Mechanisms for service-oriented resource allocation in IoT

• Autor(a) principal: Souza, Vitor Barbosa
• Data de Publicação: 2018
• Tipo de documento: Tese
• Idioma: eng
• Título da fonte: LOCUS Repositório Institucional da UFV
• Texto Completo: http://www.locus.ufv.br/handle/123456789/19074
• topic: service-oriented; Mechanisms; Arquitetura de Sistemas de Computação; Ciência da Computação

Albeit several IoT applications have been recently deployed in several fields, including environment and industry monitoring, Smart Home, Smart Hospital and Smart Agriculture, current deployments are mostly host-oriented, which is undoubtedly limiting the attained benefits brought up by IoT. Indeed, future IoT applications shall benefit from service - oriented communications, where the communication establishment between end-points is not dependent on prior knowledge of the host devices in charge of providing the service execution. Rather, an end-user service execution request is mapped into the most suitable resources able to provide the requested service. Furthermore, this model is a key enabler for the design of future services in Smart Cities, e-Health, Intelligent Transportation Systems, among other smart scenarios. Recognized the benefits of this model in future applications, considerable research effort must be devoted for addressing several challenges yet unsolved, such as the ones brought up by the high dynamicity and heterogeneity inherent to these scenarios. In fact, service-oriented communication requires an updated view of available resources, mapping service requests into the most suitable resources taking several constraints and requirements into account, resilience provisioning, QoS-aware service allocation, just to name a few. This thesis aims at proposing and evaluating mechanisms for efficient resource allocation in service-oriented IoT scenarios through the employment of two distinct baseline technologies. In the first approach, the so-called Path Computation Element (PCE), designed to decouple the host-oriented routing function from GMPLS switches in a centralized element, is extended to the service-oriented PCE (S-PCE) architecture, where a service identifier (SID) is used to identify the service required by an end-user. In this approach, the service request is mapped to one or a set of resources by a 2-steps mapping scheme that enables both selection of suitable resources according to request and resources characteristics, and avoidance of service disruption due to possible changes on resources’ location. In the meantime, the inception of fog computing, as an extension of the cloud computing concept, leveraging idle computing resources at the edge of the network through their organization as highly virtualized micro data centers (MDC) enabled the reduction on the network latency observed by services launched at edge devices, further reducing the traffic at the core network and the energy consumption by network and cloud data center equipment, besides other benefits. Envisioning the benefits of the distributed and coordinated employment of both fog and cloud resources, the Fog-to-Cloud (F2C) architecture has been recently proposed, further empowering the distributed allocation of services into the most suitable resources, be it in cloud, fog or both. Since future IoT applications shall present strict demands that may be satisfied through a combined fog-cloud solution, aligned to the F2C architecture, the second approach for the service-oriented resource allocation, considered in this thesis, aims at providing QoS -aware resource allocation through the deployment of a hierarchical F2C topology, where resource are logically distributed into layers providing distinct characteristics in terms of network latency, disruption probability, IT power, etc. Therefore, distinct strategies for service distribution in F2C architectures, taking into consideration features such as service transmission delay, energy consumption and network load. Concerning the need for failure recovery mechanisms, distinct demands of heterogeneous services are considered in order to assess distinct strategies for allocation of protection resources in the F2C hierarchy. In addition, the impact of the layered control topology on the efficient allocation of resources in F2C is further evaluated. Finally, avenues for future work are presented.