DTIO: Data Stack for AI-driven Workflows
A scalable I/O runtime that unifies disparate I/O stacks for converged HPC, Big Data, and ML workflows.
Research area: HPC
Viper: A High-Performance I/O Framework for Transparently Updating, Storing, and Transferring Deep Neural Network Models
An I/O framework that determines near-optimal checkpoint schedules and accelerates delivery of the latest model updates for DNN training and inference.
TunIO: An AI-powered Framework for Optimizing HPC I/O
An AI-powered I/O tuning framework that reduces tuning time while achieving the same performance gain through RL-driven early stopping.
LuxIO: Intelligent Resource Provisioning and Auto-Configuration for Storage Services
An intelligent storage resource provisioning and auto-configuration service for HPC multi-storage deployments.
Feature Reduction of Darshan Counters Using Evolutionary Algorithms
A study on feature reduction of Darshan counters using evolutionary algorithms to identify pertinent I/O performance predictors.
HFlow: A Dynamic and Elastic Multi-Layered Data Forwarder
A dynamic and elastic multi-layered data forwarding system leveraging a real-time data movement paradigm for HPC.
Apollo: An ML-assisted Real-Time Storage Resource Observer
A low-latency monitoring service that provides direct access to relational telemetry data for HPC middleware and applications.
Characterizing and Approximating I/O Behavior of HDF5 Applications
A poster exploring characterization and approximation of I/O behavior in HDF5 applications using genome-like sequences.
ChronoLog: A Distributed Shared Tiered Log Store with Time-based Data Ordering
A distributed shared multi-tiered log store using physical time to achieve scalable, parallel, high-performance log workloads.