ISIS Neutron and Muon Source Data Journal

This is a page describing data taken during an experiment at the ISIS Neutron and Muon Source. Information about the ISIS Neutron and Muon Source can be found at https://www.isis.stfc.ac.uk.


Comparison of Fault Injection and Beam Experiments on GPUs

Abstract: The goal of this proposal is to continue our evaluation of the resilience of modern parallel devices by comparing the experimentally measured FIT rate with the FIT rate estimated through fault injection. Our idea is to characterize the sensitivity of basic resources of GPUs (memories, functional units, conditional statements, mixed-precision hardware) through dedicated synthetic microbenchmarks. Then, we will measure the probability for faults in these basic resources through architectural level fault injectors (SASSIFI and NVBitFI). By multiplying the probability for a fault to happen in a basic resource with the amount of resources used for computation and the probability for a fault in each basic resource to affect computation we can have an estimation of the FIT rate of a code.

Principal Investigator: Dr Paolo Rech
Experimenter: Dr Fernando Fernandes Dos Santos
Experimenter: Mr Fabiano Pereira Libano
Local Contact: Dr Maria Kastriotou

DOI: 10.5286/ISIS.E.RB2000161

ISIS Experiment Number: RB2000161

Part DOI Instrument Public release date Download Link
10.5286/ISIS.E.RB2000161-1 CHIPIR 22 September 2023 Download

Publisher: STFC ISIS Neutron and Muon Source

Data format: RAW/Nexus
Select the data format above to find out more about it.

Data Citation

The recommended format for citing this dataset in a research publication is as:
[author], [date], [title], [publisher], [doi]

For Example:
Dr Paolo Rech et al; (2020): Comparison of Fault Injection and Beam Experiments on GPUs, STFC ISIS Neutron and Muon Source, https://doi.org/10.5286/ISIS.E.RB2000161

Data is released under the CC-BY-4.0 license.



UKRI


Science and Technology Facilities Council Switchboard: 01793 442000