Jerome Soumagne | Searcharxiv

arXiv2017

DAOS for Extreme-scale Systems in Scientific Applications

Exascale I/O initiatives will require new and fully integrated I/O models which are capable of providing straightforward functionality, fault tolerance and efficiency. One solution is the Distributed Asynchronous Object Storage (DAOS) technology, which is primarily designed to handle the next generation NVRAM and NVMe technologies envisioned for providing a high bandwidth/IOPS storage tier close to the compute nodes in an HPC system. In conjunction with DAOS, the HDF5 library, an I/O library for scientific applications, will support end-to-end data integrity, fault tolerance, object mapping, index building and querying. This paper details the implementation and performance of the HDF5 library built over DAOS by using three representative scientific application codes.

cs.DC↗

arXiv2015

A Remote Procedure Call Approach for Extreme-scale Services

When working at exascale, the various constraints imposed by the extreme scale of the system bring new challenges for application users and software/middleware developers. In that context, and to provide best performance, resiliency and energy efficiency, software may be provided as a service oriented approach, adjusting resource utilization to best meet facility and user requirements. Remote procedure call (RPC) is a technique that originally followed a client/server model and allowed local calls to be transparently executed on remote resources. RPC consists of serializing the local function parameters into a memory buffer and sending that buffer to a remote target that in turn deserializes the parameters and executes the corresponding function call, returning the result back to the caller. Building reusable services requires the definition of a communication model to remotely access these services and for this purpose, RPC can serve as a foundation for accessing them. We introduce the necessary building blocks to enable this ecosystem to software and middleware developers with an RPC framework called Mercury.

cs.DC↗