1. Routing

2. Distance Vector (DV) Routing
Sets up minimum distance routes to all nodes in a network
Routing tables created at each node with following fields:
The core algorithm is based on ‘Bellman Ford shortest path algorithm’
Destination
Next hop
Cost X V 14 Q B 11 ..

3. DV example: Initial State
Destination
Next hop
Cost A 1 C 3 E 9
Destination
Next hop
Cost B 1 C 5 A B C D E 1 2 9 3 5 4
Destination
Next hop
Cost C 4 E 2

4. DV example: Final State
Destination
Next hop
Cost A 1 C 3 E 9 D ?
Destination
Next hop
Cost B 1 C 5 D ? E 1 9 B E A 3 2 5 D C 4

5. DV steps
Each node advertises its routing table to neighbors(only destination & cost)
Each neighbor updates its table based on following equation
𝐶𝑜𝑠 𝑡 𝑦 𝐷𝑒𝑠𝑡 =min⁡(𝐶𝑜𝑠 𝑡 𝑦 𝐷𝑒𝑠𝑡 ,𝐶𝑜𝑠𝑡 𝑋,𝑌 +𝐶𝑜𝑠 𝑡 𝑥 (𝐷𝑒𝑠𝑡))
In the above equation,
X and Y are neighbors
Cos t y Dest denotes cost of reaching node ′𝐷𝑒𝑠𝑡′ from Y
Cost X,Y denotes weight of edge XY

6. DV example for node C Destination Next hop Cost A 1 C 3 E 9B 1 C 5 A B C D E 1 2 9 3 5 4
Destination
Next hop
Cost C 1 E

7. DV example for node C Destination Next hop Cost A 1 C 3 E 9B 1 C 5 A B C D E 1 2 9 3 5 4
Destination
Cost A 1 C 3 E 9
Destination
Next hop
Cost C 4 E 2

8. A B C D E 1 2 9 3 5 4
C’s routing table
Advertisement from B to C
Destination
Next hop
Cost A 5 B 3 D 4
Destination
Cost A 1 C 3 E 9

9. A B C D E 1 2 9 3 5 4
C’s routing table
Advertisement from B to C
Destination
Next hop
Cost A 5 B 3 D 4 E - ∞
Destination
Cost A 1 C 3 E 9
𝐶𝑜𝑠 𝑡 𝐶 𝐸 =min⁡(𝐶𝑜𝑠 𝑡 𝐶 𝐸 ,𝐶𝑜𝑠𝑡 𝐵,𝐶 +𝐶𝑜𝑠 𝑡 𝐵 (𝐸))
𝐶𝑜𝑠 𝑡 𝐶 𝐸 = min ∞, 3+9 =12
New route identified

10. C’s routing table
Destination
Next hop
Cost A B 4 3 D E 12
New route added

11. A B C D E 1 2 9 3 5 4
C’s routing table
Advertisement from B to C
Destination
Next hop
Cost A 5 B 3 D 4 E 12
Destination
Cost A 1 C 3 E 9
𝐶𝑜𝑠 𝑡 𝐶 𝐴 =min⁡(𝐶𝑜𝑠 𝑡 𝐶 𝐴 ,𝐶𝑜𝑠𝑡 𝐵,𝐶 +𝐶𝑜𝑠 𝑡 𝐵 (𝐴))
𝐶𝑜𝑠 𝑡 𝐶 𝐴 = min 5, 3+1 =4
Route updated

12. C’s routing table
Destination
Next hop
Cost A B 4 3 D E 12
Route updated

13. DV example for node C Destination Next hop Cost A 1 C 3 E 9B 1 C 5 A B C D E 1 2 9 3 5 4
Destination
Next hop
Cost C 4 E 2
Destination
Cost B 1 C 5

