Curriculum
Computer Science for Societal Challenges and Innovation, XXXVIII series
Grant sponsor
Dept of Mathematics (PNRR)
Supervisor

Tullio Vardanega
Co-supervisor
s
Anna Spagnolli; Michele Scquizzato
Contact
edoardo.tinto@phd.unipd.it

Project description
Traditional high-performance computing (HPC), are based on high-performance solutions like clusters and supercomputers, which are expensive to realise and maintain. Cloud computing introduces a different model, where resources are provisioned on a pay-per-use basis; this model is still centralized, because all the computation is performed by cloud datacenters, but it is based on a cloud infrastructure and, it provides lower access barriers to HPC resources. This centralized cloud computing model is no longer suitable to address the needs of applications encompassing the so called cloud-edge continuum, were the computation is expected to move, as well as the data, from and towards the devices connected with the cloud, namely the edge of the network, due to convenience, for example to leverage the availability of quiescent (unused) resources located at the edge, or to necessity, for example to reduce response latency by performing certain computations closer to the end user. Apt technologies have been proposed lately, which allow moving computation in a safe and efficient manner across the continuum, but without offering real-time guarantees on the timing behaviour of this intermediate layer. The main contribution of this project will be to develop a time-predictable runtime to be deployed in every node in the cloud-edge continuum