TCP QIC: Quadric Increase Congestion Control algorithm

Implementation and simulation

Girish Paliwal Swapnesh Taterh
[email protected] [email protected]
Amity University Rajasthan Jaipur Amity University Rajasthan Jaipur

Abstract: Now these days most of the devices having internet connections that increased the traffic over the server and TCP provide the
connection-oriented connection. TCP may be experiencing issues with using the total bandwidth of the communication channel. Numbers
of congestion control proposals have already been suggested to reduce this problem. This paper presents the implementation of a quadric
increase congestion control algorithm and it’s a simulation through the ns3. This algorithm is based on binary increased congestion control
algorithm (TCP BIC). In ns3 TCP QIC is separately implemented and tested with the different congestion control algorithms. These
congestion control algorithms are TCP Westwood, BIC, NewReno, scalable and Illinois. The performance of the TCP QIC has the
significance over the other congestion control algorithms in respect of throughput, goodput, delay variance and round-trip time.
Keywords: Congestion control algorithm, TCP, NS3, TCP QIC: Quadric Increase Congestion control algorithm.

1. Introduction 2. TCP Basics

Over the past decades, so much work has been already done This provides a connection-oriented connection between
to enhance the TCP performance and the result many different sender and receiver. The Connection-oriented communication
congestion control algorithms. The number of congestion means TCP send packets between computers provides an
control algorithms are already developed and implemented after acknowledgment number through the sequence number.
finding some enhancement result over the existing congestion Unacknowledged TCP packets are retransmitted based on the
control algorithms [1]. The implementation and comparing them information about successfully received sequence number and in
with other congestion control algorithms is troublesome without some cases, round trip time out of the packet is used for
network simulators. retransmission of the packet. TCP congestion control algorithm
In the real world scenario, it is very difficult to compare proposed over the last 30 years, roughly to maximize the
algorithm performance on the basis of actual devices it is throughput and avoiding the congestion at the internet routers
possible with the simulators it is less time to consume and less and provide the services effectively and efficiently.
expensive and given the moral as actual results. Computer TCP congestion control algorithms have been having some
simulator provides valuable insight into potential common features these features are on the basis of changing in
implementation before extending the considerable effort that congestion window acknowledgment received round trip time of
may be required to build a technology simulator allow us to a packet. On the basis of these parameters each and every TCP
measure and compare the existing algorithm as well as we can congestion control algorithm wants to provide good throughput
introduce a new algorithm and compare in light of existing and Plus latency time and use the maximum capacity of the
algorithms. It provides a much more flexible way to implement channel [4].
a new algorithm and finding out the related result on the basis of TCP congestion window control the amount of packets a
some existing networking topology [2]. TCP flow they've in the network whenever you want the
This paper presents an implementation of the QIC congestion congestion window is defined using additive increase
control algorithm that is based on the TCP BIC. In order to the multiplicative decrease AMD mechanism that probes the
ns3 TCP BIC that is most widely used by the Linux operating available bandwidth dynamically adapting to changing network
system [3]. I enhance the congestion control algorithm TCP conditions the AMD mechanism work well when the sender
back that use the methodology of the congestion window continuously has data to send as is typically the case for TCP
increase in a binary manner whereas I will change a little bit in used for bulk data transfer current TCP implementation have a
this algorithm to check the performance of quadric congestion variety of the behaviors for setting up after an ideal period some
control window increase the effect is immense to attract our current TCP implementation slow start after an ideal period
attention. longer than the RTO estimates as suggested in [5] RFC 2581.
The rest of this paper is organized as in the following While other implementations aren't reduced their congestion
sections the basic of the TCP in section II, section III related window after an ideal period. Once the congestion window is
work of the TCP congestion control algorithm, section IV reduced the slow start threshold remains as a memory of the
simulation setup, and section V have the simulation result matrix recent condition window. sending of back again to back packets
the last section is the conclusion of the present quadric are facing packets out over the period of a circular trip time and
congestion control algorithm performance. old congestion window that has not been fully using for some
time can't be trusted being an indication of the bandwidth
currently readily available for a floor we'd contain that the The detailed description of TCP BIC algorithm could be
mechanism to piss out packet allowed by the congestion window found in [10]. A diagram of this algorithm during the
are largely used to determine the correct size of the congestion Congestion Avoidance phase is shown in figure 1
window. [6]
The Mechanism to determine the RTO Round Trip Timeout,
in the most up-to-date ideal period could also be implemented
using the timer every packet received RTO after check always
per packet. Figure 1 TCP BIC congestion avoidance representation

3. Related Work
That is a very challenging task to style a fresh protocol that
may satisfy all of the criteria like in congestion control these
criteria are RTT fairness TCP friendliness and scalability [7]. A
protocol should adopt its window control with regards to the size
of the window. TCP Bic the congestion control problem is
viewed as a research problem [8]. The TCP BIC is binary
increase congestion control that consists two parts binary search
increase and additive increase.

