Overview

USRC hosts a summer research symposium showcasing the work of summer interns, full-time staff, and PIs.  Please join us and see the work going on (LANL badge required - external collaborators feel free to contact usrc-contact@lanl.gov for assistance).

Summer Research Symposium 2019

• USRC 2019 Summer Symposium Program
• 2019 High Performance Computing Intern Mini-Showcase (related)

• Topics covered included:

  • Deep I/O: Smart Networks for Fast Storage
  • Improving SaNSA: Spark Integration and Anomaly Detection in HPC State Analysis
  • Shortening Hamming Codes to Better Correct 2-bit Errors
  • Differential Privacy for Supercomputer Sensor Data
  • Profiling HPC Application Resilience using DisCVar
  • Examining Contextual Based Error Correction Techniques in CLAMR
  • Performance Characterization of DRAM-NVM Hybrid Memory Architecture for HPC Applications using Intel Optane DC Persistent Memory Modules
  • Algorithm Learning with the Diagonal Neural GPU
  • Revere: HPC Job Failure Early Alert
  • FI-VIS: Towards Understanding Fault Propagation through Visualization
  • In-Situ Partitioning for Range Queries
  • Tiered Stripeset: Data Availability During Failure Bursts
  • Providing order to the world: Range query for KV-SSD
  • Petavision: Interpolating Video and Up-Sampling Simulations
  • Analyzing Excessive Memory Faults on Trinity and Trinitite
  • KrakenBoot: Firmware-Level Cluster Provisioning via UEFI Surgery

Photos from this event can be found here.

Summer Research Symposium 2018

• Topics covered included:

  • Node resilience analysis on HPC clusters via state transition from multiple fused data sources
  • Compiler based intelligent data placement for heterogeneous memory architectures to improve on performance and resilience
  • A foundation for automated placement of data
  • Investigation of dynamic routing protocols to route network traffic between compute and IO nodes with a cluster
  • Scalable in-situ indexing mechanism for manycore platforms
  • Assessing Ansible to replace core components of the HPC software stack
  • Software engineering, debugging, testing, and enhancements to the DECAF-FSEFI software fault injection system to improve its functionality
  • Using machine learning for system log analysis applied to clusters
  • Building Monte Carlo simulations, using MCNP6, of neutron scattering in HPCs to be used in the determination of error rates due to cosmic radiation
  • Improving IO and recovery performance of storage systems using declustered RAID in ZFS
  • Software development for controlling power to nodes in the raspberry pi cluster
  • Enhancement of system log analysis by identifying the most likely source code origin of each message via cloud search tools
  • Comparing the explanations of LIME and an authentic approach using internal structures of a random forest developed at LANL called “LogAn” for the purpose of determining the potential and reliability of quantitative automatically-generated explanations.
  • Analysis of context-sensitive correction with error correction codes applied to adaptive mesh refinement application
  • Creating fast storage endpoints for the next generation of HPC systems
  • Applying machine learning to supercomputer telemetry data
  • Evaluating storage system performance impacts of hardware acceleration
  • Parallel simulation of mosquito-borne diseases using highly-asynchronous programming model based on OpenSHMEM
  • Using CI/CD tools to automate a cluster installation
  • Using Raspberry Pis and Amazon Web Services to test the launching of OpenMPI across clusters
  • Pioneering the testing of the Open Build Service which will potentially be used to create dedicated clusters to build packages and virtual machine images in the future
  • Applications of deep neutral networks to high performance computing

Photos from this event can be found here.

Summer Research Symposium 2017

• Topics covered included:

  • Node resilience analysis on HPC clusters via state transition from multiple fused data sources
  • Compiler based intelligent data placement for heterogeneous memory architectures to improve on performance and resilience
  • A foundation for automated placement of data
  • Anomaly Detection in System Logs
  • Burst Buffer Simulation In Dragonfly Network
  • Radiation Induced Effects on HPC Hardware
  • Reining in Complex Memory
  • DeltaFS - Parallel In-situ Data Indexing
  • Error Injection with a DIMM Fault Injector
  • Inexact Computing for Pre-Post_Moore's and Post-Moore's Time Frames
  • Effects of Cosmic Ray Neutrons on Modern HPC Components
  • To Share or Not To Share: Comparing Burst Buffer Architecture
  • Soft Error Resilience and Failure Recovery for Continuum Dynamics Applications
  •  Metadata Load Balancing and Key-Value Stores
  • Probabilistic Graphical Models for DRAM Faults
  • Interpretable Context-Aware Anomaly Detection for High Performance Computing Systems Monitoring Syslog
  • Multi-scale Analysis of Resilience and Energy Efficiency for Scientific Applications
  • Temporal and Spatial Analysis of SEDC Energy Data
  • Overview of USRC, the Ultrascale Systems Research Center 

Photos from this event can be found here.