Experimental Analysis of Concurrent Multi-path Transmission

Schemes

Varun Kumar Sharma

, Lal Pratap Verma

, Mahesh Kumar

Department of Computer Science and Engineering,The LNM Institute of Information Technology, Jaipur, Rajasthan, India

Department of Computer Science and Engineering,Moradabad Institute of Technology, Moradabad,

Uttar Pradesh, India

Department of Computer Science and Engineering,Jaypee University of Engineering and Technology, Guna, Madhya

Pradesh, India

Keywords: Concurrent Multi-Path Transmission, MPTCP, SCTP, Wired Networks.

Abstract: The mobile hosts having multi-interfacing capability can increase their performance (i.e., throughput) by

effectively making use of concurrent transmissions over multiple available network paths. Nevertheless, this

policy severely faces degradation in application-level throughput performance because of highly dissimilar

characteristics (i.e., different delay and available bandwidth) of multiple network paths. In particular, the

dissimilar paths’ characteristics issue will persistently cause data to be received disordered (i.e., Out-Of-

Order (OOO)) and due to negligibility of this issue, serious degradation in application-level throughput

performance will occur. Consequently, these mentioned issues will further create the buffer blocking

problem in the system and hence degrades the performance to a greater extent. In this paper, we evaluate

and present an analysis of the concurrent transmission policies’ performance over varying path

characteristics. And we will understand what is the behaviour of the suggested concurrent transmission

policies when the paths’ characteristics are highly varied.

1 INTRODUCTION

The traditional Layer-4 (Transport Layer) protocols,

TCP (Postel, 1981) and UDP (Postel, 1980), are

fundamentally ignore the use of multi-homing

feature. Specifically, TCP allows dynamic binding

to only single network address at each end of the

connection. At that time when TCP’s functionality

was suggested, the multi-interfacing capability were

known to be inefficient because of expensive

hardware requirements, hence, multi-homing feature

was beyond the scope of interest for the researchers.

Then, as the time passes, the desire for

communication (data exchange) to be highly fault

tolerant between end-to-end hosts, have conveyed

multi-homing feature within the scope of interest for

the researchers (Iyengar et al. 2006) (Sharma et al.

2019) (Verma and Kumar, 2017) (Verma et al.,

2018) (Wallace and Shami, 2014)

The classic protocols engaged in heterogeneous

network interface utilization are SCTP (Stewart et

al., 2000) and MPTCP (Raiciu et al., 2011) (Ford et

al., 2013) (Ford et al., 2011) (Paasch, and

Bonaventure, 2014). In direction to transmission

association establishment inclusive of numerous

paths amid two end hosts, SCTP and its Concurrent

Multi-path Transfer extension (CMT-SCTP) (Verma

et al., 2018)exploit multi-homing. SCTP principally

supports multi-homing competence and offers

functionalities such as congestion and flow control,

reliability and ordered data delivery (Natarajanet al.,

2013). MPTCP offers the capability of concurrent

utilization of numerous available network paths

amid two end hosts, which has gained immense

industrial attention and absorption from societies of

research and standardized bodies (i.e., IEEE and

IETF). In particular, the researchers have suggested

the idea of combining the advantages of TCP and

CMT in MPTCP. In this perspective, CMT

dynamically exploits numerous available interfaces

to effectively schedule data in concurrent fashion.

Hence, CMT has an exceptional capability of

providing significant fault tolerance, bandwidth

aggregation and proper load balancing demands to

multiple resource constraint applications.

Nonetheless, the classical data transmission policies

Sharma, V., Verma, L. and Kumar, M.

Experimental Analysis of Concurrent Multi-path Transmission Schemes.

DOI: 10.5220/0010561600003161

In Proceedings of the 3rd International Conference on Advanced Computing and Software Engineering (ICACSE 2021), pages 5-11

ISBN: 978-989-758-544-9

(e.g., CMT, CMT during Path-Failure (CMT-PF)

(Natarajan et al., 2008)and CMT-SCTP) schedule

the data over network paths in round-robin fashion.

