CESD SAGES Scottish Alliance for Geoscience, Environment & Society SAGES and ECDF Simon Tett & Mike Mineter Centre For Earth System Dynamics Simon.tett@ed.ac.uk/m.mineter@ed.ac.uk 1 Overview CESD • What, why, who of SAGES and CESD • Initial steps • Possibilities for the future 2 What is SAGES? CESD • Multi-institution, multi-disciplinary collaboration at the forefront of earth and environmental research. • Goal: link discoveries in earth and environmental science with the needs of society. – Pool expertise and resources – Address complex issues outside individual group’s scope 3 Centre for Earth System Dynamics: What? CESD • “The modelling skills and range of expertise in each Theme will be integrated through the creation of a virtual centre for Earth system dynamics”. • CESD will: – carry out modelling of the Earth System – support research undertaken by SAGES partners 4 Centre for Earth System Dynamics: Who? CESD • CESD has to have defined membership, not least because it is opening resources and services to users in many institutions • Who is in CESD? – SAGES members, via a lightweight application to CESD Director, Prof Simon Tett. – CESD members have been invited to register their postdocs and PhDs to use CESD services 5 CESD CESD requirements and initial steps • Support many very diverse groups in SAGES encouraging collaboration • Some requirements for HPC – different groups, different patterns of use – intermittent or consistent - and TBytes of data for storage and sharing • Seek common core of information technologies, services and infrastructures – Relatively easy to adopt – Confident of persistence beyond SAGES’ current phase • So – share one cluster as a SAGES-wide platform for collaboration, computation and storage 6 CESD services CESD Services to support collaboration SAGES groups in different locations, organisations, and research domains seek to work together Each group has its own data, programs, models. Shared infrastructure for computation and data storage and sharing Some of these need to be shared and coupled across groups. 7 Why use ECDF? CESD • High performance resources for individual and collaborative computation / data storage • Easy to explore before committing • Fast start-up to doing work cf. buying own cluster • Able to support users from our SAGES partners • Important added benefits: – ECDF services – people and software – EASE authentication – Strengthens exchanges with initiatives in different subject domains – Could open options to share data/programs – via UK or EU grids 8 CESD Initial Roadmap CESD Support porting of other prioritised models 4 Establish ECDF use for all CESD partners 2 Membership Support use of ported models Initial model on ECDF: 4 3 HadCM3 climate model 1 Establish Wiki Authorisation 9 HadCM3 CESD • HadCM3 developed in mid-90’s and “operational” at Met Office from 1997 on T3E – Models atmosphere, land-surface, ocean and sea-ice. • Resolution is: 3.75x 2.5 degrees in the atmosphere and 1.25x1.25 degrees in the ocean. • In widespread use in UK community though mainly on national super-computers i.e. HPCX. • Our task: port and test model on ECDF • Model scales well up to about 32 processors on ECDF. • We get about 20 simulated years/real-time day on ECDF. • Estimated cost is £200/simulated century + storage costs 10 Using multiple processors CESD Uses domain decomposition – each subgrid is allocated to different processor • Sub-grids are processed concurrently • Processes exchange subgrid boundary data 11 HadCM3 for SAGES CESD • Ported HadCM3 to ECDF – Collaborating with NCAS – Necessary to engage with UK community via NCAS • We use it under license – owned by Met Office • Benefit from NCAS support, others configurations of the model,.. And we will contribute to that UKwide pool • CESD supports users of the model – Documentation (CESD Wiki) – “Hand-holding” while get started 12 SAGES > Σ(parts) CESD • CESD’s support for collaboration – Wiki: Exchange of ideas, know-how, metadata (about data and models…) – ECDF • Enables collaboration – controlled sharing of data, programs, codes • Encourages cooperation (different research domains) scope for common tools?? Based on some common data standards?? – Modelling Officer • Coupling of models – major objective – Loose or tight coupling…. 13 Modes of coupling: tight CESD E.g. better climate / ice simulations Model A Model B loop loop { do something else { do something Exchange data } Exchange data } 14 Modes of coupling: loose CESD Model A B consumes data from A: Either • While A is running • From archive loop { do something Write data } Model B loop { do something else Read data } 15 What next? CESD • Monitor effect of CESD – are we supporting new science? (including new ocean carbon cycle might be an early “yes”) • Need metadata service – PROVENANCE of models – Metadata about the data.. – Must • minimise extra steps for researcher • Something simple, easy to tailor for different models/data…. – Would be nice to adopt a University or national service for this!! 16 Summary CESD • Establishing a basis – CESD wiki: https://www.wiki.ed.ac.uk/display/cesdwiki/CESD+Home – Using ECDF in trial mode • Supported by EASE • Climate model – first CESD model en route ECDF • Beginning to buy time /storage on ECDF to meet CESD commitments to SAGES • • • • More models and support for users to follow Forming longer-term strategies as SAGES develops Seeking resonances with other projects Wanting University (National?) services for metadata and provenance SAGES web site: http://www.sages.ac.uk 17