This paper presents about the congestion control algorithms which were derived
when in October of ’86 the network speed dropped and was called ‘series of
congestion collapses’. It basically states about achieving network stability by
following a ‘packet conservation principle’. The sudden factor of thousand drop
bandwidth fascinated them that they put seven new algorithms into the 4BSD TCP.
This paper is a brief description of slow start, round-trip-time variance,
dynamic window size on congestion. And the algorithms should obey ‘packet
conservation principle’. If the packet conservation fails it can lead to three
problems which are:
1.The connection doesn’t get to equilibrium.
2.A sender sends a new packet before an old packet has been received.
3.The equilibrium can’t be reached because of resource limits in the
The first algorithm states about the
connection to be in equilibrium. For each acknowledged new data, increase the
congestion window by one packet. And while sending the minimum of receiver’s
window and congestion window is sent. This algorithm makes sure that a
connection on network will source data to flow twice the rate .
The second problem which states that a sender sends a new packet before
an old packet has exited can be achieved by estimating a good round trip time that
helps to survive heavy load. The author has also discussed about retransmitting
of packets more than once but for this there is only one thing i.e. exponential
The author has stated that a packet loss occurs due to congested network
and time out. For this there is a congestion strategy that in a network endpoints should have the policy that a
congestion is occurring and it should increase the utilization if it didn’t receive
signal from network and should decrease the utilization if it received the signal.
This helps in utilization of bandwidth in the network and also helps in
And to avoid congestion the author has stated that congestion window
should be half the current window size i.e. multiplicative decrease on any timeout and on
any acknowledgement of new data congestion window should be increased by 1/cwnd
i.e. additive increase.
In the paper, the author has mentioned about the future work that is
gateways to avoid the congestion control. This approach is applicable to
identify the congestion problems and it will also work for heavy loads on
The paper overall addresses the problem about congestion in the network
and a way to control them. The paper doesn’t discuss about the wireless networks
congestion control as it is focused on wired network. Overall, the paper states
all the problem and the possible solutions for congestion avoidance and control
and states about the future works for congestion.