However, these approaches have not includes the

path variable properties (e.g., available BW, path

quality, and delay) of multiple network paths and

transmit the data packet blindly. Consequently, these

scheduling policies undeniably cause severe OOO

data delivery at receiver which affects the

performance significantly. Subsequently, OOO

delivery causes unnecessary fast retransmissions and

redundant congestion window (cwnd) reductions as

well. Additionally, it leads to the issue of buffer-

blocking problem in the network. The buffer-

blocking problem severely restricts the possibility of

data communication (exchange) and ultimately

makes the connection idle, which subsequently

degrades the performance in terms of average

throughput and goodput respectively. Also, it leads

to higher transmission delays and increases spurious

retransmissions in the network.

The problem of redundant fast retransmissions

can be dodged by designing and applying an

efficient scheduling strategy. However, the most

standard and common scheduling policy (i.e. round-

robin scheduling policy) have not included the path

dissimilar characteristics andschedules the

transmission abruptly. Hence, the researchers have

suggested numerous data scheduling policy (Sharma

et al. 2019) (Verma and Kumar, 2017) (Verma et al.,

2018)(Paasch, 2014) which effectively consider

varying path characteristics, and hence reduces the

re-ordering (packet) at the receiver’s end. In this

paper, we evaluate and present an analysis of the

CMT, CMT-PF and MPTCP’s performance over

varying path characteristics.

The structure of this work as follows. Section 2

discusses the related works in the field of CMT.

Section 3 discusses the simulation setup and

environment used to test different CMT schemes.

Section 4 discusses the experimental results. Section

5 finally concludes this work.

2 RELATED WORK

In order to describe about the foremost issues

affecting several SCTP and MPTCP based proposals

is the main objective of this section. Furthermore,

we will look at what the researchers have suggested

to reduce all the issues related with SCTP and

MPTCP based proposals in this section.

2.1 SCTP based Proposals

SCTP does not support CMT; hence, Iyengaret al.

(Iyengar et al. 2006) suggested an optimize version

SCTP also known as CMT-SCTP. CMT-SCTP

minimizes the reordering problem in CMT such as:

superfluous fast retransmissions; SCTP sender

wrongly interprets the reason for a packet loss, that

is, network congestion. Consequently, SCTP sender

needlessly reduce its cwnd size which majorly

influences the network’s average throughput

performance. However, there could be another

reason for the unordered packet, i.e., a packet could

be delayed somewhere at the longer path. Moreover,

excessive acknowledgment (ACK) traffic and

receiver buffer blocking are other serious issues

which are effectively handles by CMT-SCTP.

Natarajan et al. (Natarajan et al., 2008) (Natarajan et

al., 2013) have presented buffer-blocking problem

associated with CMT, and authors show “how this

problem severely hampers the performance of CMT

during permanent (long) and short-term packet

losses”. Then, Verma and Kumar (Verma et al.,

2018) investigated that the load distribution policy

of CMT and CMT-PF blindly schedules the load

over network paths without considering the path’s

bandwidth and delay variations. Hence, the authors

have given an adaptive data chunk distribution

policy for CMT (A-CMT) which effectively

schedules the transmission load over paths

considering both delay variations and available

bandwidth factors. Still, an effective and deep

evaluation of these above-mentioned approaches is

indeed needed in highly lossy environment (e.g.,

Mobile Ad-hoc Networks (MANETs)). Hence, Xu et

al. (Xu et al., 2013) have given Quality-Aware CMT

(CMT-QA) for heterogeneous environment and

extensively evaluates their approach compared to

CMT and CMT-PF. Nevertheless, CMT-QA solely

focuses on throughput performance enhancement

and lacks concentration over fairness towards other

competing TCP traffic flows. In addition, CMT-QA

determines the packet losses only at Layer-4 level,

which makes this approach highly inefficient in

wireless environment, where appropriate buffer-

overflow induced and channel characteristics

induced loss classification (see details in (Sharma

and Kumar, 2017) and references therein) is highly

required. Also, many of the other proposals (Wallace

and Shami, 2014) (Perotto et al., 2007), likewise

CMT-QA, depends exclusively on Layer-4 Quality

of Service (QoS) based parameters and also these

policies do not consider other reasons for a packet

drop apart from buffer overflow, hence, these

ICACSE 2021 - International Conference on Advanced Computing and Software Engineering

policies suffer lower performance. Recently,

Network-Coding (NC) based Layer-4 policies (Xu et

al., 2015) (Xu et al. 2016) (Xu et al. 2017) have been

proposed and has by now been confirmed a

proficient method to solve buffer-blocking issue.

