Reduction of Transaction Failures in a Constrained Distributed Payment Processing System in East Africa Through Machine Learning and a Federated Timeout Interval Negotiation Protocol

The use of web services on the Internet has become ubiquitous, and as a result, the challenge of providing a satisfactory quality of experience for users has increased. However, the same transport protocol design has been employed over the past three decades, with the representational state transfer (REST), hypertext transfer protocol (HTTP), and running on top of transmission control protocol (TCP). As a result, the traditional congestion control mechanisms such as TCP New-Reno, despite efforts to optimize TCP performance by enhancing the core network congestion control methods, do not apply to generic congestion issues for a mixed network architecture. The constraint results from the rule-based strategy employed, where efficiency is linked to pre-set mapping among the observed state of the network and the pertaining activities. The challenges are exacerbated in Africa due to the constrained network environment throughout the continent. Constrained devices usually come with microcontrollers that are 8-bit with a limited amount of both random access and read-only memory, while networks with a constrained environment such as internet protocol v6 operating on low-power wireless individual area networks often exhibit high packet failure rates and throughput of 10 kbit/s. As a result, there is a significant reduction in the probability of packet delivery. Consequently, the constrained network environment suffers a delay in transmission, resulting in timeout failures, since clients will not wait indefinitely for packets to arrive. Over the past decade, Ethiopia has seen a significant shift from traditional face-to-face banking to mobile/online banking, mainly due to the growth of mobile phone usage and ease of use (Jerene & Sharma, 2019). However, Ethiopia has limited internet speed, lightweight devices, and scarce electricity. Therefore, the high volume of intrabank mobile/online transactions coupled with the constrained network environment has reduced the quality of experience and quality of service. In this research, machine learning was used to design a congestion control protocol called timeout interval negotiation protocol (TINP), an extension of the message communication protocol, to reduce transaction processing failures in constrained distributed payment processing technologies. TINP automatically identifies the optimal timeout interval strategy, given the observation of the observed environment, in a real-time manner. It enables clients to adjust the pre-set timeout intervals based on just-in-time feedback. The protocol utilizes this capability to let clients learn various strategies to better acclimate to varying networking behaviors instead of automatically following static timeout intervals. The framework comprises three components: 1) a machine learning model to predict network congestion; 2) an algorithm to identify the optimal timeout interval; and 3) bidirectional communication between the client and server to federate and negotiate the optimal timeout interval in real time

Author

• Merdassa, Naoll Addisu

Language

• en

Keyword

• Congestion
• HTTP
• Constrained
• Webservice
• Machine Learning
• timeout

Date created

• 2023

Type of Work

• Dissertation

License

• All rights reserved

Rights statement

• In Copyright

GW Unit

• Engineering Management

Degree

• D.Engr.

Advisor

• Etemadi, Amir
• Widener, Donald V.

Committee Member(s)

• Widener, Donald V.
• Yassan, Rebecca
• Etemadi, Amir
• Dimitriou, Ari

Persistent URL

•  https://scholarspace.library.gwu.edu/etd/fb4949256