CS 257: Wireless Networks andMobile Computing

Srikanth Krishnamurthy

T, Th: 3:40 p.m. to 5 p.m.

Office hrs: W 10-11 a.m.

Info

• My E-Mail: krish@cs.ucr.edu

• Office Hours: Wednesdays 10 a.m. to 11 a.m.

• Web : http://www.cs.ucr.edu/~krish

Course Details• Discussions based

• Evaluation by means of Project

• 55 % for Report

• 25 % for Presentation (Oral and Written)

• 10 % for Participation/ Discussion

• 10 % Quiz

• Need to have taken CS 164 -- good understanding ofcomputer networks

• Able to read technical papers -- understand, evaluate.

What do you do ?

• Task 1: Critical Evaluation of Literature

• We form groups of two.

• Each group looks at one paper from eitherINFOCOM, MOBIHOC, SENSYS or MOBICOM of thelast two years.

• Need to be thorough -- understand and read otherrelevant papers!

• Each group must clearly articulate almost all thedetails in the paper and emphasize on what theylearned from the paper.

Others who don’t present in aparticular class...

will express themselves andparticipate in discussions.

Task 2: Project• Consists of the majority of the grade -- so I expect someeffort.

• Possibility 1: An original idea in the area of wirelessnetworks.

• Can be very simple -- but needs to be new.

• Motivation -- Last year in 257, one paper was publishedin ICCCN 2006.

• Prior projects have yielded papers.

• You should be clear why are you doing this ? Why theidea is enticing ?

• I will offer help in framing the problem working with you.

• Teams of 1 or 2.

What do I expect in the the originalproject?

• Could be Analysis Oriented

• Simple constructions using either graph theory, stochastic modelsor queuing models.

• Protocol design and evaluation via simulations

• New stuff though!

• If you are doing simulations, no point in repeating other peopleswork.

• Implementations

• Difficult – need not be original

• Lack of resource may be a problem -- but we have a wirelesstestbed.

• Ultimately your enthusiasm is what counts.

• Possibility 2: A Survey on a Specific Topic

• Needs to be comprehensive

• A rich set of papers -- articulate thecontributions in each, differentiate between theefforts.

• Examine the possibilities -- having explored thisarea, what did you learn ?

• Are there any remaining problems left tosolve.

• If not, why ? Why is this a closed area ?

• Teams of 1 or 2.

Project Schedule

• A Project Abstract -- a brief description of what you intendto do -- more like an abstract and a statement of work is due inthree weeks ( October 19)

• Teams embarking on original research projects might want tosend me e-mail and schedule appointments or drop by to discuss.

• Final Term Paper due on the last day of class -- in the formof a technical paper (just like the ones you read) --- no morethan 8 pages -- IEEE 2 column format.

• The term paper needs to be formal -- abstract, intro, yourproposed work etc etc.

• Write well: I need to understand what you have done.

ABSOLUTELY NO EXTENSIONSDue on the last day of class and

Presentations

Tentative:

• Each presentation will last 40 minutes. Firstmember will present in 15 minutes and secondmember will present in the next 15 minutes.

• Ten minutes for discussion questions/ answers.

• Two groups will present in a class.

Presentations for Project

• Each member will present for 15 minutes each.

• Need to cover relevant work.

• Ten minutes for discussion at the end.

• NOTE: Need to make sure that your presentationis done in the time allocated. A good rule of thumbis to have two minutes allocated for a slide -- so in15 minutes you probably want to have no more than8 slides.

What will I do ?

• Give you some thoughts on what wireless networks areabout -- pointers to books papers that you may want toread.

• Cover some interesting papers in wireless networks --main focus on modeling networks, cross layer design.

• Note that it is impossible to provide an extensivecoverage of what has happened in wireless in a quarter !

• So in summary -- it is your course -- the more effortyou put into it, the more you will learn.

• Emphasis on knowing about what is exciting today aboutwireless -- stepping stone for research.

• Some of my senior graduate students will present someinteresting papers as well.

Grading Policies

• Previous project endeavors have lead toconference papers -- that should be your goal!

