Routing Algorithm Objectives

1
Routing Algorithm Objectives
1
3
6
A
4
B
Host
2
Switch or router
5

• Rapid and Accurate packet Delivery
• Adaptability to varying traffic loads
• Congestion avoidance
• Adaptability to link failure or reconfiguration
• Low overhead

Figure 7.23
2
Routing Alternatives

• Static Versus Dynamic
• Static routes are manually configured. (not
  practical in large or changing networks)
• Dynamic routes are automatically determined
  (complex and potentially unstable)
• Centrally controlled or distributed
• Centrally controlled routes for whole network
  are determined in one place and distributed (easy
  to optimize globally)
• Distributed each router determines routes using
  information exchanged with neighbors (more
  scaleable, but potentially unstable)
• Virtual Circuit and Datagram?
• Datagram each packet is routed using routing
  information
• Virtual Circuit Route is determined once for
  connection using same routing information.

3
Virtual Circuit Example
Virtual Circuit
2
7
1
8
B
1
3
3
A
1
6
5
5
4
2
VCI
4
3
5
2
5
C
6
D
2

• Each packet carries a VCI identifying its virtual
  circuit
• Each router maps VCI to new link and VCI on that
  link

Figure 7.24
4
Virtual Circuit Routing Tables
Node 3
Incoming Outgoing
Node 6
Node 1
node VC node VC
1 2 6 7
Incoming Outgoing
Incoming Outgoing
1 3 4 4
node VC node VC
node VC node VC
4 2 6 1
3 7 B 8
A 1 3 2
6 7 1 2
3 1 B 5
A 5 3 3
6 1 4 2
B 5 3 1
3 2 A 1
4 4 1 3
B 8 3 7
3 3 A 5
Node 4
Incoming Outgoing
node VC node VC
2 3 3 2
Node 2
Node 5
3 4 5 5
Incoming Outgoing
Incoming Outgoing
3 2 2 3
node VC node VC
node VC node VC
5 5 3 4
C 6 4 3
4 5 D 2
4 3 C 6
D 2 4 5
Figure 7.25
5
Datagram Routing for same routes
Node 3
Node 6
Destination Next node
Node 1
Destination Next node
1 1
Destination Next node
2 4
1 3
2 2
4 4
2 5
3 3
5 6
3 3
4 4
6 6
4 3
5 2
5 5
6 3
Node 4
Destination Next node
1 1
2 2
Node 2
Node 5
3 3
Destination Next node
Destination Next node
5 5
1 1
1 4
6 3
3 1
2 2
4 4
3 4
5 5
4 4
6 5
6 6

• Each Packet carries full destination address
• Routing table must identify every destination
  (large)

Figure 7.26
6
Hierarchical Routing
Destinations reached from this link
Hierarchical Network (destinations grouped by
address)
4 entries
Tables specifiy only unique prefix of destination
Non-hierarchical Network
16 entries
Tables specifiy all destinations
7
Link State vs Vector Routing

• Distance vector routing
• Routers maintain tables with cost and next link
  to reach each destination
• Routers exchange these distance vectors to update
  best route
• Link state routing
• Routers maintain information about their links
• Routers broadcast link information to all others
  to allow optimal routes to be constructed

8
Routing based on shortest path

• Each link has an associated cost
• Packet delay (length and/or queuing)
• Link congestion (average buffer length)
• Inverse of link speed
• Objective is to pick the route with the lowest
  total cost
• Difficult to compute the true optimum
• Costs change rapidly (congestion, broken links)
• Routing algorithms compute an approximation.

Figure 7.28
9
Bellman Ford Algorithm

• A shortest path from x to y consists of the sum
  of the cost of a link from x to some node (a)
  plus a shortest path from a to y.
• The shortest path from X to Y can be found by
  examining all links from X and minimizing the
  sum of the cost of that link and the shortest
  path from the node it reaches to Y.
• The shortest path can be found by iteration.

10
The algorithm

• For all nodes i set Di 0 for d and ? otherwise
• For all nodes i,j set Cij to
• The cost of the direct link between i and j
• ? otherwise
• For each i not the destination do
• Di min(CijDj) for all j (not equal to i)
• Repeat step 3 until no more changes.

