A Review on role of Congestion Control Techniques in Internet of Things

Gomed Kumar Pathak¹, V.K. Patle²

^1,2SOS in Computer Science and IT, Pt. Ravishankar Shukla University Raipur, C.G.

¹gomed.pathak@gmail.com, ²patlevinod@gmail.com

Abstract:  Internet of Things has picked stream over industry, education and research, which provide a platform to connect various heterogeneous devices over internet. Devices may be sensor node related to environmental data collection, prediction, industry systems etc. Nowadays connection of various devices increases over the internet exponentially which causes congestion controls in Internet of Things. Congestion control plays a vital role in the transmission of large amount of data. In the era of Internet of things (IoT) which embellished our life still faces various challenges. Such as interoperatibility, security and congestion control.  In the networks working on Internet of things due to the sensitivity of traffics would become extremely complicated as there is a genuine risk of congestion collapse in the absence of adequate congestion control mechanisms. This paper presents a review of congestion control techniques, types of congestion control techniques, congestion avoidance, performance matrix and supporting tools in Internet of Things.

Keyword: Internet of Things (IoT), Congestion Control Techniques, Congestion Avoidance.

Introduction: -

The Internet of Things (IoT) refers to the fast expanding network of Internet-enabled physical objects found in homes, businesses, and "smart cities." Since these IoT devices have limited resources, an increase in network traffic could overwhelm the system [1]. When there is a lot of data being sent and received on an Internet of Things network, it slows everything down. Congestion develops for these reasons, in the long run [2]. One of the most crucial challenges is developing a framework for identifying and classifying network congestion, which has implications for both optimal bandwidth use and the routing path selection process. Consequently, it is crucial to categorise and predict congestion and take suitable preventative measures accordingly [3] to ensure efficient network communication. Thus, studies that classify and anticipate network congestion are crucial, especially for IoT networks. Congestion prediction in IoT networks often uses tried-and-true methods developed for WSNs (WSN). The development of high-powered computers and complex algorithms has allowed for the prediction and management of traffic flows; these developments are the topic of this study. In instance, measuring the traffic volume on an active Internet of Things network can be expensive. Due to the need for many types of network connectivity, the Internet of Things (IoT) is receiving a great deal of interest from researchers. The goal of the Internet of Things is to create a network of interconnected gadgets and services that can exchange and store data and information. There is no longer anyone who does not have some method of connecting to the internet these days. Throughout the last 20 years, researchers and businesses have been working on the concept of the Internet of Things[4][38]. Many other issues, such as self-organization, scalability, data quantities, power supply, data interpretation, wireless communications, and interoperability, are also being studied in relation to the Internet of Things[5]. The purpose of the Internet of Things is to facilitate normal social functioning. The Internet of Things encompasses a wide range of applications, from smart grids and cities to environmental and health care monitoring systems. CompTIA has predicted that by 2020[6], there will be as many as 50 billion internet-connected gadgets in use worldwide. The implications for computing and communication networks could be enormous. Most Internet-connected gadgets, however, aren't particularly powerful and have limited memory and power, among other drawbacks [7]. As a result, creating standardised protocols that allow for these capabilities is crucial. In Fig. 1, we see the IoT's underlying structure.