• I am very liberal with the grades if I see thatyou have spent time “enthusiastically” towards thecourse and the project.

Why should we have differentnetworking strategies with wireless?

• High error rates -- wire-line protocols assume that errorrates are very low.

• Mobility -- Nodes can move around -- so networkdynamically changes (extent may vary).

• Wireless medium

• Arbitrate the access -- simultaneous transmissions canlead to interference.

• Security becomes a major issue.

• Synchronization issues.

• New Applications -- Sensors, Traffic Control etc.

• Inter-layer interactions if traditional layered protocols areused.

In this lecture....

• Preliminary discussion on wireless networks, channelmodels, spectrum sharing, cross layer design.

• Reference: Any book on wireless communications andnetworks.

Acknowledgements: Slides from PrashantKrishnamurthy, Dept of Info Sciences andTelecommunications, Univ. of Pittsburgh.

Model of a wireless communication system asviewed at the physical layer

SourceSource Encoder

ChannelEncoder

Modulator

DestinationSource Decoder

ChannelDecoder

Demod-ulator

Radio C

hannel

What networking folks assume

Disk ModelsSensingRange

Transmissionrange.

Fading Channels

Large Scale FadingSmall Scale

Fading

Path LossShadow Fading

Time Variation Time Dispersion

Amplitude fluctuationsDistribution of amplitudes

Rate of change of amplitude“Doppler Spectrum”

Multipath Delay SpreadCoherence Bandwidth

Inter-symbol InterferenceCoverage

Receiver Design (coding)Performance (BER)

Receiver DesignMaximum Data Rates

Actual Models

Performance degradation and mitigation

EqualizationDS-Spread Spectrum

OFDMDirectional Antennas

Inter-symbol Interferenceand

Destruction of signal

TimeDispersion

OfSignal

Error control codingInterleaving

Frequency hoppingDiversity

Bit error ratePacket error rate

TimeVariation

OfSignal

Fade Margin – Increasetransmit power

or decrease range

Received Signal StrengthShadowFading

MitigationTechnique

Performance AffectedIssue

Path Loss Models• Used very commonly to estimate link budgets, cell

sizes and shapes, capacity, handoff criteria etc.• “Macroscopic” or “large scale” variation of RSS• Path loss = loss in signal strength as a function of

distance– Terrain dependent (urban, rural, mountainous)– Site dependent (antenna heights for example)– Frequency dependent

• Usual characterization: Lp = L0 + 10α log10(d) (in dB)– The parameter α is called the “path loss gradient”

or exponent– The value of α determines how quickly the RSS

falls

Shadow fading

• The path loss is NOTthe same for all d– There is a variation

about the mean– This variation has a

“distribution”• Experiments show that

the distribution islognormal– In dB the distribution is

normal– Usually modeled as a

zero mean RV withstandard deviation σ 100 101 102

50

60

70

80

90

100

110

120

130

140

distance in m

path loss in dB

Small scale fading• Multipath = several delayed

replicas of the signal arrivingat the receiver

• Fading = constructive anddestructive adding of thesignals

• Changes with time• Results in poor signal quality• Digital communications

– High bit error rates

0 5 10 15 20 25 30

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

time

ampliutude

0 5 10 15 20 25 30

-2.5

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

time

amplitude

17 17.5 18 18.5 19 19.5 20 20.5 21 21.5-1.5

-1

-0.5

0

0.5

1

1.5

time

amplitude

amplitudeloss

Small scale fading amplitude characteristics

• Amplitudes are Rayleigh distributed– Worst case scenario – results in the poorest

performance• In line-of sight situations the amplitudes have a Ricean

distribution– Strong LOS component has a better performance– Weak LOS component tends to a Rayleigh

distribution• Other distributions have been found to fit the

amplitude distribution– Lognormal or Nakagami

Rayleigh, Rician and Lognormal PDFs

f rr r

rray ( ) exp( ),= − ≥σ σ2

2

220

f rr r K

IKr

r Kric ( ) exp(( )

