Transcript MobileMAN Architecture and Protocols

QuickTi me™ and a
GIF decompressor
are needed to see thi s pi ctur e.
National Research Council - Pisa - Italy
IST - FET Initiative
QuickTime™ and a
GIF decompressor
are needed to see this picture.
MobileMAN Architecture and Protocols
Marco Conti
Italian National Research Council (CNR)
IIT Institute
2nd MobileMAN Workshop - Sophie-Antipolis 6-7 March, 2003
1
Reference Architecture
2
Enhanced Legacy MANET Architecture
Existing Applications (SMS, chating, etc..)
Middleware Services location
TCP
Security (?)
IP layer
MANET Routing
algorithms
Power Management
Cooperation
Socket API
Wi-Fi cards
Enhanced Wi-Fi cards (?)
3
MANET approach: IP centric
The problem is mainly concentrated on the IP layer and is related to
“solve mobile routing problems at the IP layer”
the main reason for doing work at the IP layer is to better support
heterogeneity and networked interoperability of lower layer technologies
support for a heterogeneous mix of technologies and devices is one of
the great successes of IP
IP layer software development often capitalizes on the rich variety of
work done within IP protocol software stacks and operating systems
therefore easing development, deployment costs, and simplifying redesign
and replication efforts.
4
Ad Hoc network realistic view
Networked communication dynamics favor smaller wireless networks or at least
networks requiring fewer hops in order to access a fixed backbone
For example, in ad hoc networks employing omni-directional antennas, recent
research
Piyush Gupta and P. R. Kumar, ``The Capacity of Wireless Networks,'' IEEE Transactions on
Information Theory, vol. IT-46, no. 2, pp. 388-404, March 2000.
has shown that as the network scales (assuming it is equally likely that any pair of
network nodes wish to communicate) as the number of communicating network
nodes grows, the achievable throughput per node pair goes to zero
As such, the prospect of deploying very large-scale, ad hoc networks based on
broadcast transmissions is not very promising.
5
Ad Hoc network realistic view (cont.)
MANET scales:
small-scale (i.e., 2-20 nodes)
moderate-scale (i.e., 20-100 nodes)
large-scale (i.e., 100+ nodes)
very large-scale (i.e., 1000+ nodes
• Small-medium scale ad hoc islands
• Ad hoc islands interconnected to the Internet
6
Ad Hoc network realistic view (cont.)
MANET etherogeneity is limited
• 802.11
• Bluetooth (too limited coverage area)
• 802.11family is de facto the standard for WLANs
• Ad hoc networking can be successful only if sufficiently dense network exists
Our view of ad hoc network:
802.11 based small-medium scale ad hoc islands
• homogeneous
• small (medium) scale
Novel MobileMAN Architecture
7
Novel MobileMAN Architecture
The rationale to go in this direction:
Problems pointed out by measurement studies of simple
802.11 ad hoc networks:
Wi-Fi in ad hoc configurations has severe problems
TCP mechanisms made the Wi-Fi system more unstable
The 3-hop horizon for ad hoc networks made
interesting
to use simpler routing mechanisms
2, 2.5 Routing protocols, simplified TCP, cross-layers functions
8
Novel MobileMAN Architecture
Proxy-based architecture for
the connection to the Internet
The Ad Hoc network is a subnet
based on “ad hoc” protocols
Applications and Middleware
Socket API
Simplified Transport Protocol
Subnetwork Forwarding and
Routing
Wi-Fi or Enhanced Wi-Fi cards
Ad Hoc Node
Proxy daemon
Socket API
Socket API
Simplified Transport
Protocol
TCP
Subnetwork
Forwarding and
Routing
IP
Wi-Fi or
Subnetwork
Enhanced Wi-Fi
Proxy
9
Novel MobileMAN Architecture
Ad Hoc Node Interconnection
to Internet without proxy
Applications and Middleware
Socket API
TCP
IP
Simplified Transport Protocol
Subnetwork Forwarding and Routing
Wi-Fi or Enhanced Wi-Fi cards
10
Novel MobileMAN Architecture (Cont.)
Current solutions are typically not “optimized”:
solutions at one layer generally not exploit
information/mechanisms available at lower layers
Es. Localization services are designed independently by
mechanisms already available at lower layers
Our idea to exploit inter-layer coordination
11
Inter-layer coordination
Example: MAC - Network Layers coordination
Let us assume that the network layer through the
routing mechanism has a (partial) knowledge of the
network topology.
We can also assume that this knowledge is typically
accurate for a 1-2 hop distance
The MAC layer could exploit this information to reduce
the impact of phenomena such as: exposed nodes,
capture phenomena etc.
12
Inter-layer coordination (cont.)
Example: MAC - Network Layers coordination (Cont.)
On the other hand knowledge available at MAC layer
can be used to select among alternative routes
It might be worth avoiding one route because a link is
currently blocked at MAC level (e.g., through the
RTS/CTS mechanism, etc.)
13
Inter-layer coordination (cont.)
Example: Cooperation /Forwarding/Routing coordination
Let us assume that the network layer through the
routing mechanism has a (partial) knowledge of the
network topology. This may include
•
Alternative routes for a given destination
•
Other info on these routes (e.g., number of hops)
The Cooperation layer by using these info can try to
classify the reliability/performance/cooperation along
these routes. A new nearness metric can be derived
14
Inter-layer coordination (cont.)
Example: Cooperation /Forwarding/Routing coordination (cont.)
Packet Forwarding can exploit the closeness metric
• to select among alternative paths
• to perform a load balancing among the routes
The feedbacks from this choice can be used to refine
the nearness metric
15
Inter-layer coordination (cont.)
Example: Forwarding/Cooperation/Transport coordination (cont.)
The transport layer might use a redundant coding technique
(see, e.g., FEC): a message is coded into n blocks but only k
(k<n) need to be received. The redundancy level can be
selected taking into account the reliability metric at the
Cooperation layer
Blocks can be forwarded on alternative links taking into
consideration the closeness/reliability metric
Finally, the results of this transmissions can be used to
refine the reliability (nearness) metric.
16
Novel MobileMAN Architecture
Shared memory
Middleware Services location
Socket API
Simplified Transport Protocol
Forwarding
Users
location
Routing (2)
Power Management
Cooperation/ Performability
Treasure hunting (with code mobility)
Enhanced Wi-Fi cards
17