Presentation on theme: "My Road to Supercomputing Holly Krutka The University of Oklahoma."— Presentation transcript:
My Road to Supercomputing Holly Krutka The University of Oklahoma
Research Background Using a computational fluid dynamics (CFD) package Fluent to simulate air flow The air flow is related to the melt blowing process of polymers Chemical engineer NOT A COMPUTER SCIENTIST!!!
Why Use CFD? Experiments had already been completed There is no affordable way to include the polymer in the air flow and take experimental measurements To be on the cutting edge of research in our field FAST AND CHEAP!!!
How Fluent Works Make a computational domain Split domain up into finite volumes Set boundary conditions Set model parameters Fluent solves the continuity equation, momentum equations, ect. Analyze simulation results
Undergraduate Research 2D simulations Symmetrical Isothermal initially Grid cells were small About 100,000 cells per grid Using a single processor on a lab computer Simulations required between 2-6 days to finish
Graduate Research Simulations become much more complex Non-isothermal 3D 700,000 – 2,000,000+ Including polymers Time dependent simulations Lots of simulations to run…and only two processors?!?!?!?
Supercomputing CFD program is actually designed to run on a supercomputer OU has a great supercomputing program One problem….still NOT a computer scientist YOU DON’T NEED TO BE!
Using Sooner CFD package was loaded on to Sooner (Brandon George, OSCER Manager of Operations ) Started using sooner.oscer.ou.edu (IBM p690, 32 POWER4 1.1 GHz CPUs, 32 GB RAM -- now decommissioned) Use Secure Shell to move my simulations to Sooner Wrote a simple script file (with help) to run the simulations Submit the simulations to the queue and wait
From Sooner to Topdawg Fluent loaded on to Topdawg (Brandon George) Later transitioned to topdawg.oscer.ou.edu (Dell Xeon cluster,1024 Xeon 3.2 GHz CPUs, 2176 GB RAM) Simulations are submitted in a queue, similar to Sooner Great improvement on speed Desktop → Sooner → Topdawg
Advantages Using more processors at a time (4 – 8) Much faster More simulations at once Much more stable My desktop computer is free to work
Conclusions You do not need to be a computer scientist to use OU’s supercomputing resources Utilizing OSCER has greatly improved both the quality and quantity of my PhD research