With this we come to finish our short argument on

suggested CMT solutions. Interested scholars and

researchers can refer to Habib et al. (Habib et al.

2016), Xu et al. (Xu et al., 2016), Wallace and

Shami(Wallace and Shami, 2012), Beckeet al.

(Becke et al., 2013) and Li et al. (Li et al., 2016)for

further profound study on different CMT related

solutions. Further interested researchers can refer to

Sharma et al. (Sharma et al., 2019), Sharma and

Kumar (Sharma and Kumar, 2017), Sharma et al.

(Sharma et al, 2018)(Sharma and Kumar, 2017)

(Sharma et al., 2018) (Sharma et al., 2012) (Sharma

et al., 2020),Kanellopoulos and Sharma

(Kanellopoulos and Sharma, 2020), and(Sharma,

2019)for congestion and energy aware solutions for

single and multi-path Layer-4 protocols as well.

2.2 MPTCP based Proposals

The multi-pathing approach was inescapable and

researchers were confident about that since

architecturescontinuallysearching more feasible

solutions. This, in return, resulted in MPTCP’s

policy development. Initially, numerous congestion

control policies have been suggested which

straightaway extended TCP NewReno for the

purpose of designing MPTCP’s policy (attributed as

Linked Increase Algorithm (LIA) (Raiciu et al.,

2011)), i.e., suggested policies directly triggers the

TCP NewReno’s functionality independently on

each sub-path. This direct extended version can lead

to severe un-fairness in the network for single-path

TCP users when the obtainable network paths share

bottleneck links with network paths used by MPTCP

users. Hence, several researchers have suggested

numerous mechanisms (Kelly and Voice, 2005)

(Han et al., 2006) (Wang et al., 2003) with the

intention of structuring an efficient multi-path

Layer-4 protocol, specifically, compatible with the

standard TCP (i.e., Coupled Congestion Control

(CCC) algorithm). These suggested algorithms in

(Kelly and Voice, 2005) (Han et al., 2006) (Wang et

al., 2003) to utilize only the best available paths to

the users and are best suited for the conditions where

similar or little variations in Round Trip Time

(RTTs) has been observed. Nevertheless, these

algorithms suffer from the problem of flappiness

(see details of flappiness and Opportunistic Linked-

Increases Algorithm (OLIA) in (Khalili et al., 2013)

and references therein) and lesser responsiveness.

Firstly, these algorithms sometimes fail to adapt

rapidly, in particular, they do not able to probe the

paths with higher channel and congestion induced

loss probabilities, hence, makes these algorithms

much lesser responsive. Secondly, these algorithms

show severe flappiness in the network. In particular,

in order to resolve the lesser responsiveness issue

associated with CCC algorithm Wischiket al.

(Wischik et al., 2011) have suggested a novel Half-

coupled congestion control mechanism which shows

more responsiveness and it is more friendly towards

single-path TCP users as well. Nevertheless, Peng et

al. (Peng et al., 2013) (Peng et al., 2016) further

claimed that Half-coupled congestion control

mechanism un-friendliness towards single-path TCP

users gets inflated during dissimilar RTT variations

of each available sub-paths. Subsequently, the

authors have identified design standard that give

assurance of uniqueness, stability of system and

existence. Their approach (attributed as Balanced

Link Adaptation Algorithm (BALIA)) mainly

focused on performance metrics such as

receptiveness, TCP friendliness and window

(congestion) variations. Singh et al. (Singh et al.,

2013) have further claimed that there are still some

performance issues have been associated with OLIA

when all the available sub-paths are congested. They

have suggested Adapted OLIA (AOLIA) policy

which effectively controlled the aggressiveness of

MPTCP in terms of cwnd growth scheme.

