Computational resource orchestration tools are one of the key enabling technologies for computational science (e-science).
The representation of tasks dependent on data producer/consumer relationships finds theoretical solidity in Directed Acyclic Graphs (DAGs) and technological implementation in engines for managing scientific workflows.
Considering a workflow consisting of two tasks (A and B, with A producing a[0..n[ file) such that B uses that data, we have that B will be executed when A is finished. This leads to inefficiency due to a lack of overlap between computation and input/output: when A has already produced a[0..m[, B could start processing a[0..m[ while A produces a [m,..,n[.
CAPIO is a middleware capable of inserting streaming I/O capabilities into workflows whose task communication is based on files, improving the overlap between I/O and computation without the need to modify the application code.
This project presents “Directed Acyclic Graph on Cross-Application Programmable I/O” (DAGonCAPIO), a workflow engine that will use CAPIO for task communication via files.
Through a predominant corporate commitment, aimed at industrial research in the field of HPC/Cloud computing convergence, DAGonCAPIO will be able to execute workflows whose tasks can be executed on on-premises, on-cloud and hybrid resources.
Since the rapid, effective and efficient response to situations that can compromise human health or environmental conservation is a primary requirement today, the functioning of DAGonCAPIO will be demonstrated through a multidisciplinary application for the simulation of transport and dispersion of inert materials similar to pollutants spilled into the sea and the subsequent prediction of the quality of filter-feeding organisms bred and intended for human consumption, through a coupled model based on artificial intelligence.
Speaker
Raffaele Montella – University of Naples “Parthenope” & CINI lab HPC-KTT
Raffaele Montella is an Associate Professor with tenure in Computer Science at the Department of Science and Technologies (DiST), University of Naples “Parthenope’” (UNP), Italy. He teaches computer architecture, web technologies, cloud computing, mobile computing, future architectures, and parallel programming paradigms. His main research topics and scientific production are focused on tools for high-performance computing, cloud computing, and GPUs with applications in computational environmental science (multi-dimensional geo-referenced big data, distributed computing for modeling, scientific workflows, and science gateways). He is the director of the High-Performance Scientific Computing Laboratory and a member of the Governing Board of the CINI Laboratory on HPC-KTT.
Event Timeslots (1)
Thu 19 – Accelerators
-
R. Montella (University of Naples "Parthenope" & CINI lab HPC-KTT)