Jul 14, 2005 ... Providing access to Internet resources (email, web). Leveraging intermittent
connectivity and ad-hoc communications. ▻ Daknet in Cambodia ...
HELSINKI UNIVERSITY OF TECHNOLOGY NETWORKING LABORATORY
Many RealMANs will be disconnected – The Need for DTN in Ad-hoc Networking
2005-07-14
RealMAN Workshop 2005
Jörg Ott
[email protected]
1
© 2005 Jörg Ott
HELSINKI UNIVERSITY OF TECHNOLOGY NETWORKING LABORATORY
Some “Extreme” Ad-hoc Scenarios Sensornets y Sparse distribution of network nodes y Rare occasions for retrieving data
` ZebraNet: Wildlife monitoring in Kenia ` Whales and seals for water monitoring ` Data mules / message ferries picking up data
Access to information resources from remote areas y Providing access to Internet resources (email, web) y Leveraging intermittent connectivity and ad-hoc communications
` Daknet in Cambodia and India ` SNC in Lapland ` Wizzy in South Africa © 2005 Jörg Ott
2
HELSINKI UNIVERSITY OF TECHNOLOGY NETWORKING LABORATORY
Some “Extreme” Ad-hoc Scenarios Sensornets y Sparse distribution of network nodes or retrieving scheduled yOpportunistic Rare occasions for data contacts
between ad-hoc nodes. End-to-end paths may never exist. ZebraNet: Wildlife monitoring in Kenia
` ` Whales and seals for water monitoring Need to perform asynchronous (store-and-forward) ` Data mules / message ferries picking up data
communications. Not limited to information exchange in exotic areas, though…
Access to information from remote Mobileresources information access forareas nomadic users: y Providing access to Internet resources (email, web) • Plentyconnectivity of personal devices around y Leveraging intermittent and ad-hoc communications
• Interesting scenarios for ad-hoc collaboration to plain ad-hoc networking, but…
` Daknet in Cambodia and India ` SNC in Lapland Lends itself ` Wizzy in South Africa © 2005 Jörg Ott
3
HELSINKI UNIVERSITY OF TECHNOLOGY NETWORKING LABORATORY
Limitations to Connectivity in RealMANs ` Willingness to cooperate y Resource constraints y Mutual distrust (spam, worms, viruses, DoS attacks, but also simple data protection)
` Capability to interoperate y Heterogeneity: independently managed, diverse devices with different feature sets Variety of suppliers for PDAs, cellphones, hybrids Different application support, operating systems even per vendor
` Availability of peer nodes y Deployment / penetration y Density of users and devices turned on y Mobility (velocity, etc.)
` Radio range + channel contention (use for different purposes) © 2005 Jörg Ott
4
HELSINKI UNIVERSITY OF TECHNOLOGY NETWORKING LABORATORY
Example 1: PDA Users
5
© 2005 Jörg Ott
HELSINKI UNIVERSITY OF TECHNOLOGY NETWORKING LABORATORY
Example 1: PDA Users
© 2005 Jörg Ott
6
HELSINKI UNIVERSITY OF TECHNOLOGY NETWORKING LABORATORY
Example 1: PDA Users Some heavy handwaving: 1.3 million PDAs were sold in Europe in Q1/2005. 20.8% market share of the Blackberry yielding 270,000 pieces. Assume all of them are in Germany: this means 0.3% of the German population bought a new Blackberry in early 2005. Berlin has 3.4 M inhabitants living on some 900 km2. This makes 10,200 Blackberries in Berlin or some 11 per km2.
7
© 2005 Jörg Ott
HELSINKI UNIVERSITY OF TECHNOLOGY NETWORKING LABORATORY
Example 1: PDA Users Some heavy handwaving: 1.3 million PDAs were sold in Europe in Q1/2005. 20.8% market share of the Blackberry yielding 270,000 pieces.
• Infrastructure networks to augment connectivity • Ad-hoc networks asGermany: stubs this means 0.3% of the Assume all of them are in German early 2005. • BUT:population Issue of bought accessa new costBlackberry added tointhe equation • Always-on is usually not an option
Berlin has 3.4 M inhabitants living on some 900 km2. This makes 10,200 Blackberries in Berlin or some 11 per km2.
