Dacksis: An efficient transport protocol with acknowledgment-assisted storage management for intermittently connected wireless sensor networks

Abstract

The benefits of hop-by-hop transport protocols and in-network storage in intermittently connected networks are well known. However, due to the limitation on the storage capability of wireless sensor networks (WSNs), the hop-by-hop transport protocols based on in-network storage but without storage management are inappropriate to be used directly in intermittently connected WSNs. The absence of storage management leads to node storage overflow resulting in a degraded performance of the network.This paper introduces an efficient transport protocol, Dacksis, which provides end-to-end reliability, congestion control and storage management for intermittently connected wireless sensor networks. The design of Dacksis is based on hop-by-hop transport and in-network storage, therefore, the system can react to detected network congestion quickly and can reduce the number of end-to-end packet retransmissions. The storage management feature of Dacksis saves node storage for unconfirmed packets, which decreases the possibility of buffer overflow and packet queuing time, and reduces energy spent on loss recovery. Dacksis forgoes link layer per-packet acknowledgments and transmits packets in blocks to reduce energy consumption. The block size of Dacksis is dynamically determined before each hop-by-hop transmission to enhance the utilization of node storage.