Journal of the Chinese Institute of Engineers, Vol. 33, No. 3, pp.

489-494 (2010) 489

Short Paper

CONGESTION CONTROL OF TRANSMISSION CONTROL

PROTOCOL BASED ON BANDWIDTH ESTIMATION

Hou-Tsan Lee*, Feng-Li Lian, and Ting-Chun Fong

ABSTRACT
This paper presents a framework for TCP congestion control, called “Bandwidth-
based TCP”, which differs from most TCP algorithms by using the bandwidth estima-
tion as the congestion measure to control the window size increment. It tries to predict
the equilibrium point of window size then make the congestion window approach this
point in a round-trip-time. First of all, an overview of TCP and AQM is introduced.
Then, the stability of the mechanisms is also investigated via linearization. Finally,
through the simulations, the performance of the proposed scheme is shown to be better
than TCP-Vegas under homogeneous and heterogeneous environments.

Key Words: Transmission Control Protocol (TCP), congestion control, round-trip-time

(RTT), bandwidth estimation.

I. INTRODUCTION Jacobsen, who proposed a mechanism called Tahoe,

The Internet plays a significant role nowadays
in our lives. Two main protocols are being imple-
mented in the transport layer of the Internet, namely,
UDP (User Datagram Protocol) and TCP (Transmission
Control Protocol). They are distinguished by their
connection type. UDP is a connectionless protocol
that suits for multimedia transmissions; on the other
hand, TCP is a connection-oriented one that is de-
signed to provide a reliable transmission policy in an
unreliable network, which may vary due to different
users, routers, bandwidths, or other cases. In recent
studies, Low et al. pointed out that 90% of the Internet
traffic is TCP-based, which is the main workhorse
and causes congestion (Low et al., 2002). To avoid
this problem, controlling the sending rate is neces-
sary and the issue has been discussed and developed
over the past quarter of a century.
In 1988, a major breakthrough was achieved by
*Corresponding author. (Tel: 886-226585801 ext. 5237; Email:
houtsan@takming.edu.tw)
H. T. Lee is with the Department of Information Technology,
Takming University of Science and Technology, Taipei 114, Taiwan,
R.O.C.
F. L. Lian and T. C. Fong are with the Department of Electrical
Engineering, National Taiwan University, Taipei 106, Taiwan, R.O.C.
which made the congestion window change dynami-
cally (Jacobson, 1988). In Tahoe, the essence is to
make the window size increase gradually until con-
gestion is detected, then it resets the window size to
one and starts to increase it again. After that, plenty
of studies were done to reach better performance based
on this idea. Two famous ones are Reno (Jacobson,
1990) and Vegas (Brakmo and Peterson, 1995). An-
other approach by Chen et al. proposed an idea that
uses the RTT to adjust the window size directly (Chen
et al., 2000). Also in (Srijith et al., 2005) and (Maor
and Mansour, 2003), the predefined thresholds were
dynamically adjusted to make TCP-Vegas more
adaptive.
On the other hand, some have put emphasis on
the router, which is in charge of delivering packets
from input links to output links rather than the sender.
The traditional, simplest, and most intuitive queue-
ing method implemented at the router is so-called Drop-
Tail, which drops the packet only when the buffer is
full. Recently, AQM (Active Queue Management)
has been the main topic in this area. As implied in
the name, researchers intend to develop another mecha-
nism to manage the queue more actively. RED (Random
Early Detection) is one of the most popular ones, where
one uses a marking probability to randomly drop the
490 Journal of the Chinese Institute of Engineers, Vol. 33, No. 3 (2010)
incoming packet before the occurrence of congestion
(Floyd and Jacobson, 1993).
The past few years have brought analytical models
into TCP/AQM systems. Given different analysis
methods, researchers can investigate the performance
of previous works and then rectify results afterwards.
This idea motivated this paper to work on a probe
into the TCP/AQM design (Hollot et al., 2001). A
new TCP framework with appropriate AQM mecha-
nisms is developed and analyzed for its feasibility.
The details of the proposed TCP mechanisms will be
explained in the following sections.
In this paper, a defect of TCP-Vegas, which has
been identified in recent years, is rectified. The model
called bandwidth-based TCP is developed to adjust
window size according to the estimated bandwidth.
In addition to presenting this idea, theoretical analy-
sis to support the proposed work is also shown in this
paper (Hollot and Chait, 2001). Furthermore, an ap-
proach based on TCP Westwood is also developed to
deal with the congestion problem. (Gerla et al., 2001).
Different operating systems and/or protocols of the
connecting computers and other devices are available
with the proposed scheme, thus one can conclude that
the environment is heterogeneous. Some control theo-
ries are quoted here to find out the equilibrium points
where the state variables will be in steady state first,
and then, linearization is used to deal with such a non-
linear system, and finally, stability can be examined
by inspecting the small neighborhood of these equi-
librium points. With the afore-mentioned analysis,
simulations are used to demonstrate that the proposed
work has some advantages over TCP-Vegas.
II. TCP-VEGAS
The flow control in TCP is window-based, where
a destination sends acknowledgments for packets that
are correctly received. A source keeps a variable
called window size that determines the maximum
number of outstanding packets that have been trans-
mitted but not yet acknowledged. When the window
size is exhausted, the source must wait for an ac-
knowledgment before sending a new packet (Low et
al., 2002). Therefore, the “window size” is an im-
portant feature that is always considered when de-
signing a TCP algorithm.
Figure 1 is an illustration of the so-called win-
dow size. In this figure, the sender sends four pack-
ets without any acknowledgements received until the
fifth transmission, so, at this time, the current win-
