In current supercomputers, storage is typically provided by parallel distributed file systems
for hot data and tape archives for cold data. These file systems are often compatible with …

Modern High-Performance Computing (HPC) data centers routinely store massive data sets
resulting in millions of directories and billions of files. To efficiently search and sift through …

A key emerging pattern in deep learning applications is the need to capture intermediate
DNN model snapshots and preserve or clone them in explore a large number of alternative …

Composition of computational science applications, whether into ad hoc pipelines for
analysis of simulation data or into well-defined and repeatable workflows, is becoming …

Our previous work, which can be referred to as EMPRESS 1.0, showed that rich metadata
management provides a relatively low-overhead approach to facilitating insight from scale …

Traditional I/O stacks feature a strict separation of layers, which provides portability benefits
but makes it impossible for storage systems to understand the structure of data. Coupling …

The file system metadata service is the scalability bottleneck for many of today's workloads.
Common approaches for attacking this" metadata scaling wall" include: caching inodes on …

Scientific data sets have grown rapidly in recent years, outpacing the growth in memory and
network bandwidths. This I/O bottleneck has made it increasingly difficult for scientists to …

In the post-exascale era storage systems, a fundamental challenge faced by the research
community is the efficient and scalable access to the stored information while meeting the …

This paper describes an attempt to use a NoSQL database engine to manage custom
metadata using a rich query interface as motivating and descriptive examples of what kind of …