3 SIMULATION SETUP AND

PARAMETERS

The simulation has been carried out on Network

Simulator-2 (ns-2) with in-built MPTCP module

originally modelled by Nishida (Nishida, 2013). The

experiments considered the wired network

environment shown in Fig. 1. As the figure

demonstrates, end hosts ‘S’ and ‘D’ are attached

with two interfaces ‘S1–S2’ and ‘D1–D2’

respectively whose configuration parameters are

listed in Table 1. Whereas, end host ‘S’ is attached

to single-homed routers ‘R1’ and ‘R3’, which

further introduce heavy cross-traffic to simulate

severe congested situation over network paths

(PATH-1 and PATH-2). For this, the routers ‘R1’

and ‘R3’ are attached with UDP with Constant Bit

Rate (CBR) traffic generating agent whose

configuration parameters are listed in Table 1. In

particular, there are two available network paths i.e.,

Experimental Analysis of Concurrent Multi-path Transmission Schemes

PATH-1 and PATH-2 with bottlenecks in the

simulated environment. PATH–1’s and PATH–2’s

bottleneck has 1 Mbps bandwidth and 45 ms

propagation delay. Additionally, PATH–1 has 1%

Packet Loss Rate (PLR) while PATH–2 has variable

PLR which varies between 1%–10%. All the

simulation results presented are estimated by

normalizing the results over hundred runs, which

makes the consequence of the loss rate and cross-

traffic on simulated policies be more accurate and

not effected by any other stochastic factors.

Furthermore, all the necessary specific simulation

parameters are listed in Table.1 below.

Table 1. Configuration Parameters.

Parameters Values

MPTCP Maximum

Segment Size (MSS)

1500 Bytes

MPTCP Sender Buffer Size 64 KB

MPTCP Receiver Buffer

Size

64 KB

MPTCP Application File Transfer

Protocol (FTP)

SCTP MSS 1500 Bytes

SCTP Data Chunk Size 1468 Bytes

SCTP Sender Buffer Size 64 KB

SCTP Receiver Buffer Size 64 KB

SCTP RTX Policy RTX-CWND

SCTP Application FTP

Queuing Scheme Drop Tail

Queue Size 50 Packets

Bottleneck Bandwidth 1 Mbps

Paths’ Propagation delay 45 ms

Background Traffic UDP

Simulation Period 200 sec.

UDP Application PATH–1: 300

Kbps, PATH–2:

400 Kbps

PLR PATH–1: 1%,

PATH–2: 1%–10%

Figure 1: Topology

4 EXPERIMENTAL ANALYSIS

Fig. 2 exhibits the performance analysis in terms of

the average throughput (Kbps) when the Packet Loss

Rate (PLR) (%) increases. The objective of this

experiment is to validate the capability of all the

simulated CMT policy to deal with packet loss,

which has considerable influence on average

throughput performance. In fact, Fig. 2 shows the

performance in terms of average throughput of all

the simulated CMT. In particular, as the PLR

increases, it radically enhances the chances of higher

cwnd growth reductions. Moreover, it also increases

the probability of higher transmission delay as well.

In particular, we can clearly observe that there is

around 29.23%, 32.52% and 29.18% drop in

throughput of MPTCP, CMT and CMT-PF

respectively as the PLR varies between 1%–10%.

This is due to the fact that all these approaches

reduce their cwndimmediately as soon as they sense

the packet loss. However, CMT-PF suggests

improved performance because it can accurately

recognize packet drop as a result of short-term route

failures. Here, the results clearly signify that the

overall throughput performance of MPTCP is pretty

much lesser than that of other CMT approaches.