14. A B C D E 1 2 9 3 5 4
C’s routing table
Advertisement from A to C
Destination
Next hop
Cost A 5 B 3 D 4 E 12
Destination
Cost B 1 C 5
𝐶𝑜𝑠 𝑡 𝐶 𝐵 =min⁡(𝐶𝑜𝑠 𝑡 𝐶 𝐵 ,𝐶𝑜𝑠𝑡 𝐴,𝐶 +𝐶𝑜𝑠 𝑡 𝐴 (𝐵))
𝐶𝑜𝑠 𝑡 𝐶 𝐵 = min 3, 5+1 =3
Not updated
C’s routing table unchanged

15. DV example for node C Destination Next hop Cost A 1 C 3 E 9B 1 C 5 A B C D E 1 2 9 3 5 4
Destination
Next hop
Cost C 4 E 2
Destination
Cost C 4 E 2

16. A B C D E 1 2 9 3 5 4
C’s routing table
Advertisement from D to C
Destination
Next hop
Cost A 5 B 3 D 4 E 12
Destination
Cost C 4 E 2
𝐶𝑜𝑠 𝑡 𝑦 𝐷𝑒𝑠𝑡 =min⁡(𝐶𝑜𝑠 𝑡 𝑦 𝐷𝑒𝑠𝑡 ,𝐶𝑜𝑠𝑡 𝑋,𝑌 +𝐶𝑜𝑠 𝑡 𝑥 (𝐷𝑒𝑠𝑡))
𝐶𝑜𝑠 𝑡 𝐶 𝐸 =min⁡(𝐶𝑜𝑠 𝑡 𝐶 𝐸 ,𝐶𝑜𝑠𝑡 𝐷,𝐶 +𝐶𝑜𝑠 𝑡 𝐷 (𝐸))
𝐶𝑜𝑠 𝑡 𝐶 𝐸 = min 12 ,4 +2 =6
Route Updated

17. C’s routing table
Destination
Next hop
Cost A 5 B 3 D 4 E 6
Updated Next-hop and Cost

18. DV properties
After one round of message exchange with neighbors, new routes to 2 hop nodes discovered
After two rounds, routes to 3 hop neighbors discovered
Converges after a few rounds if topology does not change

19. DV properties
Completely distributed algorithm – No node has global picture, yet learns shortest paths to all nodes in the network

20. Changes in Topology Nodes send updates. Two types
Triggered update: Topology change, link failure triggers an update to be sent to neighbors
Periodic update: sent even when no change happens in routing table

21. Periodic updates better than triggered updates
Destination
Next hop
Cost B 1 C A
Destination
Next hop
Cost A 1 B B C
Destination
Next hop
Cost A 1 C

22. Periodic updates better than triggered updates
Destination
Next hop
Cost B 1 C A
Destination
Next hop
Cost A 1 B B C
Destination
Next hop
Cost A 1 C

23. Periodic updates better than triggered updates
Destination
Next hop
Cost B 1 A
Destination
Next hop
Cost A 1 B C
Destination
Next hop
Cost A 1 C

24. Periodic updates better than triggered updates
Destination
Next hop
Cost B 1 A
triggered update (A to B)
Destination
Cost B 1
Destination
Next hop
Cost A 1 B C
Destination
Next hop
Cost A 1 C
B ignores

25. B send this update eve though it does not see any change in its table
If B does not send a periodic update, such as below, nodes A and C cannot learn about new paths to each other.
B send this update eve though it does not see any change in its table
Destination
Cost A 1 C
Destination
Next hop
Cost B 1 A
triggered update (A to B)
Destination
Next hop
Cost A 1 B C
Destination
Next hop
Cost A 1 C
Destination
Cost A 1 C
B ignores

26. Destination
Next hop
Cost B 1 C 2
New paths learnt A
periodic update (B to A)
Destination
Next hop
Cost A 1 B 2 B C
Destination
Next hop
Cost A 1 C
periodic update (B to C)
B ignores

27. DV  Count to Infinity problemA B 2 1 E .. A B C E D N C A 1 E ..
D - Destination
N - Next Hop
C - Cost

28. DV  Count to Infinity problemA B 2 1 E ..
C does not remove A, it doesn’t know A is down
C can corrupt other nodes with stale updates about A A B C E D N C 1 E ..
B removes A

