Wanted: Storage of HEP data via key/value storage solutions(2 September 2015)
Interface the ROOT I/O to key/value stores.
The processing of the data acquired by the LHC experiments proved to be extremely demanding during the data-taking period featuring a centre of mass energy of 7 TeV. The requirements in terms of computing resources will increase during LHC Season 2 following upgrades to the accelerator and the detectors. However, the funding for the computing infrastructure will not scale accordingly.
These circumstances pose exciting challenges, for example in the areas of scientific software performance and data storage solutions: the quest for the evolution of the current, successful HEP software stack targeting the exploitation of new hardware architectures will continue. This R&D line relates to the experiments’ data I/O and aims to explore the interplay between the ROOT* tool-kit and novel data-storage solutions based on a key/value mechanism, for instance Seagate’s Kinetic technology.
ROOT is a C++ data visualisation, analysis and storage tool-kit extremely widely used in HEP. Its I/O subsystem is unique: it enables users to serialise on disk, both in a row-wise and column-wise fashion, any C++ data structure. This solution is powerful and versatile: hundreds of petabytes of LHC collisions data have been written in ROOT format.
However, presently, ROOT can write only on files. In the context of this work, explorations will be carried out with the goal of overcoming this limitation. During a first phase, the storage of specific objects, histograms and graphs, will be considered: these entities are indeed fundamental for the continuous data quality monitoring and software validation efforts of all particle physics experiments. Benchmarks based on real-life use cases will be performed against existing and well-established technologies, such as the DQM Gui of the CMS experiment.
During a second phase, investigations and prototyping will be carried out in order to replace the existing POSIX layer of the ROOT I/O with a set of more flexible interfaces that are compliant with a key-/value-storage approach. The work plan can be formulated in a rather modular way in order to be flexible with respect to academic time scales.