Ad Hoc Wireless Routing

1
Ad Hoc Wireless Routing

• Different from routing in the wired world
• Desirable properties of a wireless routing
  protocol
• Distributed operation
• Loop freedom
• Demand-based operation
• Security
• Sleep period operation
• Unidirectional link support
• Corson M., Macker, J. Mobile Ad hoc Networking
  (MANET) Routing Protocol Performance Issues and
  Evaluation Considerations. IETF Internet Draft.
  http//www.ietf.org/internet-drafts/corson-draft-
  ietf-manet-issues-01.txt

2
Overview of Ad hoc Routing Protocols

• Globally precomputed, table based
• DSDV Destination-Sequenced Distance Vector
• WRP Wireless Routing Protocol
• GSR Global State Routing
• FSR Fisheye State Routing
• HSR Hierarchical State Routing
• ZHLS Zone-based Hierarchical Link State Routing
  Protocol
• CGSR Clusterhead Gateway Switch Routing Protocol

3
Overview of Ad hoc Routing Protocols

• On-Demand, source initiated
• AODV Ad Hoc On-demand Distance Vector Routing
• DSR Dynamic Source Routing
• TORA Temporally Ordered Routing Algorithm
• CBRP Cluster Based Routing Protocols
• ABR Associativity Based Routing
• SSR Signal Stability Routing

4
DSDV, DSR, AODV, TORA

• These four protocols were chosen for further
  study for several reasons
• Submitted to MANET for approval
• Implemented in ns-2
• Multiple performance studies have been done on
  these protocols

5
Dynamic Destination-Sequenced Distance Vector
(DSDV)

• C. Perkins and P. Bhagwat. Highly dynamic
  Destination-sequenced distance vector routing
  (DSDV) for mobile computers ACM SIGCOMM '94
  p234-244, 1994.
• Each node knows the state and topology of the
  entire network
• Routes are chosen by a metric (least delay, best
  signal strength, etc..)
• Periodically and when triggered transmits the
  entire routing table to neighbors
• Full dumps
• Incremental dumps
• Avoids loops by using sequence numbers

6
DSDV Recovery

• When a link loss is detected at node N
• the metric of the route to the destination
  through the lost link is advertised as infinity
  (the worst value), and
• An incremental update is flooded to the neighbors

7
DSDV Evaluation

• Loop avoidance
• Constant routing overhead versus mobility
• Overhead increases as the number of nodes
  increases
• DSDV can no longer find a route reliably when
  there is high mobility (lt 300s pause times)

8
Ad Hoc On-Demand Distance Vector (AODV)2

• C. Perkins. Ad Hoc On-Demand Distance Vector
  (AODV) Routing IETF Internet Draft.
  http//www.ietf.org/internet-drafts/draft-ietf-man
  et-aodv-10.txt.
• Each node only keeps next-hop information
• Source broadcasts ROUTE REQUEST packets
• Each node that sees the request and forwards it
  creates a reverse route to the source
• If the node knows the route to the destination,
  it responds with a ROUTE REPLY
• All nodes along the reply route create a forward
  route to the destination

9
AODV Recovery

• When a link loss is detected at node N
• any upstream nodes that have recently sent
  packets through this node are notified with an
  UNSOLICITED ROUTE REPLY with an infinite metric
  for that destination

10
AODV Evaluation

• Routing overhead increases as mobility increases,
  but not as the number of nodes increases
• Sends many packets, but they are small
• Costs to access the medium (RTS/CTS packets)
• Always delivers at least 95 of packets sent in
  all cases (Broch, et. al.)

11
Temporally Ordered Routing Algorithm (TORA)

• V. D. Park and M.S. Corson. A Highly Adaptive
  Distributed Routing Algorithm for Mobile Wireless
  Networks. Proceedings of INFOCOMM 97 April
  1997. http//www.ics.uci.edu/atm/adhoc/paper-coll
  ection/corson-adaptive-routing-infocom97.pdf.
• Discovers multiple routes to destination
• Separate logical copy of the algorithm for each
  destination exists on each node
• Creates a Directed Acyclic Graph with the
  destination as the head of the graph
• Requires IMEP (Internet MANET Encapsulation
  Protocol) guarantees reliable in-order delivery
  of routing messages