Specifically, MPTCP’s overall throughput

performance is 19.50% less than that of CMT-PF

and around 14.20% less than that of CMT. This is

due to the fact that both CMT and CMT-PF use

congestion control scheme independently for each

available sub-paths, and hence it leads to high

uncontrolled or aggressive cwnd growth behavior in

the network. Hence, it certainly assists CMT and

CMT-PF in terms of improved average throughput

performance. While, MPTCP uses CCC algorithm

which subsequently performs the congestion control

considering the status of each available sub-paths,

ICACSE 2021 - International Conference on Advanced Computing and Software Engineering

and hence it significantly controlled the

aggressiveness of cwnd growth for each sub-path in

the network. Hence, it leads to lesser average

throughput performance in the system. Nevertheless,

CMT and CMT-PF severely lacks of fairness against

non-CMT users, while MPTCP is far more fair

against single-path TCP users.

Figure 2: Average throughput (Kbps) performance of

simulated CMT schemes for varied PLRs.

Fig. 3 illustrates the comparative analysis of the

throughput (Kbps) when the bandwidth (Mbps)

increases. The purpose of this experiment is to verify

the competence of all the simulated CMT schemes

on increasing bandwidth values. Specifically, we

have simulated and analysed these results by

keeping PATH–1 bandwidth value constant, while,

we have varying the bandwidth values of PATH–2

in between 100 Kbps to 1.0 Mbps. Fig. 3 shows the

average throughput performance of all the simulated

CMT schemes continue to increase as bandwidth

values continue to increase. In particular, at 100

Kbps bandwidth, the average throughput

performance of MPTCP is around 22% more than

that of CMT and approximately 4% more than that

of CMT-PF. Also, MPTCP suggests comparable

performance to CMT-PF at 200 Kbps bandwidth,

while MPTCP offers 13% more average throughput

performance as that of CMT. Meanwhile, similar

average throughput performance has been observed

for MPTCP than that of CMT and CMT-PF at 300

Kbps bandwidth. Here, when the available

bandwidth of PATH–2 is limited (i.e., 100 Kbps to

300 Kbps); MPTCP’s CCC algorithm effectively

controls the cwnd growth aggressiveness on both

available network sub-paths, and hence MPTCP’s

performance is better than that of CMT and

comparable as that of CMT-PF. Since, CMT and

CMT-PF’s congestion control policy causes

highercwnd growth aggressiveness on both available

network sub-paths, and hence their policy assist in

sufficiently increasing the size of cwnd.

Consequently, their average throughput performance

is adequately better than that of MPTCP. Hence, on

limited bandwidth values these policies are more

likely to experience high packet losses and that

subsequently reduces their average throughput

performance. However, due to dependent congestion

control policy (i.e., CCC Algorithm) limits MPTCP

to aggressively utilize the available bandwidth, in

particular, it fails in sufficiently increasing the size

of cwnd, and hence MPTCP average throughput

performance is less than that of other simulated

policies. Specifically, MPTCP’s average throughput

performance is around 15% more than that of CMT

and comparable performance has been observed

with CMT-PF. While at higher bandwidth variations

(i.e., 400 Kbps to 1 Mbps), MPTCP’s average

throughput performance is 23.22% and 19.78% less

than that of CMT-PF and CMT respectively.

Figure 3: Average throughput (Kbps) performance of

simulated CMT schemes for varied Bandwidth (Mbps).

Fig. 4 shows the comparative analysis of the

throughput (Kbps) performance as the path delay

(ms) increase. The purpose of this experiment is to

verify the competence of all the simulated CMT

schemes on increasing path delay values.

Specifically, we have evaluated these results by

keeping PATH–1 delay constant, while, varying the

path delay of PATH–2 in between 10 ms to 100 ms.

Here, it has been observed that around 11% and

12.37% drop in average throughput performance of

CMT and CMT-PF respectively. Meanwhile, in case

of MPTCP, serious drop in average throughput

performance (i.e., around 54.45%) has been

observed. This effect is any increase in path delay

causes more OOO delivery at the end host (i.e.,

destination) which subsequently causes unnecessary

fast retransmissions and redundant cwnd reductions.

