Computer Science and Engineering MS Thesis Defense by Waleed Bin Qaim






Title:   Data Replication and Collection Framework for Enhanced Data Availability in

IoT-based Sensor Systems


Speaker: Waleed Bin Qaim


Time: September 10, 2018, 11:00 AM


Place: ENG208

Koç University

Rumeli Feneri Yolu

Sariyer, Istanbul


Thesis Committee Members:

Assoc. Prof. Öznur Özkasap (Advisor, Koç University)

Asst. Prof. Didem Unat (Koç University)

Assoc. Prof. Özlem Durmaz İncel (Galatasaray University)



Internet of Things (IoT) is an emerging technology aimed at bridging the gap between physical and digital worlds. IoT envisions to provide a seamless and efficient way of interacting with real-world physical objects through the Internet which highly impacts modern day life. Wireless Sensor Networks (WSNs) are primarily responsible for embedding intelligence and turning physical objects in IoT-enabled communicating entities that can be remotely accessed and controlled, thus, providing the basis for smart solutions to all the human needs. To realize the concept of IoT, efficient data generation, collection, and presentation through WSNs are crucial issues. However, WSNs face numerous challenges mainly because of the resource-constrained and unreliable nature of nodes which may result in loss of valuable sensed data. Data replication is a promising technique to facilitate efficient data management in IoT-based sensor systems. This thesis initially proposes a taxonomy of state-of-the-art data replication protocols for IoT-based sensor systems highlighting their pros and cons by classifying them into sub-categories of data retrieval, query balancing, system robustness, and data availability. Then, we propose a fully distributed hop-by-hop data replication technique for IoT-based sensor systems to avoid data loss due to node failures and enhance data availability in the network. Additionally, the proposed data replication scheme is coupled with an efficient data collection mechanism where a mobile sink node visits a relatively smaller percentage of network nodes to collect most of the network data. Extensive simulation experiments using network simulator (NS-3) show that the proposed data replication and collection framework ensures high data availability in the presence of node failures as well as provides maximum data collection efficiency. Comparative simulation results reveal that in comparison to two other state-of-the-art approaches, namely, greedy and random replication techniques, the proposed framework improves data availability with maximum gains of about 15% and 34%, respectively. Similarly, it improves average replicas created in the network with maximum gains of about 18% and 40%, respectively. Furthermore, the proposed solution ensures a better replica spread which determines the quality of data dissemination in the network as well as facilitates efficient data collection.