IoTP: on supporting IoT data aggregation through programmable data planes

• Autor(a) principal: Madureira, André Luiz Romano
• Data de Publicação: 2020
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Institucional da UFBA
• Texto Completo: http://repositorio.ufba.br/ri/handle/ri/33626
• topic: Ciências Exatas e da Terra; Ciência da Computação; Programming protocol-independent packet processors; Programmable data plane; Data aggregation; Internet of things; Protocol for programmable data planes; Internet of Things Protocol (IoTP)

IoT devices generate large continuous data streams, which causes congestion that compromises the scalability of IoT systems. To face this problem, techniques for data aggregation propose to reduce recurring packet headers, through assembly of packet data coming from different sources. Due to the energy constraints and limitation of computational resources of devices, most proposals adjust data aggregation according to their features following multilayered-based approaches or coupling the solution to a given network protocol, but overlooking the properties of the communication link. In this work, we introduce the Internet of Things Protocol (IoTP). An L2 communication protocol for IoT programmable data planes that supports the implementation of data aggregation algorithms inside hardware switches, at the network level. Through these features, IoTP provides support for the design of efficient and adaptable aggregation schemes that can function according to network status and based on the different communication tecnologies used by IoT devices. We implemented IoTP using the P4 language and conducted emulation-based experiments through the Mininet environment. Our findings show that IoTP accomplishes a 78% improvement in network efficiency, as well as allowing control over the average delay generated by data aggregation techniques. Besides that, it was able to reduce the number of packets sent over the network, while also reducing the consumption of network devices computational resources.