CL-HPC 2018

CL-HPC 2018

Welcome to CL-HPC 2018: CloudLighing HPC Industrial Brief and Workshop -- From HPC to Cloud and AI, NTNU-Trondheim, Norway, Jan. 15, 2018

Venue: Dept. of Computer Science, NTNU-Trondheim --- All talks in Room 454, IT Building (Sem Sælandsvei 9, Trondheim) 

This event covers an industrial briefing of the work HPC-Lab members did as part of the EU H2020 project CloudLightning, which included looking at an OPM (Open Porous Media) application. The event will also include several talks by HPC-Lab's current and future collaborators discussing future developments related to HPC, including applications, hetereogeneous computing and AI.

Program    

9:30-9:45 Mingling / coffee/ tea  (*) 

(*) Maxeler is not attending since diverted to Oslo in storm Sunday night

9:45-10:00   Welcome  (Dr. Anne C. Elster, HPC-Lab)

10:00-10:30 OPM and HPC in the Industry (Dr. Alf B. Rustad, Statoil Research, Trondheim)

10:30-10:45 Waffle Break

10:45-11:30 CloudLightning @NTNU: From OPM Containerization to Enabling HPC in Clouds (Dr. Malik M. Khan, HPC-Lab)
 
11:30 - 12:15 High Performance Seismic Modelling with Maxeler Dataflow Technology (Dr. Tobias Becker,  Maxeler Technologies, London, UK) (*** will likely be postponed due to storm)
 
11:00-12:30 Overview of Current & Future HPC-Lab projects /Tour of HPC-Lab
 
 
12:30-13:30 Lunch
 
**** Dr. Gavin Taylor is also coming in from out of town, so afternoon scheduling will depend on his and Dr. Beckers ability to make it to Trondheim in the storm...
 
13:30-14:15 Visualizing the Loss Landscape of Neural Networks (Dr. Gavin Taylor, US Naval Academy)
 
14:15-15:00 Convolutional Neural Networks on Dataflow Engines (Dr. Tobias Becker, Maxeler)
 
15:00-15:15 Cake break
 

15:15- 15:30 CFD Simulation of Multiphase Reacting Systems (tent)  (Dr. Tian Li, Dept of Energy and Process Engineering, NTNU)

15:30 - 16:30  TBD /Informal Panel/Wrap-up

-------

Abstracts

OPM and HPC in the Industry, Dr. Alf B. Rustad, Statoil Research

Abstract

High performance computing has played a very important role in the oil and gas industry for decades. Traditionally, seismic processing has been the most prominent area, but today reservoir simulation is a fast growing use case. Reservoir simulation is solving the flow equations in the reservoir, and hence the way historic and future production of the reservoir is calculated. Oil and gas reservoirs are subject to large uncertainties, and handling these uncertainties have resulted in reservoir engineers now routinely simulating a hundred realizations where only one or a few were run earlier. In addition, there is a persistent need to increase detail and complexity in existing reservoir models. In sum, this means that the computational need within reservoir simulation increases rapidly.

The flow equations solved by reservoir simulators are systems of partial differential equations with typically no analytical solutions available, so they need to be solved numerically. Moreover, they are very hard to precondition efficiently and they are very hard to parallelize. This means that the software is highly specialized and thus very expensive with a very high barrier of entry. The OPM project aims to solve these challenges through open source development, enabling multiple communities with different competence to collaboratively come up with reservoir simulators more capable than any single community can muster. In this talk we will share the current status of OPM, and give some forward looking predictions.

---------------------

For further information, contact the organizer:

Prof. Anne C. Elster, HPC-Lab Director

E-Mail: elster@ntnu.no
Phone: +47 981 02 638