HARE: Self-deploying and Adaptive Data Streaming Analytics in Fog Architectures
Fog architectures are composed by millions of edge, near-user devices that, as a whole, can enable insightful online data analysis (e.g., traffic safety in vehicular networks). The large, fluctuating volumes of data such systems produce shifted the traditional “first the data, then the query” (database) mentality to “first the query, then the data”, leading to the data streaming processing paradigm. Modern stream analysis frameworks scale to hundreds or thousands of homogenous servers. While tailored to Cloud infrastructures, nonetheless, these frameworks do not embrace Fog architectures’ heterogeneity and dynamicity (and cannot thus leverage their huge cumulative computational resources). HARE aims at shaping a new analysis approach tailored to Fog architectures, shifting the “first the hardware, then the query” attitude to “first the query, then the hardware”. HARE will allow for efficient data analysis to happen at millions rather than thousands of devices, providing self-deploying and adaptive capabilities and releasing applications developers from specifying how to best deploy applications and how to balance their load or provision and decommission resources. HARE will be successful by building on top of the applicant’s experience in parallel, distributed and elastic streaming analysis, fine-grained synchronization, leveraging of heterogeneous and diverse hardware and thanks to his connections with the industry (enabling for valid use cases and proper evaluation).
Vincenzo Massimiliano Gulisano (contact)
Associate Professor at Chalmers, Computer Science and Engineering (Chalmers), Networks and Systems (Chalmers)
Swedish Research Council (VR)
Funding Chalmers participation during 2017–2020