dow size is determined to be four, i.e., the number of
outstanding packets in the network that have been
transmitted but have not been acknowledged.
Tahoe, Reno, and Vegas are the most three sig-
nificant versions in the TCP design evolution. Both
Sender Receiver
Packet
W
Acknowledgment
Fig. 1 Illustrative example of window size
TCP-Tahoe and TCP-Reno change the adjustment of
their window size by self-generated congestion. TCP-
Vegas uses the round-trip-time (RTT) information to
vary the adjustment. The following subsections pro-
vide a quick overview of these three TCPs.
TCP-Vegas employs three techniques to produce
a higher throughput but lower losses than TCP-Reno.
These modifications are summarized as follows
(Brakmo and Peterson, 1995):
1. New Retransmission Mechanism—Uses more ac-
curate RTT estimation to decide to retransmit a
dropped segment.
2. Congestion Avoidance Mechanism—Gives a
method to measure and control the amount of ex-
tra data this connection has in transit, then adjusts
its transmission rate accordingly.
3. Modified Slow-Start Mechanism—Modifies TCP’s
slow-start to avoid packet losses while trying to
find the available bandwidth during the initial use
of slow-start.
The algorithm of TCP-Vegas is shown in Fig. 2.
Instead of self-generated congestion, as in TCP-Reno,
TCP-Vegas uses a more clever method of window size
adjustment to prevent packet losses. Once the first
acknowledgment is received, the sender calculates the
RTT and takes this value as a reference value, called
RTTmin. Henceforward, every RTT is compared with
it to get a performance index, called delta. When the
delta is less than a predefined value alpha (a kind of
threshold value), it implies the network is so smooth
that the measured RTT is close to the RTTmin, so the
sender can increase the RTT window size by one; on
the contrary, the sender decreases the window size by
one when the delta is larger than another constant value
beta. If the delta is between the alpha and beta,
 H. T. Lee et al.: Congestion Control of Transmission Control Protocol Based on Bandwidth Estimation
Slow-start
W_init ← 1;
delta ← (W/RTTmin-W/RTT)*RTTmin;
for each ack
if (delta < gamma) then
W ← W+1;
else
enter Congestion Avoidance
end if
Congestion Avoidance
delta ← (W/RTTmin-W/RTT)*RTTmin;
for each ack
if (delta < alpha) then
W ← W+1/W;
else if (delta > beta) then
W ← W-1; (only execute once in a RTT)
else
W ← W;
end if
Fast Retransmit
if (dup_ACKs)
retransmit the lost packet
W ← W/2;
enter Congestion Avoidance
end if
Fig. 2 Algorithm of TCP-Vegas (Brakmo and Peterson, 1995)
Vegas considers that the current state is good enough
to stay in, therefore freezes the window size.
III. PROBLEM FORMULATION
TCP-Vegas has shown its benefits over TCP-Reno.
However, some problems inherently exist in TCP-
Vegas that prevent it from being the most popular TCP
protocol. In TCP-Vegas, the increment of window
size is one of 3 alternatives, 1, 0, and -1. When ap-
plying this protocol to different network scenarios,
only three fixed increments are not sufficient to make
the window size get a steady-state value efficiently
all the time, i.e., in different situations, using this kind
of strategy may cause too conservative or too aggressive
behavior. The objective in this paper is to design a
TCP protocol that can achieve the following two goals:
1. Improve the convergence efficiency. Once a TCP
algorithm is designed, no doubt the mechanism can
be written as a state space form. By means of these
state equations, the equilibrium point of each state
variable is found afterwards. Note that the equi-
librium point of window size implies that no mat-
ter how the state trajectory changes in the transient
state, the window size still converges to a point
or a space at last. Therefore, the first objective of
this paper is to lower or reduce unnecessary varia-
tions in window size as much as possible, by al-
lowing more flexible behavior than TCP-Vegas, such
that the increment of window size could depend
491
Sender Router 1/B Receiver
B
1/B
Fig. 3 The badwidth estimating method
on the degree of network congestion, not only just
1,-1, and 0. The flexible behavior is expected to
reduce the cost of convergence, then improve the
convergence efficiency. Furthermore, if an equi-
librium point of window size is given, the best so-
lution is to approach that point directly without
producing any fluctuations.
2. Rectify the drawbacks in Vegas. Three problems
about using Vegas have been mentioned in the
previous section. Hence, the proposed TCP proto-
col is expected to remedy these defects, especially
in fairness.
IV. BANDWIDTH-BASED TCP DESIGN
When a system is controlled, some common de-
sign flows are usually used to analyze it. One con-
structs a model, finds out the equilibrium points, and
tries to approach the points or spaces. Using such a
design flow to deal with the TCP problem is tough
work as long as the number of users in the network
keeps changing. With respect to a sender, there is no
way to know how many users are sharing the link.
Hence, one can not compute the exact equilibrium
points of window size and queue length. In other
words, since the sender never knows how many users
are in the network, one is also unable to adjust the
window size in a logical way to approach the equi-
librium point, which dynamically varies its value.
In this section, a new TCP version called “Band-
width-based TCP” is proposed for tackling the prob-
lem identified above. By this mechanism, the sender
can estimate the available bandwidth first and subse-
quently predict the equilibrium state of congestion
window under this situation. Compared to TCP-
Vegas, the slow-start phase is slightly modified while
using a whole new adjustment manner for window
size in the congestion avoidance phase. This approach
includes the following three key features:
1. Modified Slow-Start,
2. Bandwidth Estimation, and
3. Bandwidth-based adjustment of window size.
Figure 3 illustrates the bandwidth estimation
method. If the bandwidth of a router is B (packets/
492 Journal of the Chinese Institute of Engineers, Vol. 33, No. 3 (2010)

