Ian Adams discusses DAOS-Distributed Asynchronous Object Storage at Intel As we enter the exascale era , current storage techniques and technologies are struggling to cope with the demands of modern HPC and “Big Data” applications. Hundreds of thousands of processes and many petabytes of data may be transferred during a single application run. Even with new storage technologies, the overhead of many storage systems is coming to dominate the access latency for many IO operations. To explore how to address these issues, we are working on DAOS: Distributed Asynchronous Object Storage. DAOS leverages new technologies and techniques, utilizing byte-addressable non-volatile memories and high-speed network fabrics to enable fine-grained versioning and automatic caching and tiering of HPC applications and data at exascale. In our talk, we will give a broad overview of the current DAOS architecture and some of the challenges and solutions we’ve encountered thus far. https://github.com/daos-stack