Spray and Wait: An Efficient Routing Scheme for Intermittently Connected Mobile Networks

1
Spray and Wait An Efficient Routing Scheme for
Intermittently Connected Mobile Networks
Thrasyvoulos Spyropoulos, Kostantinos Psounis,
and Cauligi S. Raghavendra EE Department,
USC spyropou, kpsounis, raghu_at_usc.edu
2
Routing in Intermittently Connected Mobile
Networks (ICMN)
Network Characteristics
Routing

• Network is sparse and partitioned
• Nodes follow stochastic mobility model
• Mobility is not enforced (e.g. Zhao et al. 04)
• Mobility is not predictable (e.g. Jain et al. 04)

• Exploit node mobility
• store-carry-and-forward?
• Replicate message
• send multiple copies
• to whom and when?

1
12
D
13
S
14
2
16
11
15
3
7
4
5
8
10
3
Existing Proposals

• Flooding everyone gets a copy (Epidemic Routing
  - Vahdat et al. 00)
• Note optimal delay only when traffic is very
  low!
• Reducing the overhead of flooding
• Randomized Flooding (Y. Tseng et al. 02)
  handover a copy with probability p lt 1
• Utility-based Flooding (A. Lindgren et al. 03)
  handover a copy to a node with a utility at least
  Uth higher than current
• Can use p and Uth to tradeoff transmissions for
  delay, BUT

Dilemma low p / high Uth? significant delay
increase high p / low Uth? degenerates to
flooding
4
Existing Proposals (contd)

• Single-copy solutions (Spyropoulos et al. 04)
• Only one copy per message at any time
• Randomized, utility-based, hybrid, etc.
• Significantly reduced transmissions, BUT high
  delay
• Summarizing
• No existing protocol has both low transmissions
  and low delay!

5
Efficient Routing Design Goals

• Performance goals
• perform significantly fewer transmissions than
  flooding-based schemes under all conditions
• better delay than existing single and multi-copy
  schemes close to optimal
• Additional goals
• scalability good performance under a wide range
  of values for various parameters (e.g. number of
  nodes)
• simplicity require little knowledge about the
  network

6
Outline

• Our Approach Spray and Wait
• Optimizing Spray and Wait
• Simulations
• Conclusion

7
Spray and Wait

• Redundant copies reduce delay
• Too much redundancy is wasteful and often
  disastrous!
• Spray and Wait send only L copies
• Spray phase spread L message copies to L
  distinct relays
• Wait phase wait until any of the L relays
  finds the destination (direct transmission)

• Important questions to be answered
• How are copies distributed? How many?
• What is the effect on delay?
• Can all our design goals be met? (e.g.
  scalability)

8
Spraying Matters

• Source Spraying Slowest
• source distributes all L copies one by one
• Binary Spraying Optimal
• source starts with L copies
• whenever a node with n gt 1 copies finds a new
  node, it hands over half of the copies that it
  carries
• proof of optimality see paper
• intuition when movement is I.I.D., any two nodes
  will find on average an equal number of potential
  relays in the same amount of time

9
Spraying Matters!
(analysis)
100x100 network with 100 nodes

• Efficient spraying becomes more important for
  large L
• Few copies suffice to achieve a delay only 2x the
  optimal!

10
Delay of Spray and WaitAn Upper Bound

• Assume independent random walks/random waypoint
  and no contention
• To keep things simple
• Some DTN applications are close to this model
• e.g. taxis forming a content distribution
  network for exchanging traffic conditions, clips,
  etc.
• Exact delay can be calculated using a system of
  recursive equations, but is not in closed form
  Derive a bound

Probability a wait phase is needed
Wait Phase
Spray Phase
M number nodes, L number of copies
11
Performance of Spray and WaitDelay of Wait Phase
Expected Meeting Time (from stationarity)
D
EDdt expected delay of Direct Transmission,
which is known (Spyropoulos et al. 04)
S
L relays
12
Performance of Spray and WaitDelay of Spray Phase

• M nodes
• L copies

D
S
Tight if L ltlt M
13
Choosing The Right Number of Copies (L)
Minimum L such that EDsw a EDopt

• Method 1) Calculate L from bound
• Method 2) Approximate HM-L (Taylor series) and
  solve resulting 3rd degree polynomial equation

Minimum L to achieve expected delay a times the
optimal (M 100)
14
What If Network Parameters Are Unknown?Online
Parameter Estimation
IDEA use meeting times statistics
Method
Estimator
Note Optimal L depends on M only
Applies to any mobility model with exponential
meeting times
15
Scalability of Spray and Wait
a 2
Number of Copies L (M 100)
a 5
a 10

• Spray and Wait M? ? ?
• In contrast in flooding, transmissions grow
  linearly with M
• Less than 10 need a copy to achieve 2x delay!

16
Simulations (contention, waypoint model)

• Simulated schemes
• Epidemic routing (epidemic)
• Randomized flooding (random-flood)
• Utility-based flooding (utility-flood)
• Spray and Wait (spray wait(L xx) )
• Seek and Focus (seek focus) Spyropoulos et al.
  04
• Simple slotted collision detection MAC protocol
  to handle contention

Adjusted individual protocol parameters per
scenario to achieve a good transmissions-delay
tradeoff
17
Scenario A Effect of Traffic Load
(500x500 grid, M 100 nodes, Tx Range 10)
increasing traffic
18
Scenario B Effect of Connectivity
(500x500 grid, M 200, medium traffic)
Spray and Wait is better with respect to both
metrics under all load and connectivity scenarios
considered !

• Spray and Wait clearly outperforms all schemes
  for all connectivity levels, in terms of both
  transmissions and delay
• Spray and Wait is considerably more scalable
• Performance of other schemes varies greatly with
  connectivity
• Spray and wait (i) fixed transmissions, (ii)
  decreasing delay

19
Limitations of Spray and WaitRestricted Mobility

• So far weve assumed that every node may go
  anywhere inside the network
• But, Spray and Wait may get in trouble if
• nodes mobility is restricted inside a local area
• nodes mobility is unrestricted but nodes move
  extremely slow
• Solution? Spray and Focus (work in progress)
• Spray L copies to L relays
• Route each copy using a single copy utility-based
  scheme (instead of direct transmission)

20
Work in Progress Performance of Spraying Schemes
in a Very Localized Scenario
Lessons learned Case 1 - highly mobile
nodes Spray and Wait is adequate (close to
optimal) Case 2 slow moving nodes/local
movement Spray and Focus is the winner (utility
contains a lot of information here)
21
Conclusions and Future Work

• Conclusion
• Spray and Wait yields lower delay than existing
  flooding and utility-based schemes, and
  significantly reduces transmissions
• delays close to the optimal can be achieved with
  few copies
• theory and simulations prove that it is scalable
• It is simple can be optimized with little
  knowledge about the network
• Future Work
• Performance of all protocols under more realistic
  mobility models that exhibit correlation in space
  and/or time
• Preliminary simulations show Spray and Focus
  performs well
• Good utility function at the focus phase is the
  key
• Extend theory for such scenarios (non-exponential
  meeting times)
• Extend theory to model contention

22
The End

• Thank You!

23
Target Applications(Delay Tolerant Networks)

• Sensor networks for habitat monitoring and
  wildlife tracking
• ZebraNet sensor nodes attached on zebras,
  collecting information about movement patterns,
  speed, herd size, etc.
• Boatnet
• Ad hoc networks for low cost Internet provision
  to remote areas/communities
• Africa, Saami, etc.
• Inter-planetary networks
• extend the idea of Internet to space
• Ad-hoc military networks