Author : Safae Bourhnane
Publisher :
Page : 276 pages
File Size : 15,30 MB
Release : 2017
Category : Cloud computing
ISBN :
Smart Grid (SG) is emerging as a promising technology leveraging energy-efficiency. In SG, electricity producers and consumers are equipped with sensors disseminating data in a real-time mode. These data fall within the realm of Big Data. To process SG Big Data, HPC (High-Performance-Computing) turns out indispensable. This thesis is two-fold: In the first part, we investigate Big Data acquisition via the deployment of a wireless sensor/actuator network that can be further optimized for real-world scalability. In the second part, we investigate the pros and cons of using Containers vs. VMs (Virtual Machines) for Big Data processing. Indeed, as Cloud Computing (CC) is getting more and more popular, it came clearer that virtualization is the key technology enabler behind it. Cloud services are usually provided via the instantiation of virtual machines (VMs). However, the latter have proven to consume a considerable amount of resources, namely memory. Also, the process of instantiating VMs is somehow heavy and usually takes more time which is of a direct impact on the overall performance of the system. As a promising alternative, containers are emerging to take over VMs as they consume significantly less resources, thus boosting system performance. Still, as any other new technology, containers introduce new challenges. In this thesis, we highlight the tradeoffs that one has to deal with when considering VMs and Containers. For HPC as a Cloud Service, we introduce a new hybrid approach that combines both containers and VMs. We led appropriate experiments on a real-world Cloud testbed, using Openstack, as a proof-of-concept.