3.1. Binary search increase

Starting points with this search our current maximum
window size and maximum window size usually Wmin and W
max are the windows I just before the past recovery. The
algorithm rapidly computes the midpoint between Wmax and
Wmin set the present window size to the midpoint and check for
feedback in the proper execution of packet loss. 4. Simulation set up
If there is a package loss and then present threshold
In this section, I have evaluated and analyzed the different
minimum increments. This technique is called binary search
type of congestion control algorithms as TCP Westwood, BIC,
increase allow bandwidth propping to become more aggressive
NewReno, scalable, Illinois, and QIC. I am using a simple
initially when the difference from the present window size to the
network scenario and applied through the network simulator
mark window size is large and become less aggressive as the
NS3. To test the behavior of each Protocol in the bottleneck
present window size gets nearer to the mark window size [9].
bandwidth network, I used simple simulation Technology
consisting of a single sender and a single receiver interconnected
3.2. Additive Increase
to each other through a router at the other end; this is shown in
To make certain convergence at RTT fairness the binary Figure 2. Here I create a point to point connection between the
search in trees having an additive increase strategy also. Once sources and sink and simulate according to different bandwidth
the distance from the midpoint to a current minimum is too big channel between server to the router and varying the time of
increase window size right to the midpoint. Once the distance simulation. Each TCP congestion control algorithm is an
from the present window size of the mark is less compared to analysis on the basis of the congestion window, Round trip time
the Smax, during those times window increase right to the mark and finally gets the throughput. Table 1 shows the network
[9]. simulator parameters that used in simulation through NS3. [11].

3.3. Slow Start

The TCP congestion control algorithm runs a “slow start” Figure 2 TCP congestion Simulation network Scenario
strategy to probe for a fresh maximum, once the present window
size is more than Wmax but smaller than Wmax+Smax. The
congestion window increases in each RTT round in steps. 10 Mb/s, 0.45 ms 2Mb/s,0.01ms
Under the fast convergence, the binary search increase
combines in multiplicative decrease converges to great amount
search. Considering that the large window reduces more and
multiplicative decrease. The fast convergence strategy the binary
search increase following a window reduction new maximum
and minimum are set. Than Readjust the new condition window Table 1 Simulation variables and its values
maximum size to be the same as the new target window that is
the midpoint and then we adjust the mark and apply binary
search it has an effect to reduce the increasing rate of the big Parameters Values
window allows the smaller window to catch up this strategy Bottleneck Bandwidth 2 Mb/s
Time of Simulation 50 seconds
called first convergence [9].

2
 Queue Type PfifoFastQueueDisc On the basis of the simulation result, it is studied that the
Nodes 2 throughput of the TCP QIC having some substantial value in
Delay 0.01 ms comparing to another congestion control algorithm. This is
TCP Congestion Control TCP Westwood, BIC, NewReno, represented by the figure 3. In this graph the throughput
Algorithms scalable, Illinois, and QIC representation in increasing order.
Access Bandwidth 10 Mb/s
Access Delay 0.45 ms
Figure 3 Throughput comparison between congestion control
algorithms in increasing order

5. Result evaluation matrix

Where the performance of the congestion control algorithms
evaluated three major performance metrics is really as average
throughput, average goodput, average Queue delay and the
typical package reception. These metrics are calculated on the
bottleneck link in both directions. To pull these values trade
shows are established along the network device on the router
every time a packet is received by the unit the method. Which
calculates performance metrics assignment of the timestamp to
packet while they enqueued and dequeued is implemented by
the callback function.

5.1. Throughput
It's measured as the ratio of the full total amount of receiving
data and the full total simulation run time [11].
Average Throughput =Total data received /
Simulation time

5.2. Goodput
Figure 4 Goodput graph for bottleneck bandwidth 2Mb/s in
In computer networks, goodput is the application- level increasing order
throughput. The quantity of data considered excludes protocol
overhead bits as well as retransmitted data packets

5.3. Delay (End to End)

dend-end= N[ dtrans+dprop+dproc+dqueue]
Where
dend-end= end-to-end delay
dtrans= transmission delay
dprop= propagation delay
dproc= processing delay
dqueue= Queuing delay
N= number of links (Number of routers - 1)
Note: we have neglected queuing delays.
The average end-to-end delay measured by the full total time
taken by sent packets, divided by the full total time taken by
received packets.

