Henning Schulzrinne with Arezu Moghadam Suman Srinivasan Jae Woo Lee and others Columbia University WINLAB 20th December 2009 Challenges for years 2039 Changing usage H2H ID: 565432
Download Presentation The PPT/PDF document "The Future of Wireless: Reaching the Unr..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
The Future of Wireless: Reaching the Unreachable and Adaptive Wireless Networks
Henning Schulzrinne(with Arezu Moghadam, Suman Srinivasan, Jae Woo Lee and others)Columbia University
WINLAB 20th - December 2009Slide2
Challenges for years 20...39
Changing usage: H2H M2MMore than just first-mile accessUser-focused designInterconnecting mobile serviceCovering the white spotsWINLAB 20th - December 2009Slide3
Wireless networks now
WINLAB 20th - December 2009Slide4
Emerging wireless applications
WINLAB 20th - December 2009Slide5
Changing usage
WINLAB 20th - December 2009
voice
web
M2MSlide6
More than just Internet Classic
Networkwirelessmobilitypath stabilitydata
units
Internet “classic”
last hop
end systems
>
hours
IP
datagrams
mesh networks
all links
end systems
> hours
mobile ad-hoc
all links
all nodes, random
minutes
opportunistic
typical
single node
≈
minute
delay-tolerant
all links
some
predictable
some
predictable
bundles
store-carry-forward
all nodes
all nodes
no
path
application data unitsSlide7
Reaching the unreachable
WINLAB 20th - December 2009Slide8
White spaces (real world)
WINLAB 20th - December 2009
$60 for 5 GB
$12/GBSlide9
Internet
?
?
D
Contacts
are
opportunistic
intermittent
802.11 ad-hoc mode
BlueToothSlide10
Web Delivery Model
7DS core functionality: Emulation of web content access and e-mail deliverySlide11
Search Engine
Provides ability to query locally for resultsSearches the cache index using Swish-e libraryStores query for future contactsSlide12
Email exchangeSlide13
BonAHA framework
Node 2Node 1
key21 = value21
key22 = value22
key23 = value23
key24 = value24
key11 = value11
key12 = value12
key13 = value13
key14 = value14
[2] node1.get(key13)
[1] node1.register()
[3] data =
node1.fileGet(
value13);
BonAHA
[CCNC 2009]Slide14
Generic service model?
Opportunistic Network Framework – get(), set(), put(), rm()
ZigBee
BlueTooth
mDNS/
DNS-SD
DHTs?
Gnutella?
ApplicationSlide15
Bulletin Board System
Written in Objective-C, for iPod TouchSlide16
Local MicrobloggingSlide17
Problem – lack of group
communication model for mobile DiTNs?Any cast communication modelEmergencies Traffic congestion notifications Severe weather alerts Traditional multicast as a group communication model
Fails!
No knowledge of the topology
No infrastructure to track group memberships
Communication
with communities of
interest
Even a harder problem!
Market news, sport events
Scientific articles
Advertisement about particular
products
Epidemic routingSlide18
Interest-aware Communication
Jazz
Jazz
Jazz
Rock
Rock
Communication with communities of interest
Interest-aware
m
usic
s
haring
a
pplicationSlide19
UI of Interest-Aware Music and News Sharing Application for 7DSSlide20
Problem 1 of interest-aware: Greedy!
SX
Y
Y
D
1
1
1
3
3
3
3
wireless contact
data transfer
Y
a
b
c
d
e
f
g
2
D
4
4
D
D
X
X
X
Y
h
5
DSlide21
Energy issues
Interest-aware algorithms transmit until end of contact Battery life remains a problem for mobile devices!
Source: TIAX, portable power conferenceSlide22
Solution – PEEP
Still interest-awareInterest vectors; binaryLearning interests: feedback from user, # data items of each category, play times for music files, or LSATransmit-budgetAmount of data items allowed for transmission at each connectionHow to divide the transmit budget? PopularityShould be estimated
1
2
Items of interest?
Others?
1
0
0
1
1
1
0Slide23
Criteria to assign budget?
Only interest-awareMight waste budgetInterest-aware + randomly selectedInterest-aware + popularity estimationIdeal case: we know the global popularityBudget designation (e.g., 50%)
1
2
Items of interest
1
2
Items of interest
random
1
2
Items of interest
popular
1
2
interests
popularSlide24
Popularity estimation
Contact window NHistory of the users’ interestsAverage or weighted averageExample: C=6, N=8Replace the oldest
1
0
1
0
0
1
1
0
0
1
1
1
0
1
0
0
0
0
1
0
0
1
0
0
0
0
1
0
0
0
0
1
0
0
0
0
1
1
0
0
0
0
1
0
1
0
0
0
.62
.37
.37
.25
.12
.25Slide25
Evaluation of PEEPSlide26
Adaptive networks
WINLAB 20th - December 2009Slide27
Spectrum management
WINLAB 20th - December 2009What happens at field level makes the spectrum even tighter. "Stop and consider," said Mendelsohn, "that each coach on the field has a beltpack with four frequencies per pack, with about 10 coaches per team. Then the quarterbacks have two per pack. That's 42 frequencies for each team right there; so with two teams, that's about
84
frequencies." But that's hardly all. "Then add another 15 frequencies for the referees, the chain gang and security frequencies. That's
99
— before counting the TV broadcasters, which require
40
frequencies each, minimum," he said. "Then there are another
15
for home and away radio, and
20
more for various broadcasters doing stand-ups before and after the game. "And what most people forget about is,"
Mendelsohn
said, "that all of this RF is basically contained within and around just 100 yards."