 Congestion Control

In an effort to keep the network from becoming overloaded, each TCP implements a set of behaviours known as congestion control, which are defined by algorithms. The Transmission Control Protocol (TCP), which employs a variety of different congestion control techniques, is used to carry packets in and of itself. The most common transport layer protocol used on the internet is TCP [8]. Depending on how they detect congestion, different categories of congestion control algorithms can be identified [9]. Loss-based algorithms are ones that detect congestion when packets are dropped from full buffers. Delay-based algorithms use round trip time 4 (RTT) measurements where the RTT is the amount of time taken for a packet to go from sender to receiver and for an acknowledgement to be sent back to the sender. A fluctuation in the RTT indicates delays due to increased buffering. Hybrid is the name given to a class of algorithms that exploit both the previous methods for congestion detection. Some Hybrid algorithms might classify their type of congestion control algorithm based on their unique congestion detection method. In [9], several new congestion control algorithms come out. They are known as TCP Slow Start, Congestion Avoidance, Fast Retransmit and Fast Recovery. TCP is widely deployed as the network's transport protocol [10]. We can classify congestion control algorithms into subclasses according on their methods of congestion detection [11]. Congestion is identified via loss-based algorithms, which discard packets from overflowing buffers. Round trip time 4 (RTT4) is the elapsed time between when a packet is transmitted and when the acknowledgement is received, and is used by delay-based algorithms to determine how long a delay actually is. There have been delays because of higher buffering if the RTT has fluctuated. One category of algorithms, called hybrids, combines the strengths of the two aforementioned congestion detecting strategies. Congestion control algorithms can be categorized by some Hybrid algorithms, depending on how they detect congestion. The new congestion control algorithms presented in [10] are extensive. TCP employs four such mechanisms, namely the Slow Start, Congestion Avoidance, Fast Retransmit, and Fast Recovery. To put it simply, Reno is no longer a potent TCP congestion control algorithm for modern high-speed networks like 10G Ethernet. There is no way that Reno can grow quickly enough to maximize its bandwidth usage. In geometry, a line with a cubic function has two distinct morphologies, one convex and one concave. In the absence of a loss event, this is the time it takes for the function to increase to a maximum value. Whether it's an ACK, a lost packet, or the end of the allotted time for retransmission, congestions are always signaled in the aforementioned two methods. Las Vegas used RTT to solve some of Reno's issues. In Reno, unless there is packet loss, the growth of will continue. Optimizing is measured differently in Vegas; throughput is tracked ( ). The effective bandwidth (ERE, eligible rate estimate) of the channel is estimated in Westwood in a manner similar to that used in Las Vegas, with a concentration on RTT. Westwood, on the other hand, calculates the bandwidth based on ACK arrivals over a range of different measurement intervals. [11]. WinSock is not publicly available since Windows is not open source.

Types of Congestion Control Algorithms

Congestion control majorly categorized as Protocol dependent Congestion control and Offloading based control congestion

Congestion Management Based on Protocols: The standard application protocol CoAP is used by the IoT's Congestion Control Protocol at the Application Layer.

Protocol for Controlling Congestion at the Network Layer Congestion can be managed with the aid of a load-aware routing protocol design.

Congestion Manage by Offloading Procedures The term "offloading" refers to the practise of passing the workload of a traffic node onto other nodes. This method helps manage congestion, lessens waiting time, and boosts efficiency. For the sake of traffic smoothing, gateways and other network infrastructure account for the in service delay by adding extra queues and processing time. To avoid communication breakdown and unnecessary packet loss in a crowded network, offloading requests has been proposed [12].

Congestion in IoT

The Internet of Things (IoT) is a worldwide infrastructure for connecting computers and other networked devices that can exchange and process data. The devices can be anything with its own identity (IP address) on the Internet, including desktop computers, tablets, smart phones, and more. Smart homes, smart cities, smart grids, industrial monitoring systems, healthcare monitoring systems, environmental monitoring systems, etc. are just a few examples of the many ways the Internet of Things can be put to use. Communication over the Internet must be regulated by a set of protocols in order for the IoT to be implemented. In order to facilitate communication between its many endpoints, IoT supports a wide variety of application protocols. Three examples of application protocols for the Internet of Things (IoT) that rely on TCP (Transmission Control Protocol) [13] for data transmission are XMPP (Extensible Messaging and Presence Protocol) [14], MQTT (MQ Telemetry Transport) [15], and RESTful HTTP [16].Transport layer control protocols can be split into two distinct camps. One subclass of  CA algorithms, called loss-based CA, relies on packet loss as a congestion signal. Author [38] proposed an Adaptive Congestion Window (ACW) for IoT devices. The design of the ACW depends on three parameters: sending rate, receiving rate and the available bandwidth of the path. Author[39] proposes a novel STCP approach to control congestion in the IoT environment. In this approach, a new window initialization technique is used based on the current available bandwidth of the path in order to reach the available bandwidth as fast as possible. It is limited to very small buffers, less than or equal to one maximum segment size. The authors [40]propose a new technique to reduce the delay in multipath TCP (MPTCP) by reducing the number of transmissions using an Opportunistic Routing (OR) technique. The OR routing model is implemented to increase the throughput and reliability of wireless networks via the use of the broadcasting method.