That ultimately leads to the issue of buffer-blocking

problem and reduces the average throughput

performance drastically. In particular, there is slight

drop in average throughput performance has been

observed for CMT and CMT-PF because their

aggressive cwndgrowth policy rapidly manages to

utilize the available channel bandwidth well on time.

Experimental Analysis of Concurrent Multi-path Transmission Schemes

However, MPTCP lacks in effectively achieving the

channel utilization due to its less aggressive

cwndgrowth policy, and hence its average

throughput performance gets seriously affected.

Figure 4. Average throughput (Kbps) performance of

simulated CMT schemes for varied Path Delay (ms).

5 CONCLUSIONS AND FUTURE

SCOPE

This paper evaluated and presented an analysis of

the CMT, CMT-PF and MPTCP’s performance over

varying path characteristics. And along with that, we

understood how dissimilar characteristics (i.e., PLR,

bandwidth and path delay) of multiple available

network paths made the difference to the old

schemes given. We revealed that old data scheduling

scheme (i.e., round-robin scheduling) and lesser

aggressive cwnd growth behavior significantly

affected MPTCP’s average throughput performance.

Since, both CMT and CMT-PF independently adapt

their cwndgrowth, hence, this aggressive cwnd

growth behavior assists both policies in effectively

utilizing channel bandwidth. Consequently, their

average throughput performance is significantly

more than that of MPTCP. Still, it is not rational to

imply that we must go for only CMT and CMT-PF

scheme but not for MPTCP. If we talk about fairness

in particular, the CMT and CMT-PF policy

significantly fails in achieving fairness to non-CMT

users, while, MPTCP performs considerably well by

achieving fairness to single-path TCP users.

The current suggested method makesutilization

of all the accessible sub-flows. Still, there may be

the possibility that better throughput performance

may be achieved by dynamically eradicating based

on their individual congestion. Furthermore,

methodical and accurate assessment of the traffic

scheduling and path management concern is

unquestionably required prior to widespread

deployment of these concurrent transmission polices

in actual Internet environment.

REFERENCES

A. Singh, M. Xiang, A. Konsgen, C. Goerg, and Y. Zaki.

(2013) “Enhancing fairness and congestion control in

multipath TCP.” In 6th Joint IFIP Wireless and Mobile

Networking Conference (WMNC), 1–8.

A.Ford, C.Raiciu, M.Handly, and O.Bonaventure. (2013)

“TCP extensions for multipath operation with multiple

addresses.” IETF RFC 6824 (Technical Report),

https://tools.ietf.org/html/rfc6824. Accessed on 12

October 2017.

A.Ford, C.Raiciu, M.Handly, S.Barre, and J.Iyengar.

(2011) “Architectural guidelines for multipath TCP

development.” IETF RFC 6182 (Technical Report),

https://tools.ietf.org/html/rfc6182. Accessed on 1

October 2016.

C.Paasch, and O.Bonaventure. (2014) “Multipath TCP.”

Communications of the ACM 57 (4): 51–57.

C.Paasch, S. Ferlin, O. Alay, and O.Bonaventure. (2014)

“Experimental evaluation of multipath TCP

schedulers.” In: Proceedings of the 2014 ACM

SIGCOMM workshop on Capacity sharing workshop,

27–32.

C.Raiciu, M.Handly, and D.Wischik. (2011) “Coupled

congestion control for multipath transport protocols.”

IETF RFC 6356, https://tools.ietf.org/html/rfc6356.

Accessed on 14 December 2018.

C.Xu, J.Zhao, and G-M.Muntean. (2016) “Congestion

Control Design for Multipath Transport Protocols: A

Survey.” IEEE Communications Surveys & Tutorials

18 (4): 2948–2969.

C.Xu, P.Wang, X.Wei, and G-M.Muntean. (2017)

“Pipeline network coding-based multipath data

transfer in heterogeneous wireless networks.” IEEE

Transactions on Broadcasting 63 (2): 376–390.

C.Xu, T.Liu, J.Guan, H.Zhang, and G.Muntean. (2013)

“CMT-QA: Quality-aware adaptive concurrent

multipath data transfer in heterogeneous wireless

networks.” IEEE Transactions on Mobile Computing

12 (11): 2193–2205.

