Landslides, Couches and Particle Physicists
Short version:
Researchers from Bristol University have developed pioneering landslide modelling software which is used by governments in developing countries for disaster prediction and response planning. The software, input data and simulation results are currently managed in an ad-hoc manner and users typically lack the resources required to handle this effectively. We describe a collaboration between Bristol University's Geographical Sciences department and Particle Physics group, where techniques developed for the management and analysis of petabyte-scale datasets generated by the Large Hadron Collider are being applied to the problem of effective landslide modelling and risk reduction. A framework for hosting simulation software, based on the BigCouch fork of Apache CouchDB and the DIRAC workload management system, is being developed which provides features for managing complex input and output data and supports the parallel execution of simulations across a number of pluggable compute backends.
---
Long version:
Landslides caused by tropical storms can be catastrophic events that disproportionally affect developing countries with inadequate disaster management resources. Researchers from Bristol University have developed pioneering landslide simulation software, used by academic researchers, industry and governments to simulate slope behaviour under various storm conditions. The simulation results provide a cost benefit analysis that is essential for planning effective preventative measures and response strategies.
Currently the software, input data and simulation results are managed in an ad-hoc manner and making them available to all stakeholders is difficult. Simulations are run in Bristol where input data provided by field agents is validated by experts and simulation results are stored. While data volumes are relatively small, around 10TB/year for approximately 100 slopes, governments of developing countries typically lack the resources required for effective management of this data. In countries like St Lucia, a tropical storm is likely to cause significant damage to telecommunications equipment, removing access to the data and computing infrastructure right when it is urgently needed.
We describe a collaboration between Bristol University's Geographical Sciences department and Particle Physics group, where techniques developed for the management and analysis of petabyte-scale datasets generated by the Large Hadron Collider are being applied to the problem of effective landslide modelling and risk reduction. A framework for hosting simulation software is being developed which provides features for managing complex input and output data and supports the parallel execution of simulations across a number of pluggable compute backends.
The framework architecture leverages the DIRAC workload management system, developed by the LHCb collaboration based at CERN, to provide access to WLCG/EGEE grid infrastructure and Amazon EC2. The BigCouch fork of the Apache CouchDB document database provides the necessary central management of simulation data, software releases and simulation requests, and standalone CouchDB instances provide field engineers and government planners with a "personal" copy of the relevant data.
