Three-Level Stochastic Modeling of High-Performance and Distributed Computing Systems

3 maggio 2019
Versione stampabile

Time: h.12.00am
Venue: Via Sommarive 5 - Polo Ferrari 1 (Povo, TN) - Room Ofek

Speaker

  • Alexander Rumyantsev, Senior Researcher, Institute of Applied Mathematical Research (Karelian Research Centre of the Russian Academy of Sciences), Petrozavodsk

Abstract

High-performance and distributed computing systems are tools for computations acceleration used both in science and business environments. These two classes are mainly distinguished by reliability of computation elements (nodes, processors), speed/latency of interconnect, and cost of ownership. While high-performance computing systems are specially made for computations, distributed computing systems are utilizing idle resources of regular enterprise/private infrastructure (e.g. personal computers, laptops, tablets). However, in both classes, system performance and cost are the main objects to study. 

This talk delivers a brief overview of recent studies and brings some new results in the field of stochastic modeling of high-performance and distributed computing systems. A three-level approach is a group of interconnected and cross-validated models of analytical, simulation and technical level, that allow to transfer theoretical results towards practical applicability.

Contact: enrico.blanzieri [at] unitn.it (Enrico Blanzieri)