Angelos-Christos Anadiotis, EPFL @ PITLab
Enabling adaptive data processing in the IoT network and the cloud servers
Friday, May 10 2019, 12.30
Big Data today are generated by several devices in the context of different applications. This talk will focus on Internet of Things (IoT) devices, as they are considered one of the major Big Data sources, and on transactional applications, as they are a common type of applications that either generate new or update old data and they are strongly connected with business intelligence workflows. In the first case, this talk will describe a network operating system which enables the execution of in-network MapReduce workloads over heterogeneous IoT devices, by adapting packet forwarding both to the workload and the network characteristics in order to offload the network from transferring all generated data to the cloud. In the second case, this talk will describe a system which enables fine-grained elasticity in scale-up cloud servers hosting transactional engines, rendering them adaptive to variations coming both from the side of the workload and of the computing resources availability. Moreover, it will provide an overview of use cases as well as further applications that can exploit elasticity in a scale-up context. Finally, this talk will propose a vision for leveraging theory coming from computer networks in order to jointly optimize query execution and resource scheduling in distributed data management systems.