Cloud HPC - Benefits and Best Practices

UberCloud webinar about the benefits of using cloud hpc for computer aided engineering workflows

https://www.theubercloud.com/cloud-hpc

Transcript

1. Cloud HPC Benefits & Best Practices for Mechanical Engineers
2. 1700 2018 1980s 1800 Lessons for us: 1. The revolutions are coming faster 2. The previous revolution’s technologies get commoditized 3. To reap the benefits of the current revolution, outsource the old (i.e.) computing STEAM ELECTRICITY COMPUTING DATA & INTELLIGENCE Industrial Revolutions
3. UberCloud Mission To make high performance computing (HPC) a reality for every engineer.
4. DESIGN PROTOTYPING & TESTING Current CAE Workflow & Challenges - over-simplified models - iterative process - lack of computational resources - no dedicated IT - difficult to work remotely - difficult to collaborate effectively CAE Pre-Processing Post-Processing Simulation
5. Cloud CAE Workflow & Benefits Pre-Processing Post-Processing - accurate models - more simulations, faster - best in class computational power - support from experts - access results from anywhere - collaborate worldwide CAE Pre-Processing Post-Processing Simulation Simulation
6. Cloud Experience
7. Cloud HPC Experience – Fully Interactive GUI
8. Full GUI For interactive design exploration Software Maintenance Software upgrades multiple times a year Ongoing Operations Adding/removing nodes from the cluster (this is a big benefit and money saver since the entire cluster is not needed 7x24 at peak capacity) Support Break fix services when cluster nodes fail Training Training on how to use the services (transfer files, monitor jobs, etc) Initial Setup Help Setting up and validating software License compliance. Benchmarking Cloud HPC Best Practices CLOUD CAE
9. Analysis The UberCloud-Azure Environment performed 53% faster than a 2 year old on-premise cluster with identical core counts Case Details •  Transonic flow around an airfoil. Flow is 2D – the mesh is extruded to give 3D meshes of various sizes o  Turbulent SST, ideal gas, heat transfer o  Default advection scheme (high resolution) •  Global mesh size: 104,533,000 nodes 103,779,720 hexahedral elements Benchmark Details •  Suitable for 100s of partitions •  Number of iterations: 5 •  Uses Large Problem Partioner •  Partitioning memory requirement: 28GB •  Partitioning time: 15 min on Intel 5670 •  Solver memory requirement (total): 140GB Analysis UberCloud container environment consistently showed strong scalability from 32cores to 512 cores on Microsoft Azure H-Series instances with InfiniBand On-premise Cluster Azure-UberCloud ANSYS 100M Airfoil Benchmark 53% faster Comparison of wall clock solution (solver) time as reported by CFX benchmark
10. FASTER TIME TO MARKET Flexibility lets you use the appropriate hardware based on the urgency of the need GLOBAL ACCESSIBILITY Use the resources anywhere. Make better use of your licenses UNLOCK NEW CAPABILITIES Unlimited hardware means you can run multiple analyses in parallel, do Design Optimizations, use new codes Why Cloud HPC is better CONFIDENCE IN YOUR ANALYSIS Access hundreds of different hardware configurations that are renewed constantly lets you run the analysis the way you want
11. Page No-27 Last week at SC17, Hyperion Research announced that the UberCloud and the Stanford Living Heart Project have won the Hyperion Award for Innovation Excellence. November 20, 2017 THE LIVING HEART PROJECT A TRANSLATIONAL RESEARCH INITIATIVE TO REVOLUTIONIZE CARDIOVASCULAR SCIENCE THROUGH REALISTIC SIMULATION
12. Cloud HPC Case Studies and Resources visit: https://www.theubercloud.com/case-studies
13. https://www.theubercloud.com/cloud-cae contact us: help@theubercloud.com