C.Xu, Z.Li, J.Li, H.Zhang, and G-M.Muntean. (2015)

“Cross-layer Fairness-driven Concurrent Multipath

Video Delivery over Heterogeneous Wireless

Networks.” IEEE Transactions on Circuits and

Systems for Video Technology 25 (7): 1175–1189.

C.Xu, Z.Li, L.Zhong, H.Zhang, and G-M.Muntean. (2016)

“CMT-NC: Improving the Concurrent Multipath

Transfer Performance using network coding in

wireless networks.” IEEE Transactions Vehicular

Technology 65 (3): 1735–1751.

D. Kanellopoulos, V. K. Sharma. (2020) “Survey on

power-aware optimization solutions for MANETs.”

Electronics 9: 1129.

D. Wischik, C. Raiciu, A. Greenhalgh, and M. Handley.

(2011) “Design, implementation and evaluation of

congestion control for multipath TCP.” In: Proceedings

of the 8th USENIX conference on Networked systems

design and implementation, 99–112.

F. Kelly, and T. Voice. (2005) “Stability of end-to-end

algorithms for joint routing and rate control.” ACM

ICACSE 2021 - International Conference on Advanced Computing and Software Engineering

SIGCOMM Computer Communication Review 35 (2):

5–12.

F.Perotto, C.Casetti, and G.Galante. (2007) “SCTP-based

Transport Protocols for Concurrent Multipath

Transfer.” In IEEE Wireless Communications and

Networking Conference, 2969–2974.

H. Han, S. Shakkottai, C. V. Hollot, R. Srikant, and D.

Towsley. (2006) “Multi-Path TCP: A Joint Congestion

Control and Routing Scheme to Exploit Path Diversity

in the Internet.” IEEE/ACM Transactions on

Networking 14 (6): 1260–1271.

J. Postel (1980) “User datagram protocol.” IETF RFC 768,

https://tools.ietf.org/html/rfc768. Accessed on 15 May

2019.

J. Postel (1981) “Transmission control protocol.” IETF

RFC 793, https://tools.ietf.org/html/rfc793. Accessed

on 15 May 2019.

J.R.Iyengar, P.D.Amer, and R.Stewart. (2006) “Concurrent

multipath transfer using SCTP multihoming over

independent end-to-end paths.” IEEE/ACM

Transactions on Networking 14 (5): 951–964.

L.P.Verma, and M.Kumar. (2017) “An adaptive data chunk

scheduling for concurrent multipath transfer.”

Computer Standard and Interfaces 52: 97–104.

L.P.Verma, V.K.Sharma, and M.Kumar. (2018) “New

delay-based fast retransmission policy for CMT-

SCTP.” International Journal of Intelligent Systems

and Applications 10 (3): 59–66.

M. Becke, H. Adhari, E. P. Rathgeb, F. Fa, X. Yang, and

X. Zhou. (2013) “Comparison of Multipath TCP and

CMT-SCTP based on intercontinental measurements.”

In: IEEE Global Communications Conference

(GLOBECOM), 1360–1366.

M. Li, A. Lukyanenko, Z. Ou, A. Yla-Jaaski, S. Tarkoma,

M. Coudron, and S. Secci. (2016) “Multipath

Transmission for the Internet: A Survey,” IEEE

Communications Surveys & Tutorials 18 (4): 2887–

2925.

P.Natarajan, N.Ekiz, P.D.Amer, and R.Stewart. (2009)

“Concurrent multipath transfer during path failure.”

Computer Network 32 (15): 1577–1587.

P.Natarajan, N.Ekiz, P.D.Amer, J.R.Iyengar, and

R.Stewart. (2008) “Concurrent multipath transfer using

SCTP multihoming: Introducing the potentially-failed

destination state.” in Das A., Pung H.K., Lee F.B.S.,

Wong L.W.C. (eds) NETWORKING 2008 Ad Hoc and

Sensor Networks, Wireless Networks, Next Generation

Internet, 727–734.

