Enhancing Metadata Transfer Efficiency: Unlocking the Potential of DAOS in the ADIOS context.

In HPC, data movement between applications is typically facilitated by I/O middleware, such as the Adaptable I/O System (ADIOS). This middleware leverages the capabilities of the underlying storage services, to facilitate data movement and distribution. A recent storage system, Intel DAOS, promises to deliver new capabilities for achieving performance and scalability on emerging memory/storage systems. DAOS has already deployed in University of Cambridge, TACC, and the upcoming Aurora supercomputer. This paper investigates the performance tradeoffs associated with mapping ADIOS over one of the many different DAOS interfaces and data models, and makes recommendations for their efficient use.