Literature Review

Jacobson [17] suggested a loss-based congestion control method for the TCP protocol; the Fast Retransmit and Fast Recovery approach offered by this algorithm is known as TCP Reno. When more than one packet is dropped owing to congestion, Rapid Recovery abruptly transforms into exponential congestion window lowering, leaving TCP Reno vulnerable. Recognizing this issue, Floyd et al. [18, 19] presented a new approach they termed TCP New Reno to fix the Quick Recovery issue. Mathis et al. [20] suggest a technique called TCP SACK to deal with numerous losses. The receiver can now report the total number of data packets that were successfully transmitted thanks to this protocol. Another SACK-based solution with a novel congestion control mechanism was presented by Mathis and Mahdavi [21]. There are three state variables that FACK (forward acknowledgement) keeps track of: H (highest sequence number), F (most sequences forwarded), and R (most sequences resent). High Speed Transmission Control Protocol (HS TCP) was introduced by Floyd [22] to address the efficiency issue of high-speed networks. The HS-TCP employs a factor of drop for small loss detection and a factor of increase for congestion avoidance. Kelly [23] suggests a different protocol named Scalable TCP (STCP) to solve the effectiveness problem in high speed long delay network. To solve the inter fairness issue between STCP and HS-TCP, Leith et al. [24] suggested a new congestion control technique they dubbed HTCP. The primary idea behind HTCP is to expand the size of the congestion window in the Congestion Avoidance phase by an amount equal to the elapsed time. TCP-CUBIC, an improved variant of BIC TCP [25], was proposed for congestion control by Rhee and Xu [26]. To address the issues with TCP-CUBIC, Wang et al. [27] introduced CUBIC-FIT. The TCP-CUBIC framework is expanded using a delay-based TCP model. A delay-based congestion management technique, TCP Vegas, was introduced by Brakmo and Peterson [28]. The method relies on the difference between the predicted and actual rates to arrive at an accurate prediction

Performance Matrix:

The performance of traffic in a congested environment is measured using a variety of metrics. These variables assist us in assessing network performance during periods of congestion.

Availability of Network: It calculates the energy and resources wasted on packets that are not delivered. Network efficiency varies according to the distance to the washbasin.

Efficiency in Energy (EE): The packet delivery ratio can be used to define EE.

Energy Tax (ET): The ratio of packets that successfully reach the washbasin to those are lost in the network.

Packet Loss Ratio or Delivery Ratio: Some packets are lost due to buffer overflow and packet errors; this determines the packet loss ratio.

Fairness: The topic is bandwidth distribution. It displays the variances (changes) in transmitting rates. It is reliant on equitable bandwidth distribution.

End-to-End Delay: End-to-end delay is another helpful evaluation metric for gauging how busy our network is. End-to-end delay refers to the overall amount of time that has passed since a packet was created before arriving at the base station.

Overhead for Control Packets: The expense of sending such packets that include the control information due to protocol restrictions is a good indicator of control packet overhead.

The number of packets that are successfully received by the sink node overall over a period of time.

Instantaneous Queue Size: It shows the consistency or variability of the buffer. Also used are event-based weighted queues.

Memory Requirements: The memory requirements were determined by the size of the queue, the amount of code, and the quantity of sensing units.

The Fidelity Index measures the proportion of packets that the application is intended to receive to those that are actually received [29].

Packet Loss: The average rate of packet loss provides information about how busy the network is.

Queue Length: A useful indication for congestion detection is buffer length. When the length of the buffer surpasses the limit, a predefined limit threshold is used, and the congestion signal is then activated.

Hybrid Queue Length and Channel Load: The ratio of time quantum's while a channel is being used to the total time is known as channel load.

Throughput: The number of successful transmissions on a channel is indicated by its throughput. In reality, it shows how busy the channel is still. Congestion detection is a possible application.

