Introduction to Wireless Ad

advertisement
Introduction to Wireless
Ad-Hoc Networks Routing
Michalis Faloutsos
Some slides borrowed
From Guor-Huar Lu
1
Outline
Challenges
Design Goals Specified by MANET (for now…)
Types of Routing
Protocols in Detail
Conclusion
2
Challenges
Dynamic Topologies
Bandwidth-constrained, variable capacity links
Energy-constrained
Limited Physical security
Scalability
3
Types of routing
Flat Proactive Routing
• Link state: Fish-Eye Routing, GSR, OLSR.
• Table driven: Destination-Sequenced Distance Vector (DSDV), WRP)
On-Demand or Reactive Routing
• Ad hoc On-demand Distant Vector (AODV)
• Dynamic Source Routing (DSR)
Hybrid Schemes
• Zone Routing ZRP, SHARP (proactive near, reactive long distance)
• Safari (reactive near, proactive long distance)
Geographical Routing
Hierarchical: One or many levels of hierarchy
Routing with dynamic address
• Dynamic Address RouTing (DART), L+
4
Proactive Protocols
Proactive: maintain routing information
independently of need for communication
Update messages send throughout the network
periodically or when network topology changes.
Low latency, suitable for real-time traffic
Bandwidth might get wasted due to periodic
updates
They maintain O(N) state per node, N = #nodes
5
On-Demand or Reactive Routing
Reactive: discover route only when you need it
Saves energy and bandwidth during inactivity
Can be bursty -> congestion during high activity
Significant delay might occur as a result of route
discovery
Good for light loads, collapse in large loads
6
Hybrid Routing
Proactive for neighborhood, Reactive for far
away (Zone Routing Protocol)
Proactive for long distance, Reactive for
neighborhood (Safari)
Attempts to strike balance between the two
7
Hierarchical Routing
Nodes are organized in clusters
Cluster head “controls” cluster
Trade off
• Overhead and confusion for leader election
• Scalability: intra-cluster vs intercluster
One or Multiple levels of hierarchy
8
The effect of hierarchy
Consider a protocol of f(N) complexity
If we apply it in a two level hierarchy with C
clusters
• Complexity within clusters f(Nc)
• Complexity between clusters f(N/Nc)
How should we pick cluster sizes?
Does a two-level hierarchy work?
9
Geographical Routing
Nodes know their geo coordinates (GPS)
Route to move packet closer to end point
Protocols DREAM, GPSR, LAR
Propagate geo info by flooding (decrease
frequency for long distances)
10
Geographical Routing
destination
source
Knowing GPS
coordinates
Pick next hope based on
angle of deviation
Consider other params
• Distance of next hop
• Congestion
• Reliability
Q: How do you know
location of destination?
11
Dynamic Routing: a new approach
DART Ericsson et al., L+ Morris et al
Goal: can we enforce address aggregation
But: nodes are moving
Then: address should change
12
Dynamic Routing: general idea
Separation of identity and address
• Identity is who you are
• Address is where you are
Rule for enforcing “structure” in addresses:
• near by nodes should have nearby addresses
Using the Rule, we can “aggregate” information
13
DART: in more detail
Basic idea: permanent nodeID =/= transient address
The address reflects network location
It is a proactive routing scheme, distance vector
Consequences:
• Routing is simplified: address tell me where you are
• Nodes with similar addresses are “near” each other
Challenges:
• Address allocation: When I move, change my address
• ID to Address mapping: Given an ID, find the address
14
Some more theoretical issues
15
Network Capacity
The optimal capacity of a wireless
network is
Where N nodes, and C channel
capacity
Gupta Kumar paper “The capacity of wireless networks”
What other assumptions do I need?
Transmission range: as small as I need to
Who talks to whom? Random selection
Non optimal traffic patterns and fixed transmissions
then throughput is : O(C/ sqrt(N log N))
16
Intuitive Explanation
Explanation: N nodes in the field
Destinations are random
On average N^0.5 hops per path
Each node has N^0.5 paths go through
17
Mobility increases capacity
Grossglausser and Tse (infocom 2001)
Statement: if nodes move they will eventually carry the
info where you want
Protocol:
• sender send one copy to receiver or one neighbor
• Sender and relay will at some run into destination and send the
packet
All paths are at most two hops
They show that the capacity of the network does not go
to zero
Tradeoff?