sec), the router will send out packets to the destina- Slow-start
tion every 1/B seconds. Suppose there is no loss W_init ← 1;
ε _init ← 2;
during the transmission process, then the acknowl- for each ack
edging rate must be B as well. Therefore, once an if (RTT==RTTmin) then
acknowledgment is received, the sender records the W ← W+1;
current time and calculates the ACK interval by sub- ε ←ε ;
else
tracting the same value gotten in the previous period. enter Congestion Avoidance
Computing the reciprocal, the bottleneck bandwidth end if
estimation is accomplished.
If the router fairly processes the incoming pack- Congestion Avoidance
for each ack
ets from different senders, the ACK interval received estimate the bandwidth B
by a sender should become N times longer, where N if (!drop_flag) then
is the number of senders, and the estimated bandwidth ε = ε *2^(0.5-queueing delay);
{
B(k) is equivalent to B/N consequently. else
ε ← ε +1;
When the sender enters the congestion avoid- W ← (ε +B* RTTmin -W)/W;
ance phase, it executes a steady-state prediction ac- end if
cording to the bandwidth information introduced in
Fast Retransmit
the above section. We first decompose the round-
if (dup_ACKs) then
trip-time as retransmit the lost packet
W ← W/2;
q(k) ε ← ε /2;
r(k) = T p + 1 + , (1)
B B enter Congestion Avoidance
end if
where T p is the fixed propagation delay, q(k) is the
queen length at slot k, B is the bandwidth of a router Fig. 4 The algorithm of the bandwidth-based TCP
and B1 is the processing delay. To clarify the idea,
the adjustment action is divided into three steps. (α + β)/2 – q(k) (4)
= r(k) + .
B(k)
1. Calculate the queuing delay. Refer to the RTT
equation above, subtract RTTmin from the mea- At last, the equilibrium point of window size is pre-
sured RTT to get the queuing delay, i.e. dicted as