Packet Service Time: The period of time between the arrival of a packet and its transmission accounts for the period of time needed to resolve a collision.

Scheduling Time: Scheduling Time reveals the quantity of scheduled packets.

Delay: Delays can be used to identify congestion as well [29].

Tools Related to Network Research

The goal is to describe in detail the many resources available for studying and creating new computer networks. Protocol experiments can be run in a simulation at a fraction of the cost and time. Ad hoc models and logic event driven methodologies are frequently used in network simulation. The fundamental simulation engine and a wide variety of protocol models are provided by industry standards tools like NS-3, NS-2, and OPNET. A packet switch data network's dynamic behavior can be studied with REAL, a simulator designed specifically for that purpose [30].

Developed to put different IP-over-ATM algorithms through their paces using actual traffic loads generated from empirical traffic measurements, INSANE is a network simulator. The ATM protocol stack employs Rate Controlled Static Priority (RCSP) queuing to ensure reliable operation of ATM virtual circuits. NetSim's goal is to provide a comprehensive simulation of Ethernet, including all of its complex features including collision detection and handling, transmission deferral, and the impact of station locations on network events. Maisie is a parallel discrete event simulation language written in C [31, 32] that can simulate hierarchical structures. U-Net (T. Von. Eicken, et al. 1995), the USC TCP-Vegas test-bed (J. s.Ahn, et al. 1995), and the Harvard simulator are a few more (S. Y.Wang, H.T.Kung 1999). BONeS (Cadence Inc.), COMNET III (CACI), and OPNET are all commercial simulators (MIL3).

COMNET III is a graphical, commercially available software programme that can assess and estimate the performance of any network, from a small local area network (LAN) to a large, complicated enterprise-wide system (CACI)[33].

The OPNET Optimal Network Engineering Tool (OPNET) is an all-purpose network simulator that features discrete events and an object-oriented design. For the purpose of specifying, simulating, and analysing the performance of computer and data transmission networks, it offers a full-fledged development environment.

QualNet is a for-profit implementation of the free and open-source GloMoSim simulator. QualNet's ability to scale to thousands of nodes and function on different hardware and operating systems is arguably its greatest strength[34].

Ns-2 is one of the most used network simulation tools available. It was created at the University of California, Berkeley's Lawrence Berkeley Laboratory as an object-oriented discrete-event network simulator. Ns-2 is a command and configuration interface that is written in C++ and leverages OTcl. All aspects of the simulation environment, from initialization to scenario creation to experiment execution, are managed using OTcl scripts [35].Discrete-time process simulation is the focus of J-SIM, an object-oriented library. It is most useful for simulating queueing networks [36].Discrete event simulation might be complicated, but OMNeT++ makes it easier than ever The simulator is built out of smaller pieces, or modules. In OMNeT++, you can utilise either basic modules or compound modules. A computer network's behavior can be simulated with software like NS3, a network simulator. In addition, the client can display the network's topology in order to single out the system's hubs and trace the connections between them. In addition, it offers a narrated packet-flow animation. It's run in C++ with an optional Python scripting API for customization. It's an open-source project that's free to use and structured around research groups that work on improving and maintaining it[37].

Conclusion:

Due to the sensitivity of traffic, networks working on the Internet of Things would become extremely complicated, with a genuine risk of congestion collapse in the absence of adequate congestion control mechanisms. In this paper we have studied various congestion control techniques, types of congestion control techniques, congestion avoidance, performance matrix and supporting tools in Internet of Things.

References

1.     Thompson, W.L. & Talley, M.F. Deep learning for iot communications. 53rd Annual Conference on Information Sciences and Systems (CISS), 2019, pp. 1-4. doi: 10.1109/CISS.2019.8693025

2.     Al-Kashoash, Hayder & Ahmed,Abdulmohsin. Congestion control for 6LoWPAN wireless sensor networks: Toward the internet of things. University of Leeds, 2017. PhD Thesis. doi: 10.1007/978-3-030-17732-4

3.     Wang, M.; Cui, Y.; Wang, X.; Xiao, S. & Jiang J. Machine learning for networking: Workflow, advances and opportunities. IEEE Network, 2017, 32(2), 92-99. doi: 10.1109/MNET.2017.1700200

4.     O. Salman, I. H. Elhajj, A. Chehab, and A. Kayssi, “QoS guarantee over hybrid SDN/non-SDN networks,” in 2017 8th International Conference on the Network of the Future (NOF), pp. 141–143, 2017.

