Ch. 12 Routing in Switched Networks

1
Ch. 12 Routing in Switched Networks
2
12.1 Routing in Packet Switched Networks

• Routing Algorithm Requirements
• Correctness
• Simplicity
• Robustness--the ability of the network to deliver
  packets via some route in the face of localized
  failures and overloads.
• Stability--does not over react to network
  changes.
• Fairness--as related to all other users.
• Optimality--as related to some criterion.
• Efficiency--as related to processing overhead.

3
12.1 Elements of Routing Techniques

• Performance Criteria
• Number of hops, cost, delay, throughput.
• See Table 12.1
• Decision Time
• Virtual Circuit--at connection establishment.
• Datagram--before packet is placed in outgoing
  buffer.
• Decision Place
• Each node, central node, originating node.

4
12.1 Elements of Routing Techniques (cont.)

• Network Information Source
• None, local, adjacent nodes, nodes along the
  route, or all nodes.
• Network Information Update Timing
• Continuous, periodic, major load change, topology
  change.

5
12.1 Routing Strategies

• Fixed Routing
• A route is selected for each source-destination
  pair of nodes.
• A central routing directory can then be
  distributed to the various nodes.
• Routes are not changed unless topology changes.
• Simple (advantage) but inflexible (disadvantage.)

6
12.1 Routing Strategies

• Fixed Routing Example (Fig. 12.2)
• Refer back to the network in Fig. 12.1.
• Central directory lists all the routing
  information.
• Each column of the central directory becomes the
  Next Node columns in the nodal directories.

7
12.1 Routing Strategies (p.2)

• Flooding (Fig. 12.3)
• A packet is sent out on every outgoing link
  except the link that it arrived on.
• Duplicates must be discarded.
• Hop counter could be used.
• Very robust (advantage.)
• High traffic loads are generated (disadvantage.)

8
12.2 Routing Strategies (p.3)

• Random Routing
• An outgoing link is selected at random (based on
  a probability distribution.)
• Requires no use of network information
  (advantage.)
• Actual route will not be least-cost or
  minimum-hop route (disadvantage.)

9
12.2 Routing Strategies(p.4)

• Adaptive Routing
• These algorithms react to changing conditions of
  the network, for example failures and congestion.
• Advantages--can improve performance and aid in
  congestion control.
• Disadvantages--complex, requires extra "overhead"
  traffic to collect information, and may react too
  quickly (unstable.)

10
12.2 Routing Strategies (p.5)

• Adaptive Routing(cont.)
• Schemes can be characterized by
• Source of Network Information
• Local--Fig. 12.4 Isolated Adaptive Routing
• Minimize Queue Length Bias
• Adjacent Nodes
• All Nodes
• Distributed or Centralized Control

11
12.2 Examples Routing in Arpanet

• First Generation Distant Vector Routing
• Distributed adaptive algorithm (1969)
• Performance criteria--estimated delay (from queue
  length).
• Version of the Bellman-Ford Algorithm.
• Problems did not consider line speed, queue
  length is not an accurate measure of delay, and
  the algorithm responded slowly to congestion and
  delay increases.
• See Fig. 12.5, 12.6 and discussionpage363.

12
12.2 Internet Routing Examples (p.2)

• Second Generation (Link-State Routing)
• Distributed adaptive algorithm (1979).
• Performance criteria--delay (direct
  measurements).
• Version of Dijkstra's Algorithm.
• Problem did not work well for heavy loads.

13
10.2 Routing Strategy Examples (p.3)

• Third Generation ARPANET (1987)
• The average delay is measured and transformed
  into estimates of utilization.
• The link "costs" were calculated as a function of
  utilization--this helped to prevent oscillations.
• Fig. 12.7--traffic could oscillate from link A to
  link B and back.

14
12.3 Least-Cost Algorithms

• The Problem
• Given a network of nodes connected by
  bi-directional links, where each link has a cost
  associated with it in each direction, define the
  cost of a path between two nodes as the sum of
  the costs of the links traversed. For each pair
  of nodes find the path with least cost.
• Solutions
• Dijkstra's Algorithm (1959)
• Bellman-Ford Algorithm (1962)

15
Dijkstra's Algorithm

• Define
• Nset of nodes in the network.
• ssource node.
• Tset of nodes so far incorporated by the
  algorithm.
• w(i,j)link cost from node i to node j w(i,i)0
  and w(i,j)? if the nodes are not directly
  connected.
• L(n) cost of the least-cost path from node s to
  node n that is currently known to the algorithm.

16
Dijkstra's Algorithm (p.2)

• Three Steps (repeated until MN)
• Step 1 Initialize Variables
• T s.
• L(n)w(s,n) for n ? s.
• Step 2 Get Next Node
• Find the neighboring node (x) which has the
  least-cost path from node s and incorporate that
  node into T.
• Step 3 Update the least cost-paths.
• L(n) min L(n), L(x) w(x,n) for all n ? T.
• See Table 12.2a and Fig. 12.9.

17
Bellman-Ford Algorithm

• Define
• s the source node.
• w(i,j)link cost from node i to node j.
• hmaximum number of links in a path at the
  current stage of the algorithm.
• Lh(n) cost of the least-cost path from node s
  to node n under the constraint of no more than h
  links.

18
Bellman-Ford Algorithm (p.2)

• Step 1 Initialize
• L0(n)?, for all n not equal to s.
• Lh(s) 0, for all h.
• Step 2 For each successive h,
• L h1(n) Minj Lh(j) w(j,n).

19
Comparison of the Algorithms

• Dijkstras
• Complete topology information is needed.
• Bellman-Ford
• Knowledge of link costs to each neighbor, and the
  current distance-vector of each neighbor is
  required.