) ( ), ,=− +

≥ ≥σ σ σ2

2 2

2 0 220 0

−−=

2

2

2

)(lnexp

2

1)(

σµ

σπ

x

xxf LN

I0(x) is the modified Bessel function of the first kind of order zero

Time variation of the channel

• The radio channel is NOT time invariant– Movement of the mobile terminal– Movement of objects in the intervening

environment

• How quickly does the channel fade (change)?– For a time invariant channel, the channel does not

change – the signal level is always high or low– For time variant channels, it is important to know

the rate of change of the channel (or how long thechannel is constant)

Fade rate and fade duration

• The signal “level” is the dB above or below the RMS value• Fade rate determines how quickly the amplitude changes

(frequency Doppler Spectrum)• Fade duration tells us how long the channel is likely to be “bad”• Design error correcting codes and interleaving depths to

correct errors caused by fading

“level”

6 crossings of the “level” in 30 seconds

time under the “level”

packet

0 30

time

Coping with Fading: Diversity• Idea: Send the same information using several

“uncorrelated” paths or forms– Not all repetitions will be lost in a fade

• Types of diversity– Time diversity – repeat information in time spaced so as to

not simultaneously have fading• Error control coding!

– Frequency diversity – repeat information in frequencychannels that are spaced apart

• Frequency hopping spread spectrum– Space diversity – use multiple antennas spaced sufficiently

apart so that the signals arriving at these antennas are notcorrelated

• Usually deployed in all base stations but harder at mobiles

Performance with diversity• If there is ideal diversity,

the performance canimprove drastically

• There are different formsof diversity combining– Maximal ratio combining

• Difficult to implement– Equal gain combining

• Easy to implement– Selection diversity

• Easy to implement5 10 15 20 25 30 35 40 45

10-6

10-5

10-4

10-3

10-2

10-1

Average Eb/N

0

Pe

BPSK under fading with diversity

BPSK-No fadingBPSK Div=2BPSK Div = 4BPSK No Div

Classification of Wireless Networks

• Cellular Networks: Organized, base stations thatare regularly placed. Mobiles communicate only withbase stations.

• Wireless LANs: Less organized; base stations oraccess points with which mobile nodes communicate.

• Ad hoc networks : No infrastructure; nodes moveand network dynamically changes.

• Sensor Networks: application specific; mobility islimited (perhaps to selected subset of nodes); tinynodes that are resource and energy constrained.

Inter Layer Dependencies

• OSI or TCP/IP stack may not be the way to go!

• Recent paper in IEEE Wireless CommunicationsMagazine by V. Kanodia and P.R. Kumar suggeststhat perhaps a new “layering” strategy is needed.

• New NSF thrust on new “layered architectures”for integrated wired/wireless networks.

• Current trend : Eliminate layers or introducehooks such that layers can interact with eachother.

An Example: TCP over ad hocnetworks

• Ad Hoc Networks will have to be interfaced withthe Internet.

• As such backward compatibility is a big issue.

• One might expect that the TCP/IP suite ofprotocols be applicable to the ad hoc domain.

• Much research on routing – IP layer.

• What are the problems with TCP ?

Problems with TCP• TCP attributes packet losses to congestion.

• What does it do when it perceives a packet loss ?

• It goes back to the Slow Start Phase and restartswith one packet.

• This would result in a degradation of TCPthroughput.

• Notice that packet losses could be due tofading/mobility. Why due to mobility ?

Packet Losses due to Mobility• When nodes move, links tend to break, and getformed again.

• When the SIR is below certain threshold, the MAClayer concludes that the link is broken.

• This would create an interrupt at the routing layer.

• Now, the routing protocol has to deduce the newlocation of the destination.

• Until it finds the new route, what happens to TCP ?

• It keeps reducing the transmission window and tryingto retransmit.

• This leads to

• Unnecessary retransmissions when there is no link

• Beginning at slow start when the link comes upagain.

• What if the destination cannot be found at all ?

• ICMP may be used to detect link failures etc.(Notice at the IP layer)

• SNMP could be used for fault management.