Q. Peng, A. Walid, and S. H. Low. (2013) “Multipath TCP

algorithms: theory and design.” In: Proceedings of the

ACM SIGMETRICS/international conference on

Measurement and modeling of computer systems, 305–

316.

Q. Peng, A. Walid, J. Hwang, and S. H. Low. (2016)

“Multipath TCP: Analysis, Design, and

Implementation.” IEEE/ACM Transactions on

Networking 24 (1): 596–609.

R. Khalili, N. Gast, M. Popovic, and J-Y. Le Boudec.

(2013) “MPTCP is not Pareto-optimal: performance

issues and a possible solution.” IEEE/ACM

Transactions on Networking 21 (5): 1651–1665.

R.Stewart, Q.Xie, K.Morneault, C.Sharp, H.Schwarzbauer,

T.Taylor, I.Rytina, M.Kalla, L.Zhang, and V.Paxson.

(2000) “Stream control transmission protocol.” IETF

RFC 2960, https://tools.ietf.org/html/rfc2960. Accessed

on 26 January 2019.

S.Habib, J.Qadir, A.Ali, D.Habib, M.Li, and

A.Sathiaseelan. (2016) “The past, present, and future of

transport-layer multipath.” Journal of Network and

Computer Applications 75: 236–258.

T.Wallace, and A.Shami. (2012) “A Review of

Multihoming Issues Using the Stream Control

Transmission Protocol.” IEEE Communications

Surveys & Tutorials 14 (2): 565–578.

T.Wallace, and A.Shami. (2014) “Concurrent multipath

transfer using SCTP: Modelling and congestion

window management.” IEEE Transactions on Mobile

Computing 13 (11): 2510–2523.

V. K. Sharma, and M. Kumar. (2017) “Adaptive Energy

Efficient Load Distribution Using Fuzzy Approach.”

Ad Hoc & Sensor Wireless Networks 39 (1-4): 123–

166.

V. K. Sharma, and M. Kumar. (2018) “Adaptive load

distribution approach based on congestion control

scheme in ad-hoc networks.” International Journal of

Electronics 106 (1): 48–68.

V. K. Sharma, S. S. P. Shukla, and V. Singh. (2012) “A

Tailored Q-Learning for Routing in Wireless Sensor

Networks.” In: IEEE International Conference on

Parallel Distributed and Grid Computing (PDGC),

663–668.

V. K. Sharma. (2019) “Energy and congestion conscious

transmissions and routing in SANETs and MANETs: A

Survey.” Engineering and Technology Journal for

Research and Innovation 1(2): 38-42.

V.K.Sharma, and M.Kumar. (2017) “Adaptive congestion

control scheme in mobile ad-hoc networks.” Peer-to-

Peer Networking and Applications 10 (3): 633–657.

V.K.Sharma, L.P.Verma, and M.Kumar. (2018) “A Fuzzy-

based Adaptive Energy Efficient Load Distribution

Scheme in Ad-hoc Networks.” International Journal of

Intelligent Systems and Applications 10 (2): 72–84.

V.K.Sharma, L.P.Verma, and M.Kumar. (2019) “CL-

ADSP: Cross-layer adaptive data scheduling policy in

mobile ad-hoc networks.” Future Generation Computer

Systems 97: 530–563.

V.K.Sharma, L.P.Verma, M.Kumar, R. K. Naha, A.

Mahanti. (2020) “A-CAFDSP: An Adaptive-

congestion aware Fibonacci sequence based data

scheduling policy.” Computer Communications 158:

141–165.

W-H Wang, M. Palaniswami, and S. H. Low. (2003)

“Optimal flow control and routing in multi-path

networks.” Performance Evaluation 52 (2-3): 119–132.

Y. Nishida. (2013) “MPTCP implementation on NS-2.”

with IETF and WIDE Project,

http://code.google.com/p/multipath-tcp. Accessed on

11 May 2017.

Experimental Analysis of Concurrent Multi-path Transmission Schemes