Category – User Experiments

Storage Research Experiment Patterns on Chameleon Cloud and Trovi

Today, two UChicago students share with us their thoughts on how to create reproducible experiments in a cost effective manner. Ray Sinurat and Yuyang (Roy) Huang talk about the experiment patterns for storage experiments they created and describe how they can serve as a basis for developing storage experiments. Best of all – they share the experiment patterns with the Chameleon community – we hope you will find them useful! 

SC: The largest Reproducibility Laboratory

Today we share a very unique user experience -- a conversation with Rafael Tolosana Calasanz who is an Associate Professor in the Department of Informatics of the University of Zaragoza, Spain and has participated in the reproducibility initiative at SC. Rafael shares with us his experiences reproducing artifacts on Chameleon and his insights on reproducibility and its importance to the modern scientific process. 

Tensor Analysis

This month's user experiment blog covers some interesting work on tensor analysis from researchers at Arizona State University.

Zeus: GPU Energy as a First-Class Resource in DNN Training

In this month's user experiment blog we get a fascinating insight into how much power training deep neural networks (DNNs) consumes – and how to make it less. The authors’ discuss research presented as part of their NSDI ’23 paper, describe how they structured their experiments on Chameleon, and explain why bare metal resources are essential for power management research. 

Ring around the edges: self-organizing overlay VPNs linking distributed edge resources

Learn how to use to easily create Virtual Private Networks (VPNs) and run unmodified middleware and applications across edge and cloud computing resources across networks with different firewalls and NATs (Network Address Translators).

Large Scale Ensemble Simulations

Researchers from Arizona State University developed DataStorm -- an easy-to-use platform for large scale ensemble simulations, which enables researchers to collaborate and achieve deep actionable insights.

Chameleon, and Simulating Self Propagating Malware to Evaluate Detection Technology

How do you develop and evaluate a new analytic on a network connection data set across large, enterprise systems without malware used to train machine learning models for cyber attacks? Researchers at the University of Virginia approach the problem by simulating self-propagating malware.

Exploring Process-in-memory Architecture for High-performance Graph Pattern Mining

Graph Pattern Mining (GPMI) applications are considered a new class of data-intensive applications -- they generate massive irregular computation workloads and pose memory access challenges, which degrade the performance and scalability significantly. Researchers at the Illinois Institute of Technology approach the problem by using the emerging process-in-memory architecture.