Flow control in data communication network

Flow control in a data communication network is the technique that controls the rate at which a sender sends data so that the receiver (or an intermediate node) is not flooded with more data than its buffers and processor can handle. Without it, frames are lost, must be retransmitted and throughput falls.

Why needed

• Sender and receiver differ in speed and buffer size.
• Receiver must also process data (check errors, pass to higher layers).
• Network nodes have limited buffers; overflow causes congestion.

Main methods

1. Stop-and-wait: send one frame, wait for ACK, then send the next. Simple but inefficient when propagation delay is large ($U = 1/(1+2a)$, $a = t_{prop}/t_{frame}$).
2. Sliding window: sender may have up to W unacknowledged frames; ACKs slide the window. Receiver can send RNR to stop. Efficient; used in HDLC, X.25, TCP.

  [0 1 2][3 4 5 6 7][8 9 ...]
  ACKed   window=5   not yet allowed

3. Start–stop: XON/XOFF characters or RTS/CTS hardware lines.
4. Credit / rate based: receiver grants credits (TCP window); or source is limited to a rate (leaky/token bucket).

Levels: hop-by-hop (between adjacent nodes, data link layer), network access (host to network) and end-to-end (between hosts, transport layer).