e@LU är ett samverkansprojekt inom Lunds universitet med avseende på einfrastrukturfrågor,
med särskilt fokus på de anläggningar och organisationer som finns vid
universitetet med koppling till e-vetenskap, t.ex. MAX IV, LBIC, HumLab och Skissernas
museum. Dessa har liknande behov vad gäller lösningar för datahantering, men också analys
av data samt modellering och simulering. Projektet omfattar hårdvara och ett antal applikationsexperter,
som ska ge stöd till olika områden.
e@LU administreras av Lunarc och lyder under dess styrelse.
En viktig del av projektet är de applikationsexperter som skall hjälpa forskargrupperna med frågor som de själva inte har tid eller kunskap att hantera. e@LU har sedan i Maj 2 st applikationsexperter inom följande områden:
Metadata, ontologies and data provenance
Science is all about data and now Science is turning into eScience, where all data is transferred to and exchanged in digital form. This opens up vast possibilities but also presents some major challenges. In order for humans and computers to understand the data outside of the context of the individual or group of scientists that created the data, and to preserve the usability over time, the issue of enriching the digital data with digital metadata including data provenance is becoming an urgent matter. This has led to a strongly developing and dynamic field where semantic interoperability is the keyword. This is the consistency of meaning and understanding within and between the large amount of data produces and data users. As the scientific data is complex by nature and requires understanding of the context of the science that produces and consumes the data this is far from trivial but it is essential for almost all sciences in the future.
The problem is relatively large in environmental and also social/medical sciences, where definitions are sometimes trivial and/or widely different among communities that study the same parameters from different viewpoints or locations. Just one example is the definition of what is “a forest in land-use classification”.
Applikationsexperter: Monica Lassi, Jörgen Eriksson, Maria Johnsson
Projektledare: Monica Lassi / Alex Vermuelen
Datahantering med särskild hänsyn till säkerhetsaspekter med avseende på humandata
The discussion rather quickly converged on what all participants see as an urgent need at the university level. The manipulation, coordination, and storage of personally identifiable data are topics that concern researchers at multiple faculties and departments, but are completely absent at the university level. These topics are very broad, and beyond the reach of a single half-time employee. The group felt strongly that all topics are vital to the competitiveness of the university and should be accounted for in the medium to long term. For the purposes of the pilot program e@LU, it was decided that a single task of designing a common strategy for data security is the most pressing.
It is the conclusion of this group that one of the positions funded by e@LU should be tasked with an evaluation of data security at the university. Within this task, it is envisioned that this person would perform an inventory of existing facilities and practices throughout the university, and assess how these practices correspond to our legal obligations. As a part of this inventory, we also believe that the person employed should serve as a key person in identifying research similarities and potential synergies – a subtask which will hopefully lead to potential collaboration across disciplines and faculties. Given this subtask, the person selected must have a good insight into the research process.
As a long-term follow-up, we envision that this person could be the seed for a new center aimed at supporting and facilitating research using personally identifiable data at the university.
Applikationsexperter: Gustav Öberg
Projektledare: Kirk Scott
Computer simulations are used to solve dynamical equations in a wide range of research areas within Lund University. A common theme for such simulations is the need to develop and run codes with efficient numerical algorithms on supercomputing systems with a large number of cores. Examples of research areas that obtain data from simulations on supercomputers are:
Climate Research. Within MERGE – a strategic research area within Earth system and climate modelling – climate models are run on supercomputers. The goal of MERGE is to establish a mul- tidisciplinary climate/Earth System modelling collaboration on climate/terrestrial biosphere interac- tions (including vegetation, land use, biogeochemical cycles and aerosols). As an example of the code optimisation challenges faced within MERGE, RCA-GUESS is a climate code that is widely used for modelling the evolution of the Earth’s climate. Porting the climate code RCA-GUESS to a new system, such as the Alarik cluster in Lund, requires substantial work to be put into optimisation. Proper adaptation of the code can speed up the calculations substantially, but qualified manpower is hard to find for undertaking such optimisation. Key Senior Staff in Climate Research: Dr. Michael Mischurow, Prof. Ben Smith, Prof. Markku Rummukainen
Theoretical astrophysics. The formation and orbital evolution of exoplanets is a major research theme in the Theoretical Astrophysics research environment at Lund Observatory. The equations for gas dynamics and gravitational N-body dynamics are solved on codes running on supercomputing centres in Sweden (through SNIC) and in the rest of Europe (through PRACE). Adapting the codes to run on an increasing number of processors requires continued development and maintenance of the numerical algorithms, e.g. for Fast Fourier Transforms. An important future research priority is to develop a code combining gravitational N-body dynamics with the hydrodynamics of gas and embedded dust and pebbles. Combining several codes in a parallel computing environment is a major challenge that would benefit strongly from the assistance of a dedicated application expert in code optimisation. Key Senior Staff in Theoretical Astrophysics: Dr. Anders Johansen, Prof. Melvyn B. Davies, Dr. Ross Church
Numerical Analysis. Research in Numerical Analysis focuses on the solution of nonlinear partial differential equations. Such equations are commonly occurring in physical processes found in engi- neering, finance, medicine and science. Two recent examples where the group have designed such schemes are optimal control of heating processes and simulations of corrosion crack formations. It is typically derived mathematically that the schemes converge with a high rate and preserve the critical structures of the original problem, e.g., energy, mass and positivity. However, to actually claim that they are scalable and efficient for large-scale applications requires the implementation and testing in a massively parallel environment, e.g., Lunarc. Such testing could be done by an applications expert within code optimisation to support the fundamental analysis done within the group. Key Senior Staff in Numerical Analysis: Dr. Eskil Hansen, Dr. Philipp Birken, Prof. Gustaf Söderlind
För att hjälpa dessa grupper sattes en liten grupp ihop med deltagare från dessa grupper för att diskutera kodoptimeringsproblem tillsammans med experter från Lunarc.
Applikationsexperter: Joachim Hein. Under 2017 kommer det till en extra expert inom området.
Projektledare: Jonas Lindemann