Improving Congestion Control of TCP for Constrained IoT Networks.

Area of Science:

• Computer Science
• Network Engineering
• Internet of Things

Background:

• TCP is suitable for IoT integration but faces challenges in constrained networks.
• The uIP TCP/IP stack uses a small window size, making retransmission timer management crucial for congestion control.
• Existing uIP TCP retransmission mechanisms struggle with variable Round-Trip Times (RTT) caused by radio duty cycling (e.g., ContikiMAC).

Purpose of the Study:

• To investigate TCP congestion control performance in uIP stacks within grid topology networks.
• To identify the causes of excessive retransmissions in uIP TCP when using radio duty cycling mechanisms.
• To propose an improved retransmission timer management scheme for enhanced TCP performance in IoT.

Main Methods:

• Examined TCP congestion control in the uIP stack using grid topology networks and the Cooja simulator.
• Analyzed the impact of ContikiMAC radio duty cycling on TCP retransmissions.
• Developed and simulated a novel retransmission timer management scheme incorporating weak RTT estimation, exponential backoffs with variable limits, and dithering.

Main Results:

• The original uIP TCP implementation results in numerous retransmissions when using radio duty cycling like ContikiMAC.
• The fixed Retransmission Timeout (RTO) in uIP TCP is inadequate for handling RTT variations caused by mechanisms like ContikiMAC.
• The proposed scheme significantly reduces retransmissions, improves throughput, and enhances fairness in conjunction with RDC mechanisms.

Conclusions:

• The proposed retransmission timer management scheme effectively addresses the limitations of fixed RTO in uIP TCP under variable RTT conditions.
• This enhancement is vital for optimizing TCP performance in resource-constrained IoT networks employing radio duty cycling.
• The findings contribute to more reliable and efficient data transmission in IoT environments.