Kestrel on Power10
for High-Performance
Simulations
From Long Processing Times to Results Up to 7× Faster
Infodataserv designed and implemented a High Performance Computing (HPC) solution on Power10 with Kestrel to Optimize Critical Simulations and Accelerate Results Generation in a High-Demand Environment.
Simulations.
Better
Decisions.
of greater possibilities
Infrastructure
Modeling
From Challenge to Results
An HPC Solution designed to Accelerate Knowledge and Decision-Making.
What changed with the new Solution?
More Capacity, More Speed, More Knowledge.
BEFORE
-
High Processing Times
-
Limited Capacity for Complex Simulations
-
Restrictions in Scenario Analysis
-
Lower Agility in Decision-Making
AFTER
-
Up to 7× Better Performance
-
Greater Processing Capacity
-
More Complex and Accurate Simulations
-
Results in Less Time
Challenge
An Institution with Critical Computing Needs that exceeded the capacity of its Infrastructure.
Evaluate Kestrel Functionality on Power10
Verify Compatibility, Compilation, and Execution of the Kestrel versions: Serial (base), Parallel (OpenMP), and AT16 (Advance Toolchain 16) on the Power10 Architecture with Ubuntu 22.04 Operating System on a Dedicated LPAR.
Optimize Processing Times
Evaluate Performance through High-Resolution and Parallel Processing (OpenMP) tests, measuring Speedup (T_serial / T_parallel) to determine the actual performance gain compared to previous Architectures.
Event: La Gasca Mudslide, Quito · January 31, 2022. A flow of mud and rocks that descended through the El Tejado ravine as a consequence of intense rainfall that caused the collapse of the intake structure in the basin. The event affected the La Gasca and La Comuna neighborhoods in Quito, with severe impact on the population.
Technical Requirements: Create an LPAR with PowerVM on Power10 · Create a SAS NL Disk Array · Install Ubuntu 22.04 with libraries for Kestrel · Install and Configure Kestrel Software
Implemented Solution
Infodataserv collaborated with the University of Bristol and the client's team to Configure the HPC Environment.
1.6 Physical Cores converted into 24 threads (SMT8), 48 GB of RAM, 60 GB of disk, and 16 Gb/s Fibre Channel I/O — isolated and optimized environment for HPC workloads.
Installation of the Ubuntu 22.04 Operating System and the required libraries. Compilation and Configuration of the Serial, Parallel (OpenMP), and AT16 versions of Kestrel.
Storage configuration with a SAS NL array to handle the large Data Volumes required by High-Resolution Simulations.
Enablement of Parallel Processing with OpenMP, leveraging the 24 available threads of the LPAR to distribute the Computational Load of the Simulations.
Compilation with AT16 to Optimize the Kestrel Code specifically for the Power10 Architecture, maximizing the use of the Processor's capabilities.
Configuration of secure Remote Access via VPN to the Institution's environment, allowing Infodataserv and University of Bristol Specialists to collaborate in System validation.
Implemented HPC Architecture
A Scalable and Optimized Infrastructure for Intensive Processing Workloads.
Results that Transformed Processing
Verified and Reproducible Performance for Complex Operational Environments.
Better Performance
Up to 7× better performance in Critical Simulations.
Optimized Processing
Greater capacity to execute demanding Computational Workloads.
More Accurate Results
Models and scenarios with a higher level of detail.
Scalable Infrastructure
Platform prepared for future Simulation and Analysis Projects.
Power10 vs. Previous Architecture — La Gasca Event
| Kestrel Version | Intel Architecture (Bristol) | Power10 (La Gasca) | Improvement |
|---|---|---|---|
| Serial (Cotopaxi Event) | 77 days | — | — |
| Serial (La Gasca Event) | 35 days | 2 days 15 h | ~12× faster |
| Parallel (La Gasca Event) | 3 days | 9 hours | 7× faster |
Main Result: The corrected Parallel Version achieved up to 7× better Performance compared to the Serial Version on Power10, reducing Simulation Time from days to just 9 hours, with physically consistent and reproducible results for Complex Operational Environments.
Technologies Used
Why Infodataserv?
High-Performance Infrastructure Specialists with Technical Vision and Proven Experience.
Power Specialization
Certified Engineers with experience in Implementing and Configuring Power environments for Critical High-Performance Workloads.
International Collaboration
Ability to work with International Institutions such as the University of Bristol on Highly Complex Technical Projects.
Measurable Results
Focus on Verifiable Performance Metrics: documented 7× Speedup in Parallel Simulations on the Power10 Architecture.
Proven Experience
More than 13 years implementing Critical Infrastructure Solutions in Government, Healthcare, Energy, and Science Sectors.
HPC Focus
Ability to design High Performance Computing environments adapted to the Specific Needs of each Organization.
End-to-End Vision
Coverage of the entire Architecture: Server, Virtualization (LPAR), Storage, Network, and Software, through a single Specialized Team.
Do you need greater capacity for critical workloads?
We Design and Implement HPC Infrastructures for Organizations that need Greater Performance, Availability, and Efficiency.
operations
without limits