q(k) q(k) (2) α+β r*

(5)
r(k) – r min = (T p + 1 + ) – (T p + 1 ) = . w* = (
2 r* – r
).
B B B B min

2. Estimate the queue size. Since the source can not After a series of estimations, the sender lets the
get exact queue information from the router, it uses congestion window adjust linearly to the prediction
the above equation to estimate the current queue value w * in an RTT and then repeats these three steps
{

size, i.e., for every round-trip-time. The initial value of ε is

set as two in the beginning. After an RTT, the sender
{
q(k) = B(k)(r(k) – r min).
{
(3) calculates the queuing delay and then the growing
factor is decided. ε is increased continuously. When
3. Predict the steady-state value of RTT and window the network is smooth, the sender should have the
size. Using the idea in Vegas to decide the extra self-awareness to increase ε to lead to a larger win-
data in the network by dn(k), if the variation of con- dow size, and decrease otherwise. I f a dropped packet
gestion window reaches a steady-state value when is detected, ε becomes half of its original value. The
dn(k) equals (α + β)/2, the extra data in the network, growing rule is switched to decrease as soon as a packet
i.e., the queue size, is equal to (α + β )/2. Further- loss is detected. The algorithm of the proposed band-
more, current round-trip-time, denoted by r * , is width-based TCP is summarized in Fig. 4.
.
kept unchanged and can be decoupled like r = Tp + To give a complete model for stability analysis,
1 + (α + β)/2 . To predict this value, rewrite the three assumptions are given.
B B 1. Ignoring the mapping from price to marking prob-
above equation in terms of measured round-trip- ability p is acceptable. Hence price is the state
time r(k), estimated queue size q(k), and estimated
{
variable which stands for the marking probability.
2. Assume that a steady-state value of ε must exist
{
bandwidth B(k):
after some drops experienced by the sender, namely,
q(k) (α + β)/2 – q(k) the exponential growing only causes transient be-
r * = Tp + 1 + +
B(k) B(k) B(k) havior.
 H. T. Lee et al.: Congestion Control of Transmission Control Protocol Based on Bandwidth Estimation 493

8 Mb/1 ms 8 Mb/1 ms State variation of 1 sender BW-based TCP/REM with buffer = 100
0.8 Mb/5 ms 100

w (pkt)
S1 D1 window
Buf = 100(pkt) 50
R1 R2
0
Fig. 5 A one sender configuration 0 5 10 15 20 25 30
100

q (pkt)
queue
50
3. The router processes the incoming packet fairly. 0
Thus the sender measures its available bandwidth 0 5 10 15 20 25 30

Epsilon (pkt)
100
precisely and it can be equivalent to B/N. epsilon
50
Now, the state equation of window size w(k) of band-
width-based TCP is shown as follows: 0
0 5 10 15 20 25 30
100
w(k + 1) = ε(k) + B(k)r min – 1 p(k)w 2(k) (6) 50
p

p
2
{ 0
Replacing B(k) with B/N, the difference equation is re- 0 5 10 15 20 25 30
written and therefore solves the equilibrium problem Time (s)
Fig. 6 State variation versus time in Bandwidth-based TCP
w(k + 1) – w(k) = 0 ,

