TCP-Tahoe: This
is a “Congestion avoidance” method in TCP protocol. The algorithms that is uses
is:
Slow-start: At the beginning It
start with making “congestions windows” value 1 “cwnd = 1” to avoid connection
start problem.
Congestion avoidance: it used
“Additive increase Multiplicative Decrease”. When a packet is lost it detects
it as a congestion and takes half of valid windows size as new threshold level
and saves. Congestion window set to ‘1’ and until reaching threshold level it
continues with Slow-start.
The most important point in Tahoe
is; it can detect the packet loss only when there is time out.
The problem is Tahoe is, to
detect if there is a congestion or not, it waits the timeout which during this
the pipeline is not used.
TCP-Reno: It is similar to Tahoe and as a difference it can do Fast
Retransmit. The Fast Retransmit is method to detect the packet lost from 3 same
ACK return (in this method for every packet it returns one ACK). If there is 3
same ACK return then it get as a packet lost and resend the packet by making
half the congestion window(cwnd = ½*cwnd) it continues. The problem in this
method is, although it is very useful in single packet loss its performance
decrease when there is more than one packet loss.
TCP-Reno
New: New version of the TCP-Reno.
When there is more than one packet loss, without decreasing congestion window
it resend the packet. It is more useful in mobile environments where packet
loss is bigger. The problem is this method is inside RTT (round trip time) it
can send only one packet.
TCP-Vegas: In this
protocol when packet send the system time is saves and when ACK returned it
also control the time and makes RTT calculations and saves it into timestamp.
It controls it in two conditions:
- · When there is same ACK return it control the time and check if there is a timeout or not, if there is it retransmits without waiting 3.ACK return.
- · When there is different ACK came it check if there is timeout, if there is it retransmits.
No comments:
Post a Comment