What is UPSCALE?
UPSCALE is a ran a set of ensembles of HadGEM3 GA3.0 atmosphere only climate simulations at resolutions from standard climate (N96, 130km) to current weather forecasting (N512, 25km). These were run at multiple High Performance Computing sites: PRADCE (HERMIT@HLRS), NERC (HECToR) and JWCRP (MONSooN). The goal of UPSCALE is to allow scientists to evaluate the value of resolution in climate modelling.
Why did the UPSCALE project choose to use JASMIN as a platform for storage and processing?
JASMIN was the only UK facility that could receive the whole UPSCALE data set (400 TB) at the required data rate (up to 4-5TB/day). In the absence of JASMIN the project would probably have been constrained by data transfer rates, seriously impacting on the achievements. UPSCALE is a project run by the High Resolution Climate Modelling (HRCM) collaboration between the Met Office and the University of Reading so a facility on which all participants can work/access the data is important.
What resources did UPSCALE make use of on JASMIN?
As an early adopter of the JASMIN Platform UPSCALE was able to act as a trial user for a number of resources:
- Virtual machines (one for Met Office and one for NCAS staff) tailored for use by UPSCALE.
- 400Tb of high-performance storage with 200Tb of backup
- High-bandwidth networks (allowing GridFTP and rsync transfers to/from JASMIN)
- Access to the LOTUS compute cluster for running large post-processing and analysis jobs
How long with UPSCALE make use of JASMIN?
The modelling runs were completed in 2012 but the post-processing and storage requirements stretch well into the future. At present there are 28 separate groups/individuals using UPSCALE results in their own projects. We expect this to continue for a number of years
A lead UPSCALE scientist said,
“The speed at which we can analyse the data is revolutionary. Elsewhere we are continually constrained by storage space and can have limited processing power attached to any storage we have. A whole host of analyses have been performed to date (see list of projects using UPSCALE data linked below). While the concept of working with the lotus cluster for scientific analysis is very new to us, the sheer power of being able to submit thousands of tasks to analyse separate parts (different runs/time periods) of the data set is a massive benefit of working with J. I believe that this facility is a demonstration of how scientific analysis in data-heavy fields, such as weather & climate science is going to proceed in the future. We would never have been able to store, nor analyse, that volume of data without the existence of the JASMIN service."
Relevant UPSCALE publications
Demory et al. Clim. Dyn. (2014) http://doi.org/r6b
Mizielinski et al, GMDD (2014) http://doi.org/r5v,
Roberts et al, J. Clim. “Tropical cyclones in the UPSCALE ensemble of high resolution global climate models” (submitted),
Vidale et al, UPSCALE science paper (in preparation)