VELOC: VEry-Low Overhead Checkpointing System

VELOC is a multi-level checkpoint/restart runtime that delivers high performance and scalability for complex heterogeneous storage hierarchies without sacrificing ease of use and flexibility.

It is primarily used as a fault-tolerance tool for tightly coupled HPC applications running on supercomputing infrastructure but is essential in many other use cases: suspend-resume, migration, debugging, roll-back and explore alternative directions.

VELOC is a collaboration between Argonne National Laboratory and Lawrence Livermore National Laboratory as part of the Exascale Computing Project.

Documentation

The documentation of VeloC is available here: http://veloc.rtfd.io

It includes a quick start guide as well that covers the basics needed to use VELOC on your system.

Availability

VELOC is part of the Extreme-scale Scientific Software Stack (E4S). Chances are it is already pre-installed on large supercomputing systems (e.g., as a module).

Furthermore, it is integrated with Spack package manager.

Bindings

VELOC supports checkpointing of complex data structures in Python through a straightforward API that hides the details of serialization using the pickle module. Furthermore, it seamlessly integrates with popular C++ serialization libraries such as Cereal and Bitsery.

Contacts

In case of questions and comments or help, please contact the VeloC team at [email protected]

Release

Copyright (c) 2018-present, UChicago Argonne LLC, operator of Argonne National Laboratory
Copyright (c) 2018-present, Lawrence Livermore National Security, LLC. Produced at the Lawrence Livermore National Laboratory.

For release details and restrictions, please read the LICENSE and NOTICE files.

LLNL-CODE-751725 OCEC-18-060