Imperial College London
Alternate

Dr Chris Cantwell

Faculty of Engineering, Department of Aeronautics

Senior Lecturer in Aeronautics
 
 
 
//

Contact

 

+44 (0)20 7594 5050c.cantwell Website

 
 
//

Location

 

Department of Aeronautics, Room 219City and Guilds BuildingSouth Kensington Campus

//

Summary

 

Publications

Citation

BibTex format

@article{Cantwell:2019:10.1007/s10915-018-0778-7,
author = {Cantwell, C and Nielsen, A},
doi = {10.1007/s10915-018-0778-7},
journal = {Journal of Scientific Computing},
pages = {565--581},
title = {A minimally intrusive low-memory approach to resilience for existing transient solvers},
url = {http://dx.doi.org/10.1007/s10915-018-0778-7},
volume = {78},
year = {2019}
}
Download

RIS format (EndNote, RefMan)

TY  - JOUR
AB  - We propose a novel, minimally intrusive approach to adding fault tolerance to existing complex scientific simulation codes, used for addressing a broad range of time-dependent problems on the next generation of supercomputers. Exascale systems have the potential to allow much larger, more accurate and scale-resolving simulations of transient processes than can be performed on current petascale systems. However, with a much larger number of components, exascale computers are expected to suffer a node failure every few minutes. Many existing parallel simulation codes are not tolerant of these failures and existing resilience methodologies would necessitate major modifications or redesign of the application. Our approach combines the proposed user-level failure mitigation extensions to the Message-Passing Interface (MPI), with the concepts of message-logging and remote in-memory checkpointing, to demonstrate how to add scalable resilience to transient solvers. Logging MPI communication reduces the storage requirement of static data, such as finite element operators, and allows a spare MPI process to rebuild these data structures independently of other ranks. Remote in-memory checkpointing avoids disk I/O contention on large parallel filesystems. A prototype implementation is applied to Nektar++, a scalable, production-ready transient simulation framework. Forward-path and recovery-path performance of the resilience algorithm is analysed through experiments using the solver for the incompressible Navier-Stokes equations, and strong scaling of the approach is observed.
AU  - Cantwell,C
AU  - Nielsen,A
DO  - 10.1007/s10915-018-0778-7
EP  - 581
PY  - 2019///
SN  - 0885-7474
SP  - 565
TI  - A minimally intrusive low-memory approach to resilience for existing transient solvers
T2  - Journal of Scientific Computing
UR  - http://dx.doi.org/10.1007/s10915-018-0778-7
UR  - https://link.springer.com/article/10.1007%2Fs10915-018-0778-7
UR  - http://hdl.handle.net/10044/1/61802
VL  - 78
ER  -
Download