Although it can be used more generally, CRUISE was designed specifically for checkpointing HPC applications on Linux clusters. CRUISE implements a POSIX-like file system in user space that stores data in main memory and transparently spills over to other storage, like flash memory, disk, or a parallel file system. It exposes file data to user-level Remote Direct Memory Access (RDMA), allowing remote processes to access files with reduced CPU interruption. CRUISE scales linearly with node count and it has been shown to deliver 1 PB/s throughput at three million MPI processes on the Sequoia system at LLNL.
Be the first to post a review of cruisefs!