Added 150K to this project plus 500 to get them out of the negative balance. -- John Roberts Argonne National Laboratory CELS Systems [email protected] On 2/23/17, 12:01 PM, "[email protected] on behalf of Bair, Raymond A." <[email protected] on behalf of [email protected]> wrote: We have converged on adding 150K to SHARP-ARC ----------------------------------------- Ray Bair Argonne National Laboratory On 2/23/17, 11:13 AM, "Yu, Yiqi" <[email protected]> wrote: Dear Ray, I appreciate that very much. I understand that heavy load on blues. I will borrow sometime from other project if we are short. Yiqi -----Original Message----- From: Bair, Raymond A. Sent: Thursday, February 23, 2017 11:11 AM To: Aithal, Shashikant M.; Yu, Yiqi Subject: Re: SHARP-ARC - Project Allocation Request Dear Yiqi, Note that your new second-half allocation will start April 1, providing more time for your project. Could you make do with 150,000 more hours now, to bridge the gap until April 1? Regards, Ray ----------------------------------------- Ray Bair Argonne National Laboratory On 2/23/17, 10:45 AM, "Aithal, Shashikant M." <[email protected]> wrote: Thanks, Yiqi. We will try to make a decision soon. Shashi ________________________________________ From: Yu, Yiqi Sent: Thursday, February 23, 2017 10:42 AM To: Aithal, Shashikant M. Cc: Bair, Raymond A. Subject: RE: SHARP-ARC - Project Allocation Request Sure, As you know 1M elements for nek5000 case is common. So we assume we need to run the case with 1M element. It normal takes 100K core hours to converge. We need to run at least 5 of these case. We still need some time for debugging and for another code. So that's how 1M comes from. Of course, If the resource don't permit, we can try to simplify the case and use URANS instead of LES to save some time. Let me know if you need any other information. -----Original Message----- From: Aithal, Shashikant M. Sent: Thursday, February 23, 2017 10:34 AM To: Yu, Yiqi Cc: Bair, Raymond A. Subject: RE: SHARP-ARC - Project Allocation Request Thanks, Yiqi. It will be helpful if you could let us know the typical time it takes for a run and how many runs you need to finish - say if a run converges in 10 hours on 256 cores, that would be 2560 core-hours. If you have to run 10 such cases, you would need some 25-30K core-hours (assuming some runs take longer or diverge). We (the LCRC-core) can get an idea of how the allotted time is going to be used and approximately how much work can get done. Please provide us with this information (based on the runs you have already done) and we can make a quick decision as to how much time we can grant your project. Thanks, Shashi ________________________________________ From: Yu, Yiqi Sent: Thursday, February 23, 2017 10:21 AM To: Aithal, Shashikant M. Cc: Bair, Raymond A. Subject: RE: SHARP-ARC - Project Allocation Request Shashi, We are going to run some multi-physics simulation(Nek5000 and some neutronic code coupled ) with SAHRP. As you may know, we don't have choice to run these calculation on Mira and Cetus since we are not able to compile the SHARP package on these cluster (very complicated situation). The time for this project is negative now. We need to run some calculations till we can get the next round time till April 1st. Please grant us as much as you can and we will appreciate it very much. Best Yiqi -----Original Message----- From: Aithal, Shashikant M. Sent: Thursday, February 23, 2017 10:14 AM To: Yu, Yiqi Cc: Bair, Raymond A. Subject: SHARP-ARC - Project Allocation Request Yiqi: Can you give us some more information about the jobs you plan to run between now and end of March and how you computed the 1M core-hour request? As you know, our old 2500 core machine Fusion, has been "retired" to make room for the new machine. So there is almost a 30% extra load on Blues. Hence we (LCRC and the allocation committee) are unable to grant large core-hour requests at this time. We might be able to grant you about 80-100K core-hours from now to March end. You might also want to consider running some of the simulations on Mira or Theta since these machines are better suited for jobs requiring large core-hours. Please let me know if you have any other questions or concerns. Thanks, Shashi ________________________________________ From: [email protected] [[email protected]] on behalf of [email protected] [[email protected]] Sent: Wednesday, February 22, 2017 1:56 PM To: [email protected] Subject: [allocations-admins] [LCRC Accounts] Project Allocation Request Hello, A change in allocation has been requested: Requester: yyu (Yiqi Yu) Project: SHARP_ARC Title: Multi-physics Simulation of SFR Core Radial Expansion with SHARP Description: As the SHARP modeling and simulation toolkit is maturing, the focus is on customer-oriented applications and further validation exercises. This allocation will help with those applications, for performing small to moderate scale simulations and for scaling up to ALCF. The SHARP toolkit is regularly compiled on Blues using Intel or GNU compilers. The components of SHARP—the Nek5000 computational fluid dynamics code and the PROTEUS neutron transport code—have already been demonstrated to scale well not only on Blues but on Blue Gene/P. This project will be used by 4-5 analysts in the NE and MCS Divisions. Current: undetermined amount Justification: FY17 efforts for SHARP are highly focused on solving challenge problems for customers and code validation. FY16 efforts were focused on code development, as there was a major release, and so computational requirements were lower than usual. If the project is successful, we anticipate that FY17 and beyond will be highly resource intensive. The components of SHARP—the Nek5000 computational fluid dynamics code and the PROTEUS neutron transport code—have already been demonstrated to scale well not only on Fusion but on Blue Gene/P. Nek5000 has demonstrated > 80% parallel efficiency on BG/P (strong scaling) out to P=131,072 processors, using only 7300 points/processor at the largest scale. PROTEUS has demonstrated greater than 90% strong space-angle scaling at the 100,000 processer level and was nominated for the Gordon Bell Prize for its parallel performance. Requested: 1000000 A specific reason has been given: Used up all the time for now. have urgent job to finish before March 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