http://www.tvtechnology.com/article/90772
Steve
Mendelsohn
, game day frequency coordinator for the NFL. Slide28
Spectrum
WINLAB 20th - December 2009http://
www.ntia.doc.gov/osmhome/allochrt.pdfSlide29
But often lightly used
29http://www.sharedspectrum.com/measurements/
NYC, August 2004Slide30
Cognitive radio is insufficient
Solution: Cognitive radio! ?Doesn’t help with dense applicationslong time scales (hours days)(geographic database solution seems most likely)each frequency still inefficiently used
automated sharing on shorter time scales
WINLAB 20th - December 2009Slide31
Mobile applications
WINLAB 20th - December 2009Slide32
Mobile why’s
Why does each mobile device need its own power supply?Why do I have to adjust the clock on my camera each time I travel?Why do I have to know what my IMAP server is and whether it uses TLS or SSL?Why do I have to “synchronize” my iPhone
?
Why do I have to manually update software?
Why
do we use USB memory sticks when all laptops have 802.11b?Slide33
Oct. 2007
33Context-aware communicationcontext = “the interrelated conditions in which something exists or occurs”anything known about the participants in the (potential) communication
relationship
time
at current location of destination
capabilities
audio, video, text, …
location
location-based call routing
location events
activity/availability
rich presence
automotive safety
sensor data (mood,
physiometric)
medical monitoringSlide34
Examples of “invisible” behaviorSlide35
Usability: Interconnected devices
any weather service
school closings
opens
(home, car, office) doors
incoming call
generates TAN
acoustic alerts
updates location
time, location
alert, events
address bookSlide36
Conclusion
Focus shifting: speed to diversity, functionality, autonomic behaviorApplications beyond voice and webmore than “Internet of things” & sensor networksSeamless user experience across cellular, WLAN & disruption-tolerant networksWINLAB 20th - December 2009Slide37
Backup slides
WINLAB 20th - December 2009Slide38
Deploying services
WINLAB 20th - December 2009
NetServ
Shared
hosting
Cloud
c
omputing
Dedicated
hosting
Colocation
Own
data center
Unit
Java task
VM
/html
server
rack
100s of racks
Provided
computation
storage
network
power
AC
computation
network
power
AC
web
server
network
power
AC
computation
storage
network
power
AC
network
power
AC
setup time
seconds
minutes
hours
day
week
years
cost
?
$1/hour
$0.10/GB
$0.10/GB-month
$20/month
$100/month
$550+/rack
$10M/yearSlide39
Networks beyond the Internet
Network modelroute stabilitymotion of data routersInternet
minutes
unlikely
mobile ad-hoc
3
τ
disruptive
store-carry-forward
< 3
τ
helpfulSlide40
Destination/delivery mode
Destination/delivery mode
Multicast
Anycast
Unicast
Interest-driven
Location-driven
Person
Location-driven
Any node that meets
conditions
e.g., any AP or
infostation to upload
Messages
7DS message delivery
Geographic routing
GeOpps
Community-
based routing
Interest-aware
communication
Geographic
routing
GeOpps
GeoDTN+Nav
Oracle-based
EBR
MaxProp
Prophet
Spray and wait
BUBBLE
SimBetSlide41
Depth and breadth
Depth and breadth
Two-hops / Source routing
More than two hops /
Per-hop routing
Single copy
Multiple copies
One-hop
Direct delivery
between a sender and a receiver
Single link
Multiple links
Flooding
Epidemic routing,
MaxProp
Shortest path
Oracle-based
Several possible paths
Oracle-based
GeOpps
GeoDTN+Nav
Prophet
SimBet
Spray and wait
EBR
BUBBLESlide42
Knowledge
Knowledge
Zero knowledge
Deterministic information
Temporal information
Spatial information
Route/destination-invariant
Mobility pattern
randomized routing
Epidemic routing
Spray and wait
7DS message delivery
Bus, train
Oracle-based
Probabilistic information
Popularity/centrality
Time-varying, dynamics are known
Time-invariant
Route-varying,
Destination- invariant
Satellite
Oracle-based
Satellite
GeOpps
GeoDTN+Nav
Oracle-based
Personal relationship
Route/destination location varying
Prophet
MobySpace
EBR
BUBBLE
SimBet
Navigation system
GeoDTN+Nav
MaxProp
Prophet