Additional 60k core hours granted. - JB On Tue, Jul 21, 2015 at 3:58 PM, [email protected] <[email protected]> wrote:
Hello,
A change in allocation has been requested:
Requester: scarcell (Riccardo Scarcelli) Project: FOA_dual_fuel Title: Efficiency-Optimized Dual Fuel Engine with In-Cylinder Gasoline/CNG Blending Description: The CFD activity is a crucial sub-task of this entire project and initially (Q1 and Q2) aims at fully characterizing the gaseous jets from DI gas injectors through comparison of numerical results against x-ray measurements carried out at APS. In the second half of FY2015 (Q3 and Q4), the comprehensive modeling of gaseous jets will be exploited to carry out investigations of the mixture formation process in a single-cylinder research engine when gasoline is introduced using port fuel injection (PFI) and compressed natural gas is directly injected (DI) into the cylinder.
Simulations will be performed with the commercial code CONVERGE and computational grids of few million cells will require the use of 32-64 cores depending on mesh resolution (the RANS approach will be initially followed). The project is expected to have two members, with 2 or 3 cases run simultaneously on daily bases. With an average of 32 cores per run and 2 cases run in parallel, the expected amount of core-hrs needed for this project for each quarter is about: 32x24x90x2=138,240.
There is not scalability data yet available for this project. However, similar engine simulations showed that with the CONVERGE software and with the expected grid resolution (few million cells) the runs should scale efficiently up to 3 nodes (48 cores). For reference please see the GDI_lean_burn project. Current: undetermined amount Justification:
Requested: 60000
A specific reason has been given: We would need additional 60,000 core-hrs for the project FOA_dual_fuel for Q4 in order to complete all our simulation tasks. In addition to continue our work to meet the deliverables, we are targeting a publication for the next SAE World Congress 2016 and most of the work for this publication needs to be done between August and September. We plan on running all our cases for this publication on 32-64 cores. Below please find the results from a scalability test:
Simulation duration = 1.5e=5 sec Computational time on 1 node - 16 cores = 4700 s Computational time on 2 nodes - 32 cores = 2700 s --> 87% Computational time on 4 nodes - 64 cores = 1600 s --> 73%
This needs to be approved and the final allocation amount decided upon.
Thank You, The LCRC Accounts System _______________________________________________ allocations-admins mailing list [email protected] https://lists.lcrc.anl.gov/mailman/listinfo/allocations-admins