TCP PERFORMANCE IN WIRELESS NETWORKS

Abstract: Transmission Control Protocol $($TCP$)$ enables end$-$to$-$end data delivery with reliability;

therefore, it is the most popular transport protocol. On the other hand, in emerging communication

networks like wireless networks, Internet of Things, $5$G$/6$G$,$ software$-$defined networking $($SDN$),$

edge computing, and multimedia applications with strict quality of service $($QoS$)$ demands, TCP is

vulnerable to performance degradation. The fundamental issue is TCP$$s inability to efficiently handle

these losses, bandwidth fluctuations, handoffs, asymmetric links, high bit error rates, user mobility,

shared medium access and delay variability hence it is crucial to fine tune TCP performance with

respect to high bandwidth$-$delay products, wireless impediments, dynamic requirements, and

complex topologies. Our objective would involve efficient and fair bandwidth sharing, minimizing

latency and jitter for multimedia, load balancing in cloud data centers, and strict QoS guarantees.

We aim to examine TCP variants in detail, and subsequently gauge their behaviour in different

network conditions as well as the key problems and challenges.

We explore potential solutions by leveraging techniques like multipath TCP and machine learning.

Multipath TCP enables simultaneous data transfer over multiple subflows, improving throughput

and resilience to failures or variations. Machine learning approaches can predict achievable rates,

latency, and QoS over different network paths, enabling proactive scheduling and adaptation of

TCP$\text{'}$s behaviour.