Please use this identifier to cite or link to this item:
|Title:||Automating Computational Placement for the Internet of Things|
|Abstract:||The PATH2iot platform presents a new approach to distributed data analytics for Internet of Things applications. It automatically partitions and deploys stream-processing computations over the available infrastructure (e.g. sensors, field gateways, clouds and the networks that connect them) so as to meet non-functional requirements including network limitations and energy. To enable this, the user gives a high-level declarative description of the computation as a set of Event Processing Language queries. These are compiled, optimised, and partitioned to meet the non-functional requirements using a combination of distributed query processing techniques that optimise the computation, and cost models that enable PATH2iot to select the best deployment plan given the non-functional requirements. This thesis describes the resulting PATH2iot system, illustrated with two real-world use cases. First, a digital healthcare analytics system in which sensor battery life is the main non-functional requirement to be optimized. This shows that the tool can automatically partition and distribute the computation across a healthcare wearable, a mobile phone and the cloud - increasing the battery life of the smart watch by 453% when compared to other possible allocations. The energy cost of sending messages over a wireless network is a key component of the cost model, and we show how this can be modelled. Furthermore, the uncertainty of the model is addressed with two alternative approaches: one frequentist and one Bayesian The second use case is one in which an acoustic data analytics for transport monitoring is automatically distributed so as enable it to run over a low-bandwidth LORA network connecting the sensor to the cloud. Overall, the paper shows how the PATH2iot system can automatically bring the benefits of edge computing to the increasing set of IoT applications that perform distributed data analytics.|
|Appears in Collections:||School of Computing|
Files in This Item:
|MichalákP2020.pdf||Thesis||11.1 MB||Adobe PDF||View/Open|
|dspacelicence.pdf||Licence||43.82 kB||Adobe PDF||View/Open|
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.