29. DV  Count to Infinity problemA B 2 1 E .. A B C E D N C 1 E .. D C A 2 B 1 E ..
C sends an advertisement to B

30. DV  Count to Infinity problemA B 2 1 E .. A B C E D N C 1 E .. A 3 D C A 2 B 1 E ..
False route to A learnt with C as next hop

31. DV  Count to Infinity problemA B 2 1 E .. D C 1 E .. A 3
B sends an advertisement to C A B C E D N C 1 E .. A 3

32. DV  Count to Infinity problem
C updates route to A with a higher cost D N C A B 4 1 E .. D C 1 E .. A 3 A B C E D N C 1 E .. A 3
This process continues until cost to non-existent destination reaches infinity
Messages sent during this time to A will be wasteful

33. Avoiding Count to Infinity Problem
Use sequence numbers, update routes only if the message has a new sequence number

34. DSDV (Destination sequenced DV)
Adds sequence numbers to routing table entries
Initial sequence number created at the destination

35. DSDV (Destination sequenced DV)B 2 1 6 E .. A B C D E D N C S A 1 2 4 E ..
D - Destination
N - Next Hop
C – Cost
S – Sequence number

36. DSDV (Destination sequenced DV)B 2 1 6 E .. A B C D E D N C S A ∞ 3 1 4 E ..
B updates A’s cost to infinity and increases sequence number

37. DSDV (Destination sequenced DV)B 2 1 6 E .. A B C D E D N C S A ∞ 3 1 4 E .. D C S A 2 B 1 6 E ..
C sends an advertisement to B
C’s update on A ignored by B because B’s data has a higher sequence number

38. DSDV (Destination sequenced DV)B 2 1 6 E .. D C S A ∞ 3 1 4 E ..
B sends an advertisement to C A B C D E D N C S A - ∞ 3 1 4 E ..

39. DSDV (Destination sequenced DV)
B’s update on route to A overrides C’s table since it has a higher sequence number D N C S A - ∞ 3 B 1 6 E .. D C S A ∞ 3 1 4 E ..
B sends an advertisement to C A B C D E D N C S A - ∞ 3 1 4 E ..
The entire network learns about latest node failures/updates

40. Limitations of DV/DSDV
Too much message flooding might be a bandwidth overhead
Convergence can be slow for mobile networks

41. Dynamic Source Routing (DSR)
On demand routing  No periodic updates
When a source wants to route packets, routes generated
Consists of two parts
Route discovery
Route maintenance

42. I G C S H B F A D J E

43. Source sends a route request
RREQ (Resource request packet is generated)
includes source, destination, intermediate nodes (initially empty)
Seq. num
Source
Intermediate Nodes
Destination

44. Seq.no, source, int nodes, destG C 2 S D S H B F A D J E

45. 2 S D I G C S H B F A 2 S D D 2 S D J E

46. B re-broadcasts after adding itself
B received 2 S D
B re-broadcasts after adding itself I 2 S B D G C S H 2 S D B F A D J E
Intermediate nodes add themselves to the intermediate nodes, and rebroadcast the RREQ

47. If sequence number was seen before, message is dropped2 S B D S H 2 S B D B F A D 2 S B D J E
If sequence number was seen before, message is dropped

48. I G C 2 S B D S H 2 S B D B F A D 2 S B D J E

49. I G C 2 S B D S H 2 S B D B F 2 S
B,F D A D 2 S B D J E

50. I G C S H B F 2 S
B,F D A D J E
When the destination receives the RREQ, it generates a Route Reply (RREP)
The packet is routed back by following the list of intermediate nodes in the RREQ

