Lecture 7: IGP and Djikstra's Algorithm

Question 1

What is the count-to-infinity problem?

Answer 1

A router goes offline during
normal operation, and the news of this failure doesn’t propagate into the network fast enough

Re-advertisement of a no-longer valid route ends up pinging
back and forth across the network, constantly increasing in
link cost in an infinite loop

Question 2

How do link-state avoid the count-to-infinity issue and provide faster convergence over distance-vector protocols?

Answer 2

The build a complete map of the network and broadcast transmissions to the entire network

Question 3

What are the 2 steps in link state protocols?

Answer 3

1. Broadcast info about your neighbors to the entire network while receiving similar broadcasts from other routers (link-state information)
2. Build the routing table with the complete network map that the link state information provides

Question 4

How do link state protocols identify its neighbors?

Answer 4

It repeatedly scanning it’s network interfaces for new connections

Question 5

What does the link-state packet contain?

Answer 5

• The address of each neighbor
• The edge cost of the connection to each neighbor

Question 6

Where does the link-state packet go?

Answer 6

It is broadcast across all network interfaces of the router so that all routers in the AS receive it and update their link state info

Question 7

What algorithm is used in link-state protocols to find the shortest path?

Answer 7

Dijkstra’s algorithm

Question 8

How does Dijkstra’s algorithm work?

Answer 8

1. Set all node costs to infinity
2. Find the shortest edge, and travel along it. Update the cost of the vertex at the end of the edge, and mark it visited
3. Update the costs of the visited nodes’ neighbors
   based on their edge costs
4. Find the next lowest cost unvisited node, and continue

Question 9

What does the first step in the shortest path to a particular destination form for the route?

Answer 9

The next hop

Question 10

What is a difference between Bellman-Ford and Dijkstra’s algorithms?

Answer 10

• Bellman-Ford can handle negative edge weights
• Bellman-Ford has higher time complexity
• The ordering of steps in finding the shortest path and whether nodes are revisted

Question 11

What is the trade off of link-state protocols?

Answer 11

Have faster route calculation and convergence at the cost of memory/network capacity