12
TORA (cont)

• Each node keeps a reference value and a height
  for each destination
• QUERY packets are sent out until one reaches the
  destination or a node with a route to the
  destination
• This node sends an update to its neighbors
  listing its height for that destination

13
TORA Recovery

• The node, N which discovers the link loss
• Does nothing because other routes still exist, or
• If the lost link is the last downstream link of
  this node
• changes its height to be the local maximum and
• transmits update packets to look for new routes

14
TORA Evaluation

• Can contain routing loops for short periods of
  time
• High routing overhead
• Does not try to find the shortest path
• When there are large numbers of sources
  transmitting simultaneously, TORA cannot find
  paths
• Congestion feedback loop
• When there is too much congestion, IMEP loses
  packets and tells TORA that the link is down
• TORA sends out more UPDATE packets to reconfigure
• More congestion is created

15
Dynamic Source Routing(DSR)

• David B. Johnson, Davis A. Maltz, The Dynamic
  Source Routing Protocol for Mobile Ad Hoc
  Networks October 1999. IETF Internet Draft.
  http//www.ietf.org/internet-drafts/draft-ietf-man
  et-dsr-10.txt.
• Routes are kept in each packet
• Routes to that point in REQUEST packets
• Full routes in data packets
• Routes are cached at each node to limit flooding
  of REQUEST packets
• Any route that is seen through a node is cached
• Source sends out REQUEST packets
• Any node which is the destination of a node which
  has a route to the destination replies with a
  route reply

16
DSR Recovery

• A Route ERROR is sent to the Source
• All nodes along the path remove that route
• Source uses a cached alternate route to
  destination or sends out request packets for a
  new route

17
DSR Evaluation

• Always delivers at least 95 of all packets sent
  in all cases (Broch, et. al.)
• Routing packets are large because of the source
  routing

18
Performance

• Broch, et al. A Performance Comparison of
  Multi-Hop Wireless Ad Hoc Network Routing
  Protocols MOBICOM 98 p85-97, 1998.
• Measurements are from simulations with 50 nodes,
  pause times from 0s (constant motion) to 900s (no
  motion), transmission rates of 4 packets/s, and
  speed of nodes at 1m/s and 20 m/s
• Load was 10 sources transmitting simultaneously,
  20 sources, and 30 sources
• Simulated in ns-2 on top of complete
  implementation of the 802.11 Medium Access
  Control (MAC) protocol Distributed Coordination
  Function (DCF)

19
PerformanceConvergence

• Convergence is the ability of the routing
  protocol to quickly stabilize the routes it
  knows
• DSR AODV always deliver at least 95 of the
  packets sent in all cases
• TORA fails to converge at less than 500s pause
  time for the 30 source experiments, but always
  converges for 10 and 20 sources.
• Congestion feedback loop
• DSDV does not converge with a pause time less
  than 300s when nodes are moving at 20 m/s

20
PerformanceRouting Overhead

• Broch, et al. A Performance Comparison of
  Multi-Hop Wireless Ad Hoc Network Routing
  Protocols MOBICOM '98 p85-97, 1998.

21
NS-2, JavaSim Evaluations

• Looking for
• Basic network/TCP implementation
• Ability to implement an application layer routing
  protocol (Gnutella)
• NS-2 is a popular network simulator
• JavaSim was recommended by a group doing similar
  research
• OMNet was pushed to the side because of the
  constant recompilation needed to run simulations

22
NS-2 and JavaSim Similarities

• Event driven simulators
• tcl like interface
• Capable of network simulation
• Outputs to NAM and xgraph formats
• Can also output to any format that's been
  programmed into it
• Bad documentation
• JavaSim no documentation other than javadoc
• ns-2 documentation does not match code

23
Differences

• ns-2 was designed as a network simulator
• Built in wireless medium support
• Uses octcl as an interface
• Does not handle application layer data well

• Designed as an all purpose simulator
• No known wireless support
• Uses jacl as an interface
• Full support for application layer data exchange

24
NS-2

• octcl interface is easy to use and set up
  simulations
• Code is confusing written in both C and tcl, no
  standard for what should be written in each
• octcl must be installed as a separate program

25
JavaSim

• Jacl interface is built-in to the simulator
• Jacl scripts are difficult to understand and not
  easy to set up simulations
• Can use actual Java applications as components