⇒ 1 p *w *2 + w * – (ε * + B(k)r min) 60
Window size comparison
2
Modified BW-based
= 1 p *w *2 + w * – (ε * + B (T p + 1 )) = 0 , 50
2 N B
BT p + 1
⇒ 1 p *w *2 + w * – (ε * + ) = 0. (7) 40
2 N
w(k) (pkt)

Besides, a third-order equation is given below 30 BW-based (20, 30)

1 w *2 + w * – (ε * + BTp + 1 ) = 0 ,
ε* N
BTp + 1 2
⇒ ε *3 – (w * – )ε * – w *2 = 0 ,
N Vegas (1, 3)
Buf 2
⇒ ε *3 – ε * – w *2 = 0 .
2N
The analysis of the stability of the proposed al-
gorithm can be proved by linearization of the state
transfer function matrix A and then A can be proved
as a Hurwitz. The complete analysis of bandwidth-
based TCP is omitted here. Whereas the defects of
TCP-Vegas have been identified, the proposed algo-
rithms which try to rectify these problems are pre-
sented in this section. Next, the simulations under
different network scenarios are demonstrated to show
the performance comparisons.
V. SIMULATION RESULTS
After presenting the design in Section IV, simu-
lations via NS-2 (The Network Simulator – Version
2) are demonstrated in this section under a heteroge-
neous environment as below, where three perfor-
mance indexes are given as the criteria to judge the
quality of a TCP protocol.
First, an observation on the time response of pro-
posed bandwidth-based TCP is taken. A simple one-
connection configuration, which is shown in Fig. 5,
20 Vegas (20, 30)
10
BW-based (1, 3)
(8) 0
0 5 10 15 20 25 30
Time (s)
Fig. 7 Window size comparison between the bandwidth-based
TCP and Vegas
is used with the settings that both the sender-router
and router-receiver are 8 Mbps with 1-ms propaga-
tion delay, and the bottleneck router-router uses a 100-
packet buffer space with 800 Kbps and 5-ms delay.
To demonstrate the analytical result, Fig. 6 shows the
trajectories of all the state variables defined in Sec-
tion IV, viz. w, q, ε , and p. For each row in Fig. 6,
the x-axis shows the time in seconds, and the y-axis
shows the window size in packets (1 kbits/packet) except
the bottom row, where the y-axis shows the marking
probability of state p instead, where p is the mapping
from price to marking probability. Fig. 6 shows the
closed relation between p and queue. Since there is
only one user occupying the link, the sender itself
can predict the equilibrium points by precisely mea-
suring the bandwidth without error.
Under the same environment, Fig. 7 shows a
window size comparison among TCP-Vegas (algorithm
494 Journal of the Chinese Institute of Engineers, Vol. 33, No. 3 (2010)
in Fig. 2), the bandwidth-based TCP (algorithm in Fig.
4), and the previous version of bandwidth-based TCP
without adaptive modification included (Eqs. 1-4),
where the x-axis shows the time in seconds, the y-
axis shows the window size in packets, response of
the bandwidth-based TCP, and the line with annota-
tion T( α , β ) denotes the time response of a TCP pro-
tocol T with two thresholds ( α , β ). Even though
every algorithm can reach a steady state in this simple
case, the selection of ( α , β ) deeply influences the
equilibrium of window size in both the previous ver-
sion and TCP-Vegas. However, the proposed band-
width-based TCP excludes this uncertainty.
The one-sender configuration is given to show
the key behavior of the proposed algorithm. From
this, the benefit of bandwidth-based TCP can be seen–
adjusting the congestion window independent of the
external threshold settings. Furthermore, the window
size is larger than the other two.
VI. CONCLUSIONS