11
Example cost to node 6
12
Least Cost routing tree
2
1
1
3
6
2
4
1
2
2
5

• Nodes exchange distance vectors
• Routing computation is distributed among nodes
• Convergence is generally rapid

Figure 7.29
13
Recovery from a broken link
2
1
3
X
6
5
2
3
4
2
1
3
2
5
4

• Nodes periodically recheck routes and discover
  outage
• During recomputation loops are possible
• All packets eventually route correctly

Figure 7.30
14
Recovery can be difficult
(a)
3
1
2
4
1
1
1
(b)
3
1
2
4
X
1
1

• When link between 3 and 4 breaks, nodes 1,2,3
  will only slowly increase cost of routes (route
  towards the node with the oldest estimate of
  cost)
• Problem known as counting to Infinity
• Modifications to algorithm to address it.

Figure 7.31
15
Split Horizon/Poisoned reverse
3
1
2
4
X
1
1

• Each node sets its cost to the destination as
  infinite in the data sent to the node on its
  shortest path.
• Prevents some forms of slow convergence

16
Dijkstras Algorithm

• Works outwards from the source in identifying the
  next closest node
• Set N s, Ds0, DjCsj
• Find i not in N with minimum Di
• Add i to N (stop if no more nodes)
• Update Djmin(Dj,DiCij)
• Go to step 2.

17
Example cost from node 1
18
Spanning Tree computed
2
1
1
3
6
2
3
4
2
2
5
Figure 7.32
19
Packet Flooding
(a)

• Packets sent on all links

Figure 7.33 - Part 1 of 3
20
Fooding stage 2
(b)

• Nodes rebroadcast received packets

Figure 7.33 - Part 2 of 3
21
Flooding stage 3
(c)

• Network becomes congested unless old packets
  removed
• Remove by time to live
• Remove packets that have already been seen

Figure 7.33 - Part 3 of 3
22
Deflection Routing
0,0
0,1
1,0
1,1

• Geometry of the network determines normal route
  (Self routing)
• Packets encountering a busy route take another
• Does not require buffering
• Many network geometries possible.

23
Source Routing
3,6,B
6,B
1
3
B
6
A
4
B
Source host
2
Destination host
5

• Source node specifies a complete route
• Reduces work by routers
• Focuses routing effort on the hosts
• Routes can be saved by adding node numbers to
  packets as they flow through the network

Figure 7.36
24
ATM Networking

• Motivation
• Packet switching is flexible but has high
  variable delay (not good for real-time streams)
• TDM Switching is predictable but wont handle
  bursty traffic and arbitrary rates.
• Compromise Solution ATM
• Small fixed size packets or cells (53 bytes) (low
  delay)
• Virtual circuit routing (small headers)
  (efficient, predictable)
• Packets fill fixed sized periodic frames.
• Synchronous traffic regular rate of cell
  arrival
• Asynchronous traffic bursty rate of cell arrival

25
ATM and TDM Networking
Voice
Data packets
MUX
Wasted bandwidth
Images
TDM
4 3 2 1 4 3 2 1
4 3 2 1
ATM

4 3 1 3 2 2 1

• TDM each source has fixed slot/bandwidth
• ATM each source produces self describing cells
  only when it has something to send

Figure 7.37
26
ATM Switching
1
1
voice
67
Switch

video
2
67
N
25
75
video
32
voice
5
25
1
32
67
data
3
39
32
39
3
video
32
61
data
6

61
2
67

N
video
75
N

• Like Virtual Circuit switching each incoming
  cell gets new Virtual Channel ID and outgoing link

Figure 7.38
27
Virtual Paths in ATM
a
VP3
VP5
a
b
ATM Sw 1
ATM Sw 2
ATM DCC
b
c
ATM Sw 3
c
d
e
VP2
VP1
d
ATM Sw 4
e
Sw switch

• Two levels of adressing Virtual PATH, and
  Virtual Channel within a path
• Allows paths to be managed as a shared resource
  and virtual channels managed locally

Figure 7.39
28
View of an ATM Link
Virtual Paths
Physical Link
Virtual Channels
Figure 7.40
29
For Next Week

• Exam covering Chapters 5,6, and 7.1-7.6
• 1 hour, like the last, 2 pages of notes
• I will not expect you to remember all the delay
  and loss formulas
• Following the exam we will cover the remainder of
  Chapter 7.
• No homework this week but next week there will be
  homework for chapter 7.