European XFEL Seminar

Towards end-to-end data-management for large scale x-ray facilities

by Prof. Brian Vinter (HPC/UCPH)

Europe/Berlin
E1.173 (European XFEL)

E1.173

European XFEL

Campus Schenefeld
Description


 
Large scale scientific facilities, including x-ray facilities, face an
extreme growth in data from instruments. With x-ray instruments data grow
exponentially with the increased size of detectors, another exponential
factor from the frequency one many sample with and finally x-ray sources
are not robust enough that a large set of experiments can be automated,
bringing a large increase in the number of experiments an instrument can
perform in a session. Thus storing the data alone is a challenge.
The challenges are furthered from the fact that detector-size has grown to
a resolution where samples, at least tomograms, cannot fit in the memory
of a PC for data-analysis, and thus must be moved onto server-class
computers with sufficient memory to hold a raw-data sample and a processed
version as well. The increase in data-rate and number of experiments also
mean that running through all samples manually easily becomes unfeasible
and some means of batch processing must be introduced.
A final challenge is that users of x-ray facilities is widening and many
of the new users are not comfortable with data-analysis and need to work
with others in that part. This means that large communities, typically
geographically distributed, need to collaborate on these, very large,
datasets.
The talk will presents our ideas for an integrated solution to the above
problems, include status and plans, and also introduce - for discussion -
an idea for such an integrated system to help fight scientific fraud.
The work is supported in parts by H2020 European training network
MUMMERING.
  

Hosts: Robert Feidenhans'l and Hans Fangohr