18
Hierarchical routing: bounds
Cluster nodes, and route between and within clusters
Location management: finding where is the node
Routing finding how to get there
Multiple levels: log(N) levels
Location Mgm: Each nodes stores O(N) locations
Routing overhead: O(log^3N)
Dominating factor: location management and not the
routing
Location mgmt handoff: O(log^2N)
See Susec Marsic, infocom 02
19
20
Types of routing
Flat Proactive Routing
• Link state Fish-Eye Routing, GSR, OLSR.
• Table driven: Destination-Sequenced Distance Vector (DSDV), WRP)
On-Demand or Reactive Routing
• Ad hoc On-demand Distant Vector (AODV)
• Dynamic Source Routing (DSR)
Hybrid Schemes
• Zone Routing ZRP, SHARP (proactive near, reactive long distance)
• Safari (reactive near, proactive long distance)
Geographical Routing
Hierarchical: One or many levels of hierarchy
Routing with dynamic address
• Dynamic Address RouTing (DART)
21
Proactive: DSDV - Destination-Sequenced
Distance Vector Algorithm
By Perkins and Bhagvat
Based on Bellman Ford algorithm
• Exchange of routing tables
• Routing table: the way to the destination, cost
Every node knows “where” everybody else is
• Thus routing table O(N)
Each node advertises its position
• Sequence number to avoid loops
• Maintain fresh routes
22
DSDV details
Routes are broadcasted from the “receiver”
• Nodes announce their presence: advertisements
Each broadcast has
• Destination address: originator
• No of hops
• Sequence number of broadcast
The route with the most recent sequence is used
23
Reactive: Ad-Hoc On-demand Distance
Vector Routing (AODV)
By Perkins and Royer
Sender tries to find destination:
• broadcasts a Route Request Packet (RREQ).
Nodes maintain route cache and use destination
sequence number for each route entry
State is installed at nodes per destination
Does nothing when connection between end points is
still valid
When route fails
• Local recovery
• Sender repeats a Route Discovery
24
Route Discovery in AODV 1
2
7
5
Source 1
3
4
8
Destination
6
Propagation of Route Request (RREQ) packet
25
Route Discovery in AODV 2
2
7
5
Source 1
3
4
8
Destination
6
Path taken by Route Reply (RREP) packet
26
In case of broken links…
Node monitors the link status of next hop in
active routes
Route Error packets (RERR) is used to notify
other nodes if link is broken
Nodes remove corresponding route entry after
hearing RERR
27
Dynamic Source Routing (DSR)
Two mechanisms: Route Maintenance and Route
Discovery
Route Discovery mechanism is similar to the one
in AODV but with source routing instead
Nodes maintain route caches
Entries in route caches are updated as nodes
learn new routes.
Packet send carries complete, ordered list of
nodes through which packet will pass
28
When Sending Packets
Sender checks its route cache, if route exists,
sender constructs a source route in the packet’s
header
If route expires or does not exist, sender initiates
the Route Discovery Mechanism
29
Route Discovery 1 (DSR)
2
<1,2>
<1>
7
<1,3>
5
<1>
Source 1
<1,3,5,7>
<1,3,5>
3
8
<1>
Destination
<1,4,6>
6
4
<1,4>
Building Record Route during Route Discovery
30
Route Discovery 2 (DSR)
2
7
5
Source 1
Destination
3
8
<1,4,6>
<1,4,6>
6
4
<1,4,6>
Propagation of Route Reply with the Route Record
31
Route Maintenance
Two types of packets used: Route Error Packet and
Acknowledgement
If transmission error is detected at data link layer, Route
Error Packet is generated and send to the original sender
of the packet.
The node removes the hop is error from its route cache
when a Route Error packet is received
ACKs are used to verify the correction of the route links
Nodes use caching: to reduce overhead.
