Utility-based Routing

1
Utility-based Routing

• Jie Wu
• Dept. of Computer and Information Sciences
• Temple University

2
Roadmap

• Mao vs. Hardy
• Why Another Routing Scheme?
• Utility-Based Routing
• Implementations
• Extensions
• Some Final Thoughts

3
Mao vs. Hardy

• Z. Mao (Serve the People)
• Knowledge begins with practice.
• Theoretical knowledge acquired through practice
  must then return to practice.
• G. H. Hardy (A Mathematician's Apology)
• The real mathematics of the real mathematicians
  is almost wholly useless.
• It is not possible to justify the life of any
  genuine professional mathematician on the ground
  of the utility of his work.

4
Implications

• Politicians (when they become politically weak)
• Start new revolutions
• (and young people become followers)
• Mathematicians (when they become old)
• Start writing books
• (and young people prove theorems)
• Professors (when they become seniors)
• Give presentations
• (and students write papers)

5
Why Another Routing Scheme?

• Why routing again?
• Because it is interesting (a non-serious answer)
• A new routing algorithm composite utility
• Benefit (of packet delivery)
• Cost (of forwarding)
• Reliability (of links)
• Timeliness (of reaching destination)

6
A Postage Example

• Best route importance of the package
• Valuable package Fedex (more reliable, costs
  more)
• Regular package Regular mail (less reliable,
  costs less)

7
A Sample Network

• Traditional metrics cost/reliability
• The minimum cost path s ? 1 ? d
• Cost 2 3 5
• Reliability 0.8 0.9 0.72
• The most reliable path s ? 2 ? d
• Cost 4 3 7
• Reliability 0.9 0.9 0.81

8
Utility-Based Routing (LuWu06)

• Each packet is assigned a benefit value, v
• s transmits a packet with benefit v to d
• Transmission cost/reliability c/p
• Utility v c if success, 0 c otherwise
• Expected utility u p(v-c) (1-p)(0-c) pv -
  c
• The best route maximizes u

Success p
Failure 1-p
9
A General Expression

• General form of u for path
• R s ( 0), , i, i1, , d ( n)
• where, PR route stability and CR route
  cost

10
How to calculate u ?

• Direct calculation
• 0.8 0.920 2 30.810
• Backward calculation
• ui pi,i1 ui1 - ci,i1 (virtual s/d)
• 0.920 3 15 (at i)
• 0.815 2 10 (at s)

11
Benefit Dependent Best Paths
Ri Pi Ci
R1 0.72 4.4
R2 0.81 6.7
R3 0.5 5.3
R4 0.57 7.7
R1 s?1?d R2 s?2?d R3 s?1? 2?d R4 s?2? 1?d
Different benefit values may have different best
paths!
12
Implementations

• Centralized Source collects global link-state
• Applies a modified Dijkstras shortest path from
  d
• Each node i maintains the maximum ui (initiated
  to zero)
• i relaxes j uj pj,i ui- cj,i until reaching s
• Wireless and mobile reactive approach
• Route discovery (from s)
• Route reply (from d)

relax
j
i
s
d
13
Extensions

• HPCC All optimal routes
• Different benefit values
• IUCC Wireless networks
• Opportunistic routing
• Network coding
• TrustCom Incentive compatible routing
• Handling selfish nodes
• ICESS Real-time responses
• Low duty cycles in WSNs

14
All Optimal Routes (HPCC)

• Requirement
• Find all optimal routes for different benefits
• Challenges
• Enumerating all benefits is infeasible
• For a given range of benefits
• Checking all paths is too expensive
• Exponential to the number of nodes

15
Intersection Point
R1 s -gt 1 -gt d R2 s -gt 2 -gt d
UR1 0.72v 4.4 UR2 0.9v-7

• Complexity O(R2)
• (R number of paths)

16
Binary Partition
Iteratively partition the benefit range into
sub-ranges
Stoppage condition r tan ? lt ? (r sub-range,
? angle between R1 and R2)
17
Wireless Networks (IUCC)

• Opportunistic routing (OR) with adjustable
  transmission range
• Relay set more than one node can relay
• Priority ETX or cost to destination

18
OR Example

• Best expected utility
• us 10 for v 20
• Priority
• s lt 1 lt 2 lt d
• The best expected opportunistic utility
• opus 14.6 for v 20
• Optimal solution
• NP-hard

19
Network Coding

• Linearly independent code at s
• a b and 2a b
• Another code at n2
• (a b) (2a b) 3a 2b
• Optimal credit min transmission
• input vs. output rate
• (n1, n2) (1, 0.5)
• (n1, n2) (0, 1)
• Optimal credit max utility if c(n1) lt c(n2)
• (n1, n2) (1, 0.5)
• Khreishah, Khalil, Wu (MobiHoc12)

n1
a b
a b
0.5
1
s
d
1
1
a b 2a b
n2
3a 2b
20
Incentive Compatible Routing (TrustCom)

• Nodes are selfish and give false private
  information
• Without reward, they will not help relay packets
• Maximize utility payment cost
• Mechanism design
• Tie self interest to societal interest
• VCG scheme enforcing the reporting of correct
  link costs
• Nodes on the optimal path
• utility remains the same when lying
• Nodes not on the optimal path
• utility reduces when lying

21
Second Price Path Auction

• Why doesnt the first price work?
• System objective inconsistent with individual
  nodes
• objectives
• The solution second price
• Losers utility is 0
• Winner is payment
• lowest cost without i - lowest cost cost of
  node i

22
A VCG Example

• Case 1 nodes on an optimal path lie
• If (s, 1) is changed to 3
• S still gets 7 6 3 4
• (same as 7 5 2 4)
• Case 2 nodes on a non-optimal path lie
• If (2, d) is changed to 1
• 2 gets 5 5 1 1 lt 3
• (utility is negative)

1
2
3
s
d
2
4
3
2
23
Real-Time Responses (ICESS)

• Energy saving on/off node
• t(s) 4, node s is up every 4 units
• Least common multiple (LCM)
• t(s) 4, t(d) 3, then LCM(t(s), t(d)) 12,
  link delay for (s, d)
• Extending utility function delay-sensitive

d
24
Low Duty Cycles in WSNs

• Utility for a delivery path R s (0), 1, 2, ,
  n-1, d (n)
• Direct computation
• Iterative computation
• forward
• backward

forward
d
s
backward
25
Probabilistic Contacts in DTNs

• Benefit is time-sensitive
• Balance delay and cost
• Probabilistic contacts
• Opportunistic forwarding
• Forwarding set is time-varying

26
Some Final Thoughts

• Is research on routing over?
• Probably yes MANETs and sensor nets
• No Other networks (e.g. DTNs and social
  networks)
• Mobility in Wireless Networks Friend or Foe ?
• Mobility as a Foe tolerating and masking
• Mobility as a Friend mobility-assisted routing

27
Some Challenges

• Future world being more wireless and mobile
• Complexity and diversity
• New challenges for architecture and protocol
  design
• From top more demand from the end user
• (e.g., mobility support)
• From bottom emerging technologies
• (e.g., new abstraction for wireless links)

28
Graph Model for Dynamic Networks

• E.g. Mobility affects network model/protocol
• Time-space view vs. space view
• View consistency in asynchronous systems
• Wu Dai (IEEE Network05) function of multiple
  views
• Evolving graph model connectivity routing
• Liu Wu (MobiHoc07, 08, 09)
• Wu (Graph and Computing10)

29
Collaborators

• Former students
• Dr. Mingming Lu
• Prof. Feng Li
• Visiting scholar
• Prof. Mingjun Xiao

30
Questions