5.4. Packet Delivery Ratio

It's a percentage of a bundle generated by the source and the In figure 4 represent the goodput graph. The throughput and
number of packets sent to the destination [11]. goodput result of the simulation 50 seconds for 2 Mb/s
Packet delivery time = Transmission time + Propagation bottleneck bandwidth. I have simulated with different cases of
delay bandwidth (2,5,10) and simulation time(25,50,80) but the
observation is that the when bandwidth increasing the
5.5. Packet Loss congestion windows is increasing significantly and congestion
The number of dropped packet when the buffered becomes does not occur for this scenario that’s why here only one
full or stored, packed time to live exceeds simulation results describe.
In the figure 5shows the end to end delays in transmission. In
this graph, the average delay calculated that show the lowest
6. TCP QIC analysis with others
average delay TCP Westwood, NewReno, and Scalable and

3
highest average delay TCP Illinois and BIC. Whereas, TCP QIC Figure 7 packet send speed
have an average delay between lowest to highest. In this respect,
the TCP QIC performance is more honest than the TCP BIC and
Illinois.

Figure 5 Average end to end delay representation

If the data packet sends speed higher but receiving speed not
similar to the sending speed in this situation congestion occur
and data packets are left out and retransmission of same data
The following figure 6 indicates the number of packets sends
packets. This is affecting the throughput and average end to end
by each TCP congestion control algorithm in increasing order
delay.
these are interpreted by the graph. It is very simple logic when
the number of packets sent by any algorithm then data more
transmitted and it shows the maximum utilization of the 7. Conclusion
bandwidth and produced maximum throughput. It is proved by This paper proposes a TCP congestion control algorithm
the above and below graphs that the TCP QIC is having an based on the TCP BIC that is a quadric congestion control
advantage in terms of throughput and data transmitting. (QIC). TCP QIC is a simulation of the different parameter
values using the ns3 simulator. The simulation metrics are
Figure 6 Packet received in 50 seconds simulation generated after 50-second simulation each of the congestion
control algorithms simulated same time. The different variety of
graph plotted on the basis of simulation Matrix result that
represents the TCP QIC having better throughput and Goodput.
TCP QIC having efficient data packet transmission speed and it
also receives a maximum number of packets within a 50 second
of simulation time. TCP QIC average delay time is higher from
lowest delay time but it is lower than the highest delay time.
Finally, I say that the TCP QIC having better performance than
the other TCP congestion control algorithms as Westwood,
Illinois, NewReno, BIC and scalable.

REFERENCES
[1] K. Nagori, M. Balachandran, A. Deepak, M. P.
Tahiliani, and B. R. Chandavarkar, “Common TCP
Evaluation Suite for ns-3,” pp. 9–16.
[2] P. Yang et al., “TCP Congestion Avoidance
Figure 7 represents the data packet transmission speed that is Algorithm Identification,” IEEE/ACM Trans.
represented by the lambda. Lambda is the data packet send per Networking, vol. 22, no. 4, pp. 1311–1324, 2014.
second. If data packet sends per second is higher that means the [3] Brett Levasseur, Mark Claypool,Robert Kinicki, “A
data packed acknowledgment received speed is higher, that TCP CUBIC Implementation in ns-3 ∗,” vol. In
show the congestion prevention and congestion control is Proceedings of the 2014 Workshop on ns-3 (p. 3).
efficiently handled by the TCP QIC rather than TCP BIC, ACM., pp. 1–12, 2014.
Westwood, NewReno, Illinois and Scalable.
[4] L. A. Grieco and S. Mascolo, “Performance
evaluation and comparison of Westwood+, New

4
 Reno, and Vegas TCP congestion control,” [9] L. Xu and K. Harfoush and I. Rhee, Binary Increase
SIGCOMM Comput. Commun. Rev., vol. 34, no. 2, Congestion Control (BIC) for Fast Long-Distance
p. 25, 2004. Networks. Piscataway, N.J: IEEE, 2004.
[5] “RFC 2581 - TCP Congestion Control,” [10] SeyedShams Feyzabadi, “Identifying TCP
[6] “RFC 3465 - TCP Congestion Control with Congestion Control Mechanisms Using Active
Appropriate Byte Counting …ABC—,” Probing,”
[7] W. Hua and G. Jian, “Analysis of TCP BIC [11] Girish Paliwal and Swapnesh Taterh, Ed., Impact of
Congestion Control Implementation,” pp. 781–784. Dense Network in MANET Routing Protocols
[8] https://www.nsnam.org/docs/models/html/tcp.html. AODV and DSDV Comparative Analysis Through
NS3: Proceedings of SoCTA 2016, Volume 1.
Singapore: Springer Singapore, 2018.