The paper presents a fail-safe NVRAM based mechanism for creation and recovery of data copies during parallel MPI application runtime. Specifically, we target a cluster environment in which each node has an NVRAM installed in it. Our previously developed extension to the MPI I/O API can take advantage of NVRAM regions in order to provide an NVRAM based cache like mechanism to significantly speed up I/O operations and allow to preload large files and operate efficiently on them. In this work, we show how to provide fail safe data write to such files using NVRAM and how to recover from failures. This provides an efficient alternative to costly checkpointing provided an application can store its consistent state in a file.
Authors
- Artur Malinowski link open in new tab ,
- dr hab. inż. Paweł Czarnul link open in new tab ,
- Maciej Maciejewski,
- Paweł Skowron
Additional information
- DOI
- Digital Object Identifier link open in new tab 10.1007/978-3-319-46586-9_11
- Category
- Aktywność konferencyjna
- Type
- materiały konferencyjne indeksowane w Web of Science
- Language
- angielski
- Publication year
- 2016
