Submission #18: And what would you do with ten thousand Raspberry Pis? ====================================================================== Authors ------- 1. Jeremy Singer (University of Glasgow) 2. Anna Lito Michala (University of Glasgow) 3. Herry Herry (University of Glasgow) Abstract -------- The introduction of ubiquitous low-cost, low-power compute devices, e.g. in the Internet of Things (IoT,) is fundamentally changing the computational landscape. Although we already see some benefits provided by special purpose IoT devices, the true capability is only realised when we can integrate large numbers of distributed devices as part of a much larger federated service. Our research is currently focusing on lightweight management infrastructures for IoT devices that are not necessary co-located. Developments in the Smart City and Smart Campus domains pose interesting research challenges in the way we maintain, manage, and use the expected cyber-physical systems that emerge. In the Anyscale Applications project [1] we have examined ways of incorporating reusable infrastructure to enable research in the smart campus environment. In the FRμIT project [2] the focus is shifted towards a management framework that allows users to maintain and manage devices that do not reside within the same physical network. In both projects we utilise single board computers (Raspberry Pi) to extend the capabilities of conventional IoT devices and increase edge computing potential. We envisage a testbed comprising tens of thousands of nodes, geographically distributed and capable of challenging remote off-grid scenarios. We aim to popularise the concept of a micro-datacenter, a minimal footprint, low-power and low-cost compute cluster that can be pushed to the edge of distributed networks, fundamentally changing the current model of the centralised datacenter. We have already built prototype micro-datacenters at multiple sites, for example a mini-HPC cluster (Southampton University), a scale model cloud datacenter (Glasgow and Loughborough University), and a decentralized sensing and communication platform for ultra-remote internet blackspots (Cambridge University). This project is about connecting together our isolated micro-datacenters to produce a federated, geo-distributed testbed. Further, to continue from our Anyscale Applications project work we are building a scalable smart campus testbed [3], using reconfigurable supersensor nodes. Currently available commercial IoT alternatives (e.g. Nokia/Siemens) have ongoing issues with security, scalability of remote management, application retargetability, and data availability. We propose disposable compute nodes, small-scale embedded programmable devices with significant compute power (e.g. Raspberry Pi), along with commodity sensors (light/temperature/CO2/sound/motion) and wireless comms (e.g LoRA, 5G). Our framework will become the enabling technology for different applications as system and user requirements evolve. For example, it can facilitate student interaction with the physical campus via smartphones, indoor location and navigation, room occupancy monitoring, campus heat maps for data measures, emergency management, on the spot satisfaction surveys, and ubicomp research. In summary, our research aims to address some of the emerging challenges in fragmented aggregations of micro-clusters, alongside the development of exemplar use cases for such systems. [1] anyscale.org [2] fruit-testbed.org [3] https://doi.org/10.1109/FiCloud.2016.16