ Nokia• S. Corson @ UMD• P.R. Kumar @ UIUC• D. Johnson @

• Zhenqiang Ye @ UCR

• Raj Jain @ Ohio State • Jens Schmitt @ Universität Kaiserslautern• A. Wolisz @ TU Berlin

• Matthias Kropff @ KOM, TUD• The other colleagues of the Mobile Networking Group @ KOM

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Contact – Feedback

Prof. Dr.-Ing. Ralf Steinmetz

Dept. of Electrical Engineering and Information TechnologyMultimedia Communications Lab · KOM

Merckstr. 25 · 64283 Darmstadt · GermanyPhone (+49) 6151 – 16 6150Fax (+49) 6151 – 16 [email protected]://www.kom.tu-darmstadt.de

Matthias Hollick

Mobile Networking

Dept. of Electrical Engineering and Information TechnologyMultimedia Communications Lab · KOM

Merckstr. 25 · 64283 Darmstadt · GermanyPhone (+49) 6151 – 16 6158Fax (+49) 6151 – 16 [email protected]://www.kom.tu-darmstadt.de

Further information• http://www.kom.tu-darmstadt.de

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


??

?

?

?

?

Outline







Appendix

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Appendix: Additional Readings (1)

Research Papers

• See course download page for selected research papers

• Laura M. Feeney, A Taxonomy for Routing Protocols in Mobile

Ad Hoc Networks, Technical Report 1999• Young Bae-Ko and Nitin Vaidya, Location-Aided Routing (LAR)

in Ad Hoc Networks, Infocom 1998

IETF Standardization

• MANET WG http://www.ietf.org/html.charters/manet-charter.html

• AODV: http://www.ietf.org/rfc/rfc3561.txt• DSR: http://www.ietf.org/internet-drafts/draft-ietf-manet-dsr-10.txt• OLSR: http://www.ietf.org/rfc/rfc3626.txt• TBRPF: http://www.ietf.org/rfc/rfc3684.txt

• The papers/standards are not mandatory for the course but you might find them interesting

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de


Appendix: Additional Readings (2)

Books

• Raj Jain: "The Art of Computer Systems Performance Analysis:

Techniques for Experimental Design, Measurement, Simulation,

and Modeling" (ISBN 0-471-50336-3)

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de







… see Appendix for othercommon mistakes


• Arbitrary selection of workload Inaccurate conclusions!

• Systematic approach is key to solve performance problem!


• The “world” model is the goal No!• “A problem well stated is half solved”!


• Chosen to be easily computed rather to be relevant?


• Should represent the actualusage of the system!


htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de



7. Wrong Evaluation Technique

8. Overlooking Important Parameters

9. Ignoring Significant Factors

10. Inappropriate Experimental Design

11. Inappropriate Level of Detail

7. Wrong Evaluation Technique

• Use the technique you know best vs. the appropriate one?

• Methods are measurement, (emulation), simulation, analysis!

8. Overlooking Important Parameters

• Just do it?• Make a list of all system/workload characteristics!

9. Ignoring Significant Factors

• Vary only well-understood parameters?• Use the factors that impact the performance most!

10. Inappropriate Experimental Design

• Naïve: each factor changed one by one cross-influences?• Improper selection Waste of time!

11. Inappropriate Level of Detail

• Too narrow vs. too broad …

• Slight variations to study detailed model!

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de



12. No (bad) Analysis

13. Erroneous Analysis

14. No Sensitivity Analysis

15. Ignoring Errors in Input

16. Improper Treatment of Outliers

12. No (bad) Analysis

• Measure, measure more, ... you’ve just produced tons of data!

• Produce summary, include analysis + measurement background!

13. Erroneous Analysis

• Average of ratios (example earlier)?• Too short simulations?

14. No Sensitivity Analysis

• Fact vs. evidence! very much depends on workload!• Clarify relative importance of parameters!

15. Ignoring Errors in Input

• Uncertainty because of inaccurate measurements?• Biased input?

16. Improper Treatment of Outliers

• Too high or too low values? real system phenomenon or outlier needs to be clarified!

htt

p:/

/ww

w.k

om

.tu

-darm

sta

dt.

de



17. Assuming No Change in the Future

18. Ignoring Variability

19. Too Complex Analysis

20. Improper Presentation of Results

21. Ignoring Social Aspects

22. Omitting Assumptions and Limitations

17. Assuming No Change in the Future

• Future is same as past? carefully limit assumptions!

18. Ignoring Variability

• Analyze only the mean performance?• E.g. take only daily average of load demands if peak hours exist?

19. Too Complex Analysis

• Reality: the more complex the model the “easier” to publish …

• Simple models are beneficial for day-to-day purposes!

20. Improper Presentation of Results

• Right metric to measure performance of analyst is not number of analyses performed but number of helpful analyses!

21. Ignoring Social Aspects

• Underestimation of presentation? social skills important!

22. Omitting Assumptions and Limitations

• Missing assumptions in report? results are falsely transferred!

Advanced Topics in Distributed Systems Mobile Ad hoc ...

Documents

Transcript of Advanced Topics in Distributed Systems Mobile Ad hoc ...