32
The Zone Routing Protocol (ZRP)
Hybrid Scheme
Proactively maintains routes within a local
region (routing zone)
Also a globally reactive route query/reply
mechanism available
Consists of 3 separate protocols
Protocols patented by Cornell University!
33
Intrazone Routing Protocol
Intrazone Routing Protocol (IARP) used to
proactively maintain routes in the zone.
Each node maintains its own routing zone
Neighbors are discovered by either MAC
protocols or Neighbor Discovery Protocol (NDP)
When global search is needed, route queries are
guided by IARP via bordercasting
34
Interzone Routing Protocol
Adapts existing reactive routing protocols
Route Query packet uniquely identified by
source’s address and request number.
Query relayed to a subset of neighbors by the
bordercast algorithm
35
Comparisons 1
Things in common:
•
•
•
•
IP based operation
Distributed operation
Loop-free routing
Very little or no support for sleep period operation
and security
36
Comparisons 2
FSR
DSDV
AODV
DSR
ZPR
Source
Routing
No
No
Yes
No
Periodic
message
Yes
No
No
Yes
(Locally)
Functioning
Proactively
Yes
No
No
Yes
(Locally)
Functioning
Reactively
No
Yes
Yes
Yes
(Globally)
37
Performance?
End-to-end data throughput and delay
Route acquisition time
Percentage of out-of-order delivery
Efficiency:
• Average number of data bits transmitted/data bits delivered
• Average number of control bits transmitted/data bits delivered
• Average number of control and data packets transmitted/data
packet delivered
38
Parameters
Network Size
Channel properties
Connectivity (average degree of a node)
Topology rate of change
Link capacity (bps)
Directionality: Fraction of unidirectional links
Traffic patterns: who talks to whom
Mobility
Fraction/frequency of sleeping nodes
39
References
Mobile Ad hoc Networking (MANET): Routing Protocol Performance
Issues and Evalution Considerations (RFC 2501)
P. Misra., “Routing Protocols for Ad Hoc Mobile Wireless Networks”,
http://www.cis.ohio-state.edu/~jain/cis788-99/adhoc_routing/
The Zone Routing Protocol (ZRP) for Ad Hoc Networks <draft-ietf-manetzone-zrp-04.txt>
Fisheye State Routing Protocol (FSR) for Ad Hoc Networks <draft-ietfmanet-fsr-03.txt>
Ad hoc On-demand Distance Vector (AODV) Routing <draft-ietf-manetaodv-11.txt>
The Dynamic Source Routing Protocol for Mobile Ad Hoc Networks (DSR)
<draft-ietf-manet-dsr-07.txt>
40
Conclusion
On-demand routing protocols (AODV and DSR) are well
established.
More analysis and features are needed (Performance
comparison between protocols, QoS extension and
analysis, multicast, security issues etc…)
Good paper (though old):
A review of current routing protocols for ad-hoc mobile wireless networks, E.
Royer, C.K. Toh
Routing Scalability in MANETs, Eriksson et al. Book chapter,
http://www.cs.ucr.edu/~michalis/COURSES/240-08/lectures/bookchapter.pdf
41
42
Fisheye State Routing (FSR)
Node stores the Link State for every destination in the
network
Node periodically broadcast update messages to its
neighbors
Updates correspond to closer nodes propagate more
frequently
43
Multi-Level Scope (FSR)
Central node (red dot)
has the most accurate
information about
nodes in white area and
so on.
•Parameters: Scope
level/radius size
•
44
ZPR architecture
ZRP
NDP
IARP
IERP
ICMP
BRP
IP
A
B Information passed from protocol A to B
A
B Exchange of packets between protocol A and B
45
Design Goals
Peer-to-peer mobile routing capability in mobile,
wireless domain.
Intra-domain unicast routing protocol:
• Effective operation over a wide range of mobile networking
scenarios and environments
• Supports traditional, connectionless IP services
• Efficiently manages topologies changes and traffic demands
46
Desired properties
Distributed operation
Loop freedom
Demand-based operation
Proactive operation
Security
“Sleep” period operation
Unidirectional link support
47
Download