Geometric Series based effective RTO estimation Technique for CoCoA

Abstract

Constrained Application Protocol (CoAP) is a standard data transfer protocol for Internet of Things. It has an in-built support for the basic congestion control mechanism that uses fixed Retransmission Time Out (RTO) for every transmission regardless of Round Trip Time (RTT), and performs Binary Exponential Backoff (BEB) when the packets get dropped. CoAP Simple Congestion Control/Advanced (CoCoA) is an enhanced congestion control mechanism over CoAP that adapts RTO based on RTT. It maintains Strong and Weak RTO estimators and uses a Variable Backoff Factor (VBF) instead of BEB when the packets get dropped. CoCoA uses an Exponential Weighted Moving Average (EWMA) to estimate the RTO for the next transmission. The weight used in EWMA is determined on the basis of whether the RTT estimated for the recent transmission was a Strong RTT or Weak RTT. However, the weights used to estimate the RTO are fixed (0.5 for Strong and 0.25 for Weak). These fixed weights lead to slow adaptation of RTO and affect the performance of the IoT applications. In this paper, we highlight the impact of having fixed weights while estimating the RTO in CoCoA. In particular, we show that the RTO in CoCoA fails to adapt quickly when the network conditions are lossless because it uses a fixed value for Strong RTO estimation (0.5). We propose a new algorithm called Geometric Series based effective RTO estimation Technique for CoCoA (GSRTC) to adapt the weight used in EWMA for estimating Strong RTO. GSRTC is integrated into CoCoA and validated against existing mechanisms using the Cooja simulator in Contiki OS and in a real testbed FIT/IoT-LAB. Our results show that GSRTC has lower Flow Completion Times (FCT), lesser retransmissions and better network throughput.