5.     A. Maheshwari and R. K. Yadav, "Analysis of Congestion Control Mechanism for IOT," 2020 10th International Conference on Cloud Computing, Data Science & Engineering (Confluence), Noida, India, 2020, pp. 288-293, doi: 10.1109/Confluence47617.2020.9058058.

6.     I. CompTIA, “Sizing up the internet of things”, (2015)[EB/OL], https://www.comptia.org/resources/sizing-up-the-internet-of-things

7.     J. Ren, Y. Zhang, N. Zhang, D. Zhang, X. Shen (2016) “Dynamic channel access to improve energy efficiency in cognitive radio sensor networks”, IEEE Transactions on Wireless Communications, 15 (5) 3143–3156.

8.     DongJin Lee, Brian E Carpenter, and Nevil Brownlee. Media streaming observations: Trends in udp to tcp ratio. International Journal on Advances in Systems and Measurements, 3(3-4), 2010.

9.     Belma Turkovic, Fernando A. Kuipers, and Steve Uhlig. Fifty Shades of Congestion Control: A Performance and Interactions Evaluation. arXiv:1903.03852 [cs], March 2019. URL http://arxiv.org/abs/1903.03852. arXiv: 1903.03852.

10.  Stevens W.: RFC 2001 - TCP Slow Start, Congestion Avoidance, Fast Retransmit, and Fast Recovery Algorithms

11.  Parziale, Lydia, David T. Britt, Chuck Davis, Jason Forrester, Wei Liu, Carolyn Matthews, and Nicolas Rosselot. TCP/IP Tutorial and Technical Overview. IBM International Technical Support Organization, 2006.

12.  Belma Turkovic, Fernando A. Kuipers, and Steve Uhlig. Fifty Shades of Congestion Control: A Performance and Interactions Evaluation. arXiv:1903.03852 [cs], March 2019. URL http://arxiv.org/abs/1903.03852. arXiv: 1903.03852.

13.  H. Byun and J. Lin, (2005) “On a fair congestion control scheme for TCP Vegas,” IEEE Communication Letters 9(2):102:105.

14.  Saint-Andre P (2011) “Extensible Messaging and Presence Protocol (XMPP): Core”. https://tools.ietf.org/html/rfc6120.

15.  MQTT Version 3.1.1 (2015) http://docs.oasis-en.org/mqtt/mqtt/v3.1.1/errata01/os/mqtt-v3.1.1-errata01-os-complete.doc.

16.  R.Fielding (2000) RESTful HTTP. http://www.infoq.com/articles/designing-restful-http-apps-roth.

17.  V Jacobson (1988) “Congestion avoidance and control”, ACM SIGCOMM Computer Communication Review, 18(4):14-329.

18.  M Allman, V Paxson, W Stevens,(1999) “TCP congestion control”. https://tools.ietf.org/html/rfc2581.

19.  S Floyd, T Henderson, (1999) “The NewReno modification to TCP’s fast recovery algorithm”. https://tools.ietf.org/html/rfc2582.

20.  S Floyd, T Henderson, A Gurtov, (2004) “The NewReno modification to TCP’s fast recovery algorithm”. https://tools.ietf.org/html/rfc3782.

21.  M Mathis , J Mahdavi ,S Floyd ,A Romanov, (1996) “TCP selective acknowledgment options”. https://tools.ietf.org/html/rfc2018.

22.  T Kelly (2003) “Scalable TCP: improving performance in high-speed wide area networks”. ACM SIGCOMM Computer Communication Review, 33: 83 – 91.

