System and application scenarios for disaster management processes, the rainfall-runoff model case study

Abstract

In the future, the silicon technology will continue to reduce following the Moore’s law. Device variability is going to increase due to a loss in controllability during silicon chip fabrication. Then, the mean time between failures is also going to decrease. The current methodologies based on error detection and thread re-execution (roll back) can not be enough, when the number of errors increases and arrives to a specific threshold. This dynamic scenario can be very negative if we are executing programs in HPC systems where a correct, accurate and time constrained solution is expected. The objective of this paper is to describe and analyse the needs and constraints of different applications studied in disaster management processes. These applications fall mainly in the domains of the High Performance Computing (HPC). Even if this domain can have differences in terms of computation needs, system form factor and power consumption, it nevertheless shares some commonalities.