Last lecture

Connection Establishment

Intro to the congestion control and loss recovery mechanism of TCP
-------------------------------------------------------------------

Today

Further discussion on TCP
-------------------------

To avoid generating an inappropriate amount of network traffic for the
current network conditions, during a connection TCP employs four
congestion control mechanisms:

 Slow start, congestion avoidance, fast retransmit and
    fast recovery. 

These algorithms are used to adjust the amount of unacknowledged data
    that can be injected into the network and to retransmit segments
    dropped by the network.


Two variables (units in terms of number of packets):

1) sender congestion window (snd_wnd) - always set to < = rcv_wnd.
= the amount of data the sender can inject into
    the network before receiving an ACK

The congestion window is increased or decreased during the transfer
    based on the inferred amount of congestion present in the
    network.


2) slow start threshold (ssthresh) 
  --- determines which algorithm is used to increase the value of
  snd_cwnd.

If snd_cwnd < ssthresh, increment snd_cwnd by 1 packet every time an ack is
received. (slow start phase)
For example, after the first ACK is received cwnd will be
    2 segments and the sender will be allowed to transmit 2 data
    packets.  This continues until cwnd meets or exceeds ssthresh (or,
    in some implementations when cwnd equals ssthresh), or loss is
    detected. In this phase, snd_cwnd is doubled for every round-trip
time (RTT). Use Fig 8.9 to illustrate this.

If snd_cwnd > ssthresh, increment snd_cwnd by 1/snd_cwnd every time an
    ack is received. (congestion avoidance phase) In this phase,
    snd_cwnd will increase by roughly 1 packet for each round-trip
    time (RTT). Use fig 8.7.  Congestion avoidance is used to slowly
    probe the network for additional capacity.

If a packet is dropped, set sshthresh to snd_cwnd/2 and set snd_cwnd
to 1.  
- TCP's default mechanism to detect dropped segments is a
timeout.  The retransmission timeout (RTO) is based on observations of
the average RTT and the standard deviation in that average.
In most commercial TCP implementations today, a coarse-grained (500
ms) clock is used - leading to long period of time the connection
idles while waiting for the timer to expire.

To remedy this, a new mechanism called fast retransmit is 
added - to trigger the retransmission of a dropped packet 
sooner than the regular timeout mechanism.