23.  D Leith , R Shorten (2007) “H-TCP: TCP congestion control for high bandwidth-delay product paths”. https://tools.ietf.org/html/draft-leith-

24.  S Ha, I Rhee, L Xu (2008) “CUBIC: a new TCP-friendly high-speed TCP variant”. ACM SIGOPS Operating Systems Review, 42:64–74

25.  ]L Xu, K Harfoush, I Rhee, (2004) “Binary increase congestion control for fast, long distance networks”. Twenty-third Annual Joint Conference of the IEEE Computer and Communications Societies, 2514–2524X.

26.  J Wang, J Wen, Y Han, J Zhang, C Li, Z Xiong (2013) “CUBIC-FIT: A High Performance and TCP CUBIC Friendly Congestion ControlAlgorithm”. IEEE Communication Latter, 17:1664-1667.

27.  L Brakmo, L Peterson (1995) “TCP Vegas: end to end congestion avoidance on a global Internet”. IEEE Journal on Selected Areas in Communications, 13:1465–1480.

28.  G Hasegawa, K Kurata, M Murata (2000) “Analysis and improvement of fairness between TCP Reno and Vegas for deployment of TCP Vegas to the Internet”. International Conference on Network Protocols, 177–186.

29.  Muhammad A. S. , Muhammad S. A., Shakeel W, Sanam B. , Usman S.,” Congestion Control in IoT” IJCSNS International Journal of Computer Science and Network Security, VOL.19 No.2, February 2019 pp 56-61

30.  S.Keshav, REAL 5.0 Overview, Cornell University, Available as:http://www.cs.cornell.edu/skeshav/real

31.  INSANE, An Internet Simulated ATM Networking Environment, Available as: http://www.ca.sandia.gov/~bmah/Software/Insane Lewis, Barnett, “An Ethernet Performance Simulator for Undergraduate Networking”, Proceeding of ACM SIGCSE Technical Symposium, 1993

32.  Lewis, Barnett, “An Ethernet Performance Simulator for Undergraduate Networking”, Proceeding of ACM SIGCSE Technical Symposium, 1993

33.  CACI Products company, COMNET III, Available at: http://www.caciasl.com/COMNET_quick_look.htm

34.  N. Sarkar and S. Halim, “Simulation of computer networks” Simulators, methodologies and recommendations. 2008.

35.  Dr. Vijay R. Ghorpade, “Comparative Study of Network Simulator: NS2 and NS3,”International Journal of Advanced Research in Computer Science and SoftwareEngineering, Vol 6, March 2016.

36.  J. A. Miller, R. S. Nair, Z. Zhang and H. Zhao, "JSIM: A Java-based simulation and animation environment," Proceedings of 1997 SCS Simulation Multiconference, Atlanta, GA, USA, 1997, pp. 31-42, doi: 10.1109/SIMSYM.1997.586473.

37.  Girish P. , Swapnesh T., N.S. Yadav,” A Design and Implementation Framework of Congestion Control Algorithm Module in the Network Simulator 3 (NS-3)”, International Journal of Computing and Applications ISSN: 0973-5704 Volume 13, Number 2, (July-December 2018), pp 371-379

38.  C. Cammin et al., "Concept for a Real-Time IoT-Architecture for Collision Avoidance in Smart Cities based on the 5G Mobile Technology," 2023 IEEE International Conference on Omni-layer Intelligent Systems (COINS), Berlin, Germany, 2023, pp. 1-7, doi: 10.1109/COINS57856.2023.10189240.
keywords: {Pedestrians;5G mobile communication;Transmitters;Smart cities;Roads;Real-time systems;Internet of Things;5G;IoT;Traffic;Smart City;Safety;Collision Avoidance;Cloud Architecture},

39.   Chappala, R.; Anuradha, C.; Murthy, P.S.R.C. Adaptive Congestion Window Algorithm for the Internet of Things Enabled Networks. Int. J. Adv. Comput. Sci. Appl. 2021, 12, 105–111. [CrossRef]

40.   Gupta, A.K.; Singh, D.; Singh, K.; Verma, L.P. STCP: A Novel Approach for Congestion Control in IoT Environment. J. Inf. Technol. Manag. 2022, 14, 44–51

41.  Aljubayri, M.; Peng, T.; Shikh-Bahaei, M. Reduce delay of multipath TCP in IoT networks. Wirel. Netw. 2021, 27, 4189–4198