Following a needs assessment performed with BeamLine managers at the end of 2004 (functions, capacity, data storage times and so on), a call for tender was launched in mid 2005 for the supply of a suitable infrastructure. The main requirements were availability to allow work to continue even if there was an interruption on the dedicated storage network, reliability to be able to re-read an item of data under any circumstances, data security to ensure data can be accessed by any person having rights to do so and extendibility as BeamLines are progressively opened.

A solution based on the Active Circle’s distributed cellular storage concept was adopted at the end of 2005 and the first section (for phase 1 Beamlines) became operational in October 2006. The Active Circle system integrates data security, continuity of service and storage optimisation functions in a standard hardware environment.

Each BeamLine has a local storage point or cell, connected both to the BeamLine network and to the central data storage cells by dedicated fibre optic cables.

The Active Circle concept is described below. Each user accesses a circle representing a virtualized unlimited file system with transparent allocation and management of physical resources. The circle is composed of a set of servers called cells, which are linked together by an IP network without any type of hierarchy applied to the servers. Each cell is an entry point to the file system which continuously monitors the presence of other cells on the circle. This means that they share storage tasks with data being replicated and distributed to other cells, and react automatically to different events concerning them, i.e. incidents, addition or removal of cells, new hardware and so on. The set-up applied in our system is as follows:

•  one storage cell per BeamLine for the most recent data with the capacity of the cell adapted to requirements of the BeamLine.
• two cells in each of the two computer rooms (RGI1 in the central building and RGI2 at the centre of the synchrotron) running a set of high-performance DDN and NEC disks and a high-capacity GRAU tape library respectively. Data is replicated in both rooms and automatically migrated from the disks to the tapes applying a strategy specified for each BeamLine. Basically this entails keeping “recent” data on disk for 100 days and keeping “long-term” data on tape for one to five years according to the volume of data generated by the line. The magnetic tapes used will be able to contain between 400 GB and 1.6 TB of data, depending on type of level of compression.

Around 30 cells are operational on the beamlines and in the RGI computer rooms. As far as possible, experimental data files are stored in NeXus format on the local storage cell via the BeamLine control and data acquisition system. 

A fail-over mechanism is in place to respond to any malfunction on a BeamLine cell. When activated, this mechanism switches the service over to one of the cells in the RGI rooms automatically in a way which is transparent for users.

A function to save data in a standard (tar) format was added in 2008, enabling data to be externalized and recovered in the event of an incident (PRA).