Routing, flow control and dynamic routing

Significance of routing

Routing decides the path that packets take from source to destination through the network of switching nodes.

• Delivers packets to the correct destination across many possible paths.
• Chooses the best path (least delay, cost or hops), so it improves throughput and reduces delay.
• Balances load over links and avoids congested parts.
• Finds alternative paths when links or nodes fail, so the network is reliable.
• Uses network resources efficiently, reducing cost.

Significance of flow control

Flow control regulates how fast a sender transmits so that the receiver or an intermediate node is not overrun.

• Prevents buffer overflow and the resulting packet loss and retransmissions.
• Matches devices of different speeds (e.g. fast server, slow terminal).
• Helps avoid congestion and deadlock in the network.
• Keeps throughput high and delay stable; ensures fair sharing.
• Methods: stop-and-wait, sliding window (HDLC, TCP), XON/XOFF, RTS/CTS, credit and rate (leaky bucket) control.

Dynamic routing

Dynamic (adaptive) routing changes routes automatically to follow the current topology and traffic. Routers exchange routing information and recompute their tables periodically or when a change is detected.

Types

• Centralized: a Routing Control Centre gathers data from all nodes, computes routes and sends tables.
• Isolated: each node uses only local information, e.g. hot potato routing (send on the shortest queue).
• Distributed:
  • Distance vector: each router sends its distance table to neighbours; updates by Bellman-Ford, $D_x(y) = \min_v \{c(x,v) + D_v(y)\}$. Simple, but slow convergence (count-to-infinity). Example: RIP.
  • Link state: each router floods its link costs to all; every router builds the whole map and runs Dijkstra. Fast convergence. Example: OSPF.

 A ---2--- B ---1--- D
  \               /
   5             3
    \--- C -----/
 Normal: A->D via B (cost 3)
 B-D fails: A->D via C (cost 8), learnt automatically

Advantages: adapts to failures and congestion, no manual work, good for large networks. Disadvantages: control traffic overhead, more processing, possible routing loops or oscillation during convergence.