Scribd
Upload
27 views3 pages

Lecture07 Report Norranat

1) The document discusses the TCP Reno congestion control mechanism and compares it to TCP Tahoe through simulations run in NS2. 2) The simulations varied bandwidth between 1-14 Mbps and delay between 1-24 ms to observe their effects on congestion window size. 3) With TCP Reno, the congestion window size is halved rather than reset to 1 upon duplicate ACKs as in TCP Tahoe, allowing the network to recover more quickly from minor congestion.

Uploaded by

norranat songsriboonsit
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
27 views3 pages

Lecture07 Report Norranat

1) The document discusses the TCP Reno congestion control mechanism and compares it to TCP Tahoe through simulations run in NS2. 2) The simulations varied bandwidth between 1-14 Mbps and delay between 1-24 ms to observe their effects on congestion window size. 3) With TCP Reno, the congestion window size is halved rather than reset to 1 upon duplicate ACKs as in TCP Tahoe, allowing the network to recover more quickly from minor congestion.

Uploaded by

norranat songsriboonsit
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

SONGSRIBOONSIT Norranat Student ID: C3IM4505

Lecture 07 Report: TCP Reno

In this lecture, the mechanism of Reno, which is a window control protocol similar to Tahoe,
was discussed. However, to solve the problem of the extremely falling window size in Tahoe, the fast
recovery is implemented in Reno. The experiment was also conducted using the NS2 program and
the result is explained as follows.

Figure 1. The topology of simulated network

The network setup is almost the same as in lecture 6 where TCP Tahoe was studied. There
are two variables to be adjusted which are bandwidth (BW) and delay (d) between node 2 and 3
according to figure 1. The effects of each variable is studied and deliberately explained in this report.
The report discusses the impact of bandwidth first followed by the delay. In addition, the difference
between the TCP Reno and TCP Tahoe is also discussed in this report.

a.) b.)

c.) d.)
Figure 2. the results with the bandwidth of a.) 1 Mbps b.) 4 Mbps c.) 11 Mbps d.) 14 Mbps and delay of 20 ms

The first parameter of interest is bandwidth. Figure 2a shows four levels of bandwidth applied
to the network which are 1, 4, 11, and 14 Mbps in a, b, c, and d respectively. In figure 2a, starting from
SONGSRIBOONSIT Norranat Student ID: C3IM4505

a slow start phase, the congestion window size (cwnd) exponentially increases and reaches the initial
slow start threshold (ssthresh), which is set to be the same as maximum window size. Then, the
congestion is detected by duplicate ACKs. Different to Tahoe, the cwnd is not reset to 1 but it is
decreased by half. Then, even though the cwnd is already decreased, it still faces serious congestion
determined by the time out. Regarding the rules of Reno, the cwnd is reset to 1 because of the
serious congestion and ssthresh is halved again. Please note that the rule here is similar to the Tahoe
algorithm. After that, the cwnd enters the slow start phase again and turns into congestion avoidance
phase at around cwnd of 16. For 1 Mbps, the repeated congestion level after the network settles is
around 40. And the cwnd after that is caused by duplicate ACKs only.
For figure 2b, the behavior is similar to in figure 2a, except that the process occurs faster
because of the faster bandwidth and the level of repeated congestion becomes higher.
For figure 2c, the bandwidth is adjusted to be higher to see a different behavior of the system.
The very first part is similar to the figure 2a and 2b. However, after the cwnd was reset by serious
congestion, the bandwidth is high enough so that the system does not meet the congestion condition
during the simulation time.
The figure 2d shows a case when no congestion occurs in the system. The cwnd
exponentially rises in the slow start phase and continues rising linearly after entering the congestion
avoidance phase after exceeding the initial ssthresh at 64.

a.) b.)

c.) d.)
Figure 3. the results with the bandwidth of 4 Mbps and delay of a.) 1 ms b.) 2 ms c.) 10 ms d.) 24 ms

In this part, the effect of different levels of delay is explained thoroughly with various levels of
delay. Overall, higher delay is related to the slower process and larger window when congestion
occurs. At the first period of simulation, the behavior is similar to what explained in the previous
section of changing the bandwidth. Specifically, the cwnd increases exponentially to initial ssthresh,
SONGSRIBOONSIT Norranat Student ID: C3IM4505

then it is reset to half and 1 due to the duplicate ACKs and time out respectively. This event applies to
all four plots, however it occurs slower in the system with higher delay. After the first period, the cwnd
linearly rises and reduces to half repeatedly. The cwnd level that congestion occurs is approximately
42.
In figure 3b, at around 0.3 second, the cwnd increases exponentially in the slow start phases,
enters avoidance congestion phase, and increases linearly. A unique characteristic is shown in figure
3b, at around 44 where the congestion occurs, the cwnd and ssthresh are halved by duplicate ACKs.
Nevertheless, the congestion still persists and becomes a serious congestion detected by the time
out. Therefore, the ssthresh is halved again and cwnd is reset to 1. Please note that the congestion at
cwnd of 44 is larger than in figure 3a of 42 due to higher delay which means less congested system.
After that the rising and dropping of cwnd and ssthresh repeat along the simulation time.
Figure 3c and 3d are the cases that the delay is adjusted to be higher so that the different
behaviors can be observed. Figure 3c illustrates the similar result as in figure 3a. Nonetheless, the
size of variables including cwnd and ssthresh are higher because of increasing delay. Finally, in figure
3d, the system does not meet any congestion during the simulation time excluding one duplicate
ACKs and one time out at the first period. This is because the delay is very high and makes the
system become a very slow process.
In conclusion, the simulation result and influence of bandwidth and delay is similar to in the
Tahoe. However, Reno solves the problem when the need to reset the congestion window to 1 even
though the congestion condition is not serious. This solution helps Reno to be able to increase the
availability of the system compared to Tahoe.

You might also like