… on the study of interference during highly parallel asynchronous I/O at node level, focusing
on … Even though the trend of I/O throughput is similar compared with Theta, the peak I/O …

… In this section we report the evaluation of our I/O system on Mira [25] and Theta [24]
supercomputers. First, we demonstrate the scalability of our system in performing parallel writes (…

… We demonstrate the scalability of our system at up to 4,096 processes on the Theta …
Parallel I/O on Theta with Best Practices. Talk at ALCF Simulation, Data, and Learning …

… Basic overview of Lustre File System and the component configuration on theta. … These
libraries offer capabilities to make managing large parallel I/O easier Pay attention to MPI-IO …

… necessary to provide continuous parallel computing over the same … of what they saw as
their best practices and major issues. These … The I/O performance was evaluated using two I/O …

… I/O racks and two Liebert PDUs are inline with the first row of compute racks (to the south).
Behind the first row of I/O … -memory parallel-programming paradigm within a node on Theta. …

… efficient parallel I/O. To … parallel I/O library and evaluate the efficacy of our approach by
using real-world simulation data and a novel approach to microbenchmarking on the Theta …

… For example, Theta has an aggregate write bandwidth of … challenging to incorporate them
into parallel I/O workflow. Besides, … Best practices are provided in the user documentation [26]. …

… Despite these upgrades, however, room remains for improvement in parallel I/O and more
… 2) Theta: Our experiments on Theta show a good I/O performance gain as well. Figure 13 …

… the aggregator count (file count) is a tunable parameter and must be set based on the target
file system, we used A = P/16 as the default for all following experiments (best for Theta). …