Rudolf is head of computing services at ESRF.
Observations today done with sophisticated instruments – including synchrotrons. A synchrotron is a super microscope to examine condensed matter. Operate from the UV to the hard x-ray spectrum. Syncrhotron light – used because it has remarkable properties: brilliance (1,000 billion times brighter than a hospital x-ray tube).
The European light source is a 6 GeV source in Grenoble. Concentrates 10,000 researchers and engineers. Cooperation between 18 countries. Annual budget of ~80 m€.
Some applications – studying the structure of spider silk; medical applications, including angiography that gives better results than conventional hospital techniques; Geophsysics, sutdying samples which undergo extreme changes; Chemistry, how do catalytic processes function; Semiconductors; Paleontology, high res microtomography of fossils.
52 Beamlines, 6222 user visits in 2007. 15,308 eight hour shifts scheduled for experiments in 2007. > 1500 peer reviewed publications/year.
Over last 10 years the data volume has increased by a factor ~300. In 2007: 300TB, ~1*10E8 files.
Storage policy is only to keep 6 months of data, even for internal users. This is under heavy discussion.
French network infrastructure has not increased bandwidth in three years. That’s a problem
Network based on Extreme Network switches, storage on NAS systems, StorageTek tape for offline. Commodity clusters in teh data center, ~400 CPUs (totally insufficient).
ESRF Upgrade Programme – 290 m€ programme.
Single most productive facility producing protein structures for the Protein Databank.
New methods: nanobeams & raster scans. Looking to increase resolution by two orders of magnitude. Petabytes of data – how to carry away the data. Easy to imagine a PB per day in 10 years, contrasted to 15 PB per year for the LHC.
Two fundamental problems:
Latency diminish teh time to measure, store and analyze data.
Add functionality – new ways to measure, store, and analyze data. “Have to get our hands dirty with grid tools”
Looking desperately for 100 Gbps network.
I/O bottlenecks in research clusters is a big issue.
ESFRI position paper on digital repositories – lots of storage and access policies.
Oren Sreebny 