© 2005 Jörg Ott
8
HELSINKI UNIVERSITY OF TECHNOLOGY NETWORKING LABORATORY
Example 2: Mobile Internet Access in the Car ` 12,044 km of Autobahn today (growing, of course ;-) y ~700 service areas (every 18 km on average) y Typically spaced at 40–60 km distance, much denser around urban areas
` Providing WLAN access of ~1–2 km at each will not suffice y Establish connectivity via additional access networks (UMTS, satellite) Always-on, always best connected (expensive, yet not ubiquitous either)
y Support relaying by other vehicles (pioneering work by FleetNet) For 18 km distance, quite a few hops needed
© 2005 Jörg Ott
9
HELSINKI UNIVERSITY OF TECHNOLOGY NETWORKING LABORATORY
Example 2: Mobile Internet Access in the Car ` 1 Jan 2005: 45 M person cars, 3.25 M new in 2004 y ~42% market share by German manufacturers y ~35% new German mid-size cars and above in 2005 (1.15 M) 2.5% of the total person cars
y Light daytime traffic on the autobahn: 6 cars per lane and kilometer 3 lanes, no lorries: 18 cars per kilometer Î < 0.5 new cars per kilometer and direction
© 2005 Jörg Ott
10
HELSINKI UNIVERSITY OF TECHNOLOGY NETWORKING LABORATORY
Example 2: Mobile Internet Access in the Car ` Sometimes (if you are lucky), there may be no other cars at all…
11
© 2005 Jörg Ott
HELSINKI UNIVERSITY OF TECHNOLOGY NETWORKING LABORATORY
Delay-tolerant Networking (DTN) ` Concept for asynchronous (store-and-forward) communications y Modeled following email exchange
` Federation of arbitrary networks interconnected by “DTN layer” y Above transport, global structured addressing scheme
` Communication by means messages (bundles) of arbitrary size y Hop-by-hop forwarding through DTN bundle routers y Along links during contacts (permanent, scheduled, opportunistic, …) No end-to-end path required due to persistent storage in bundle routers
y Different degrees of reliability (e.g. custody transfer vs. best effort) Optional end-to-end semantics (e.g. delivery notification)
y Different routing schemes conceivable y Complex security implications © 2005 Jörg Ott
12
HELSINKI UNIVERSITY OF TECHNOLOGY NETWORKING LABORATORY
Dealing with Intermittent Connectivity: Drive-thru Internet Application Clients + Drive-thru Client
Application Servers Drive-thru Proxies
Connectivity Islands (WLAN Hot Spots)
ISP A
Internet ISP B
13
© 2005 Jörg Ott
HELSINKI UNIVERSITY OF TECHNOLOGY NETWORKING LABORATORY
Dealing with Intermittent Connectivity: Drive-thru Internet Application Clients + Drive-thru Client
Application Servers Drive-thru Proxies 18000
30–90 s 20–70+ MB
16000 Data rate (kbit/s)
14000 12000
Connectivity Islands 10000 (WLAN Hot Spots)
8000
ISP A
6000
Internet
4000 2000
ISP B
0 0
200
400
600
800
1000 1200 1400
1600 1800
Distance (m)
© 2005 Jörg Ott
14
HELSINKI UNIVERSITY OF TECHNOLOGY NETWORKING LABORATORY
Dealing with Intermittent Connectivity: Drive-thru Internet Application Clients + Drive-thru Client
Application Servers Drive-thru Proxies 18000
30–90 s 20–70+ MB
16000 Data rate (kbit/s)
14000 12000
Connectivity Islands 10000 (WLAN Hot Spots)
Works fine but… 8000 ISP A • Needs direct 6000connection to an access link Internet 4000 • Does not support direct inter-vehicle communications 2000 B • Suboptimal routing due toISP indirection via Drive-thru proxy 0 0 200 400 600 800 1000 1200 1400 1600 1800 • Still “end-to-proxy” path requiredDistance (m) 15
© 2005 Jörg Ott
HELSINKI UNIVERSITY OF TECHNOLOGY NETWORKING LABORATORY
Drive-thru Internet and DTN
Mail Server
Mail Client Mobile Application Proxy
© 2005 Jörg Ott
1 DTN Router
DTN Router
Fixed Application Proxy
16
HELSINKI UNIVERSITY OF TECHNOLOGY NETWORKING LABORATORY
Drive-thru Internet and DTN
2
Internet Application
Internet Application Mobile Application Proxy
DTN Router
DTN Router
Fixed Application Proxy
17
© 2005 Jörg Ott
HELSINKI UNIVERSITY OF TECHNOLOGY NETWORKING LABORATORY
Conclusion ` Multi-hop networks involving humans (+ an existing market place) y Difficult to manage, too many influencing factors -> heterogeneity
` Deployment aspects require dealing with intermittent connectivity ` DTN provides flexible abstraction for effectively exploiting contacts y For many types of mobile applications
` Interesting open issues y Routing protocols for opportunistically connected links y Security requirements (gaining connectivity vs. many reservations) y Mapping to applications needed © 2005 Jörg Ott
18
UNIVERSITY OF HELSINKI UNIVERISITY OFTECHNOLOGY TECHNOLOGY NETWORKING LABORATORY
www.drive-thru-internet.org
© 2005 Jörg Ott
19