• But these are slow.. if links fail often, but you knowthat recovery is possible, then aborting the connectioneach time may not be the right thing to do.

Reference• G.Holland and N. Vaidya “Analysis of TCPPerformance over Mobile Ad Hoc Networks”, inProceedings of Mobicom 1999.

• Simulated the performance of TCP over ad hocnetworks and they report their findings.

• Interesting observations are made.

Effects of Mobility Patterns

• One would expect that the higher the mobility i.e.,the faster the nodes, the more the degradation inthroughput.

• However, Vaidya and Holland found that this was nottrue in all cases.

• Relative velocity counts – not absolute.

• Scenario dependent – although the general trendexists.

• In summary, some mobility patterns yield highthroughput while others yield low throughput.

Effects of Routing Protocols

• The performance also depends upon the routing protocol whichis at the IP layer.

• Presence of stale routes caused a major degradation in TCPperformance.

• Notice, this in turn depends upon the rate at which routingtables are updated (if at all).

• ARP failures – a node assumes that another node is a neighbor– but now that node has moved away.

• Asymmetry in routes (routing protocol dependent) causes ACKsto get lost – degradation due to reverse path as opposed toforward path.

DSR

• In Holland/Vaidya paper, the authors looked at aparticular routing protocol DSR (Dynamic SourceRouting).

• This is an on-demand routing scheme (we will see itsoon).

• The source searches for a route.

• Once the route is found, it is cached by theintermediate nodes and the source for sometime.

• If the route breaks, a new search is initiated.

Specific Problems Identified

• The authors noted that caching created stale routes.

• Not only this, when a source searched for a newroute, some of the nodes in between reported staleroutes – resulted in TCP backing off since staleroutes lead nowhere.

• But there is a trade-off between caching and notcaching – if nothing is cached frequent route queriescan cause congestion.

• How do we determine what is the optimal purge timefor caches.

• One conclusion that they draw is that if TCP has towork well, underlying routing has to be doneefficiently.

• Second conclusion is that the degradation isscenario dependent – speed etc. do not allow one tomake a generalized conclusion.

Using Explicit Feedback

• The Idea is similar to the use of explicit notificationsis not new – ECN or Explicit Congestion Notification inthe Internet to inform source about congestion.

• A similar scheme can be thought of which can providethe source about an explicit notification about thefailure of a link.

• This message may be called ELFN or Explicit LinkFailure Notification.

• Upon receiving this message, a TCP source can inferthat packet losses are due to link failures rather thancongestion,and therefore act differently.

How can we implement ELFN ?

• Simplest way : ICMP message to indicate that host isunreachable.

• Second possibility : Piggyback this to TCP on theRoute Failure Message.

• NOTICE: Cross Layer Dependencies.

• When the TCP layer at the source receives thismessage it disables the congestion control mechanisms.

• What does it need to do ?

• Two main questions are:

What does TCP do in response to the ELFNnotice ?

How does TCP know when the route is restored ?

TCP response to an ELFNmessage

• Enter a mode called the standby mode.

• Disable the retransmission timers.

• In this mode a packet is sent at periodic intervals toprobe whether the route has been established.

• If an ACK is received, it leaves the stand-by modeand restores its retransmission timers, and continuesas normal.

• Another possibility is to generate an explicit routerestored notification – but how ?

• Vaidya and Holland found that the ELFN messageimproved performance for all scenarios.

• Their observations were as follows:

The performance change was sensitive to probeinterval. If the interval was too large, not muchimprovement. If it was too small it leads tocongestion and further degradation inthroughput.

The performance change could depend both onthe value of the congestion window andRetransmission time-out chosen after the routerestoration.

References on ELFN• K. Chandran et al, “A feedback based scheme forimproving TCP performance in ad-hoc wirelessnetworks”, in Proceedigns of International Conferenceon Distributed Computing Systems, 1998.

• K.Chandran et al, ““A feedback based scheme forimproving TCP performance in ad-hoc wirelessnetworks”, in IEEE Personal Communications Magazine,February 2001.