ENGAGE

For those of you not involved in previous discussions - OMII and the ENGAGE project may fund the GUI development part of our earlier proposal to OMII to request support for open-sourcing GENIE. Some other aspects of this may also go ahead with seperate OMII funding, but the ENGAGE project is coming to it's end so getting a proposal together for the GUI work is the most urgent part. Steve is Steve Brewer, who I expect most of you know. He suggested gettign the ball rolling in the most expedient way possible on exploring possible funding options - in an informal email discussion. My slightly incoherent ramblings start here:

Dear Steve,

Thanks for an informative discussion on the phone earlier. I'm copying in the people who have been party to the last few emails, in the hope that they might chime in with useful info whenever they can, also Rupert Ford and Graham Riley from Manchester who have been involved in BFG2 (http://intranet.cs.man.ac.uk/cnc/projects/bfg.php, http://source.ggy.bris.ac.uk/wiki/GENIE_BFG) work with GENIE, which may integrate nicely as part of the work we put into the proposal. Have also copied in Andrew Price (SOTON e-science) who is a long-standing GENIE collaborator and whose input and funded involvement we would also welcome.

GENIE (Grid-ENabled Integrated Earth-system model) is a modular, flexible Earth-system model of intermediate complexity (EMIC), which is used extensively within the UK and internationally in Earth-system science. One important aspect of GENIE which makes it unique (or nearly so?) amongst EMICS is the modularity which currently allows experienced users to plug in new or different models of a particular Earth-system component or in theory even to use a different 'glue' code for sticking all the components together.

To enable less experienced users to access these aspects of the model, and also to facilitate ease of running the model in its 'standard' configuration, a user interface would be extremely valuable. As well as saving end-user time in manipulating XML config files and running shell scripts, for example, it will also broaden the community of researchers able to use GENIE to answer their research questions as less technical computing knowledge will be necessary. The facility to increase or decrease the complexity of options available in the GUI to produce teaching, or targeted research versions would aldo be valuable.

Steve - you suggested that the best way to go about formulating a good proposal