The DEEP-ER project features a highly scalable, efficient, and user-friendly parallel I/O system based on the Fraunhofer parallel file system BeeGFS (formerly known as FhGFS). Additionally, it provides a low overhead, unified user-level checkpointing system and exploiting the multiple levels of non-volatile memory and storage added to the DEEP architecture.
Extensions to the Posix I/O API enable applications to efficiently use the file system and the different levels of the memory/storage subsystem. These extensions originate from BeeGFS itself, the parallel I/O library SIONlib, and Exascale10 – a novel I/O concept developed by the Exascale10 Workgroup.
Building on the capabilities of the DEEP-ER I/O system and the characteristics of the OmpSs programming model, a dual-approach resiliency concept combines a coarse-grained application-based multi-level checkpoint/restart mechanism with a less intrusive and more fine-grained scheme for task-controlled recovery from component failures.
Within the DEEP-ER project, developments are made in four key software areas:
- Programming software
- I/O software
- Resiliency software
- Benchmarking & Tools
The software used in DEEP-ER is mostly open source. The developments within the project have partially also been reflected in the release updates as it is for instance the case with SIONib or BeeGFS. Some software has been entirely developed within the project, e.g. the libNAM library which enables access to the Network Attached Memory (NAM).
Below you find an overview on the software used in the project and where to get more information, where to download the latest releases and to access the documentation.
Page 1 of 2