51. I G S->B->F->D C S H B F RREP A D J E
After D previous one is F D J E

52. I G C S H
S->B->F->D B F A D J E

53. I G C S H
S->B->F->D B F A D J E

54. I G C
S->B->F->D S H B F A D J E
Now source knows the route
It appends route with the data

55. I G C
Data
S->B->F->D S H B F A D J E

56. I G C S
Data
S->B->F->D H B F A D J E

57. I G C S
Data
S->B->F->D H B F A D J E

58. I G C S H B F A D
Data
S->B->F->D J E

59. DSR optimizations
Intermediate nodes can generate RREP if they have the route in their cache I G C S H 2 S
B,F D B F A
F->D D J E

60. I G C S H
S->B->F->D B F A D J E

61. DSR optimizations RREP packets are cached
New route generation is not needed if cache has an unexpired entry
Intermediate nodes cache overhead routes for their own use

62. Route maintenance
RERR (route error) packets broadcasted to indicate broken links

63. I G C S H
S->B>F->D
S->C>G-I B F A
F->D D J E

64. I G C S H
S->B>F->D
S->C>G-I
x F->D B F A
F->D D J E

65. I G C
x F->D S H
S->B>F->D
S->C>G-I B F A
F->D D J E

66. I G C
x F->D S H
S->C>G-I B F A
F->D D J E

67. DSR limitations High overhead for large networks
Need to carry route information in each packet

68. AODV (Adhoc On-demand distance vector)
Best of both worlds of DSR and DSDV
Like DSDV it is next hop based routing, full route path not included in packet
Like DSR, does not need periodic route maintenance messages

69. Source sends Route Request (RREQ)
RREQ packet format
Seq. num
Source
Destination

70. Seq.no, source, dest I G C 2 S D S H B F A D J E

71. 2 S D I G C S H 2 S D B F A D 2 S D J E
Backward pointers created towards nodes sending/forwarding RREQ

72. I G C S H 2 S D B F A D J E

73. I G C S H B F 2 S D A D J E

74. 2 S D I G C S H B F 2 S D A D J E 2 S D
Hop Count
Generates RREP (route reply) with hop count 0
Sent back along backward pointer

75. I G C S H 2 S D 1 B F A D J E
Intermediate nodes receive and update hop-count by 1 (duplicate messages ignored as in DSR)
Forward pointers (green arrows) generated pointing to next hop for destination

76. I G C S H 2 S D B F A D J E

77. I G C 2 S D 3 S H B F A D J E
Just follow next hop links (green lines – forward pointers) during actual routing

78. AODV properties
RERR (like DSR) messages keep the network updated in case of failures
Sequence numbers prevent count to infinity problem
Widely used in WiFi-adhoc networks, ZigBee etc.

79. How to select link weights
Hop-count does not characterize a wireless link accurately
Wireless links are lossy. How can we take this into account while assigning link weights A B C D E ?

80. Delivery ratio (DR)
Characterizes fraction of packets delivered through a link
DR = 0.5 A B
Half of packets fail
If DR = 0.5, on average, we need to transmit two packets to get one successful reception

81. Delivery ratio (DR)
Characterizes fraction of packets delivered through a link
DR = 0.33 A B
66% of packets fail
If DR = 0.33, on average, we need to transmit three packets to get one successful reception

82. Delivery ratio (DR)
Characterizes fraction of packets delivered through a link
DR = p A B
100*(1-p)% of packets fail
If DR = p, on average, we need to transmit 1 𝑝 packets to get one successful reception

83. Define a new metric called ETX
ETX = expected number of transmissions on a link for successful reception B A B D C
ETX=1
ETX = 2
ETX ≈2
DR=1
DR =0.5 A D
DR = 0.51 C
DR = 0.51

84. Minimum ETX routing
ETX = expected number of transmissions on a link for successful reception B A B D C
ETX=1
ETX = 2
ETX ≈2
DR=1
DR =0.5 A D
DR = 0.51 C
DR = 0.51
Cost A-> B->D = ∑(ETX of all links along the path) = = 3
Cost A-> C->D = ∑(ETX of all links along the path) = = 4
Minimum ETX route is better. Hence ABD better than ACD

85. Minimum ETX routing
Any classical routing algorithm such as DSDV, DSR, AODV can use ETX as the edge weights for routing.