In this paper, the bandwidth-based approach in
TCP congestion control is proposed. It has three features:
modified slow-start, bandwidth estimation, and band-
width-based adjustment. In brief, it predicts an equi-
librium point of window size then approximates it based
on the measured bandwidth in every round-trip-time.
Through the simulations, the proposed scheme is shown
to have better performance than Vegas under a ho-
mogeneous and a heterogeneous environment.
Although the bandwidth-based system is superior
to TCP-Vegas according to some analysis and simula-
tion results, there are still some problems to be solved.
The key issue is the steady-state behavior. For the sake
of competing with TCP-Reno, which has better perfor-
mance in window size, the bandwidth-based TCP makes
the expected queue size grow as does TCP-Reno’s be-
havior such that window size behavior is similar, i.e., it
fluctuates within a certain range. This violates the con-
cept of TCP-Vegas– to stop increasing window size in
advance, and this problem will be tackled in the future.
ACKNOWLEDGMENTS
This research was supported in part by the Min-
istry of Education, and the National Science Council,
Taiwan, ROC, under the grants: 98-2218-E-002-008,
98-2221-E-002-160-MY3, 99-2623-E-002-007-D.
The authors would like to thank the anonymous ref-
erees whose careful reviews and detailed comments
helped to improve the readability of this paper.
NOMENCLATURE
B bandwidth of a router (packets/sec)
N the number of senders
q the queen length
r round-trip-time
Tp the fixed propagation delay
w window size
REFERENCES
Brakmo, L. S., and Peterson, L. L., 1995, “TCP Vegas:
End to End Congestion Avoidance on a Global
Internet,” IEEE Journal on Selected Areas in Com-
munications, Vol. 13, No. 8, pp. 1465-1480.
Chen, J. R., Chen, Y. C., and Lee, C. L., 2000, “TCP
Vegas-A: Improving the Performance of TCP
Vegas,” Computer Communications, Vol. 23, No.
16, pp. 1537-1547.
Floyd, S., and Jacobson, V., 1993, “Random Early
Detection Gateways for Congestion avoidance,”
IEEE /ACM Transactions on Networking, Vol. 1,
No. 4, pp. 397-413.
Gerla, M., Sanadidi, M. Y., Wang, Ren, Zanella, A.,
Casetti, C., and Mascolo, S., 2001, “TCP West-
wood: congestion window control using band-
width estimation,” IEEE GLOBECOM ’01, Vol.
3, pp. 1698-1702.
Hollot, C. V., Misra, V., Towsley, D., and Gong,
W. B., 2001, “On Designing Improved Control-
lers for AQM Routers Supporting TCP Flows,”
in Proceedings of IEEE INFOCOM, Vol. 3, pp.
1726-1734, Alaska, USA.
Hollot, C. V., and Chait, Y., 2001, “Nonlinear Stabil-
ity Analysis for a Class of TCP/AQM Networks,”
in Proceedings of IEEE CDC, Vol. 3, pp. 2309-
2314, Orlando, FL, USA.
Jacobson, V., 1988, “Congestion Avoidance and
Control,” in Proceedings of ACM SIGCOMM, pp.
314-329, Stanford, CA, USA.
Jacobson, V., 1990, “Berkeley TCP Evolution from 4.
3-Tahoe to 4.3-TCP-Reno,” Proceedings of the 18th
Internet Engineering Task Force, Vancouver, BC,
Canada.
Low, S. H., Paganini, F., and Doyle, J. C., 2002,
“Internet Congestion Control,” IEEE Control Sys-
tems Magazine, Vol. 22, No. 1, pp. 28-43.
Moar, A., and Mansour, Y., 2003, “AdaVegas: Adap-
tive Control for TCP Vegas,” Proceedings of the
IEEE GLOBECOM’03, Vol. 7, pp. 3647-3651.
Srijith, K. N., Jacob, L., and Ananda, A. L., 2005,
“An End-to-End Flow Control Approach Based
on Round Trip Time,” Computer Communica-
tions, Vol. 28, No. 4, pp. 429-440.
Manuscript Received: Mar. 18, 2008
Revision Received: Oct. 25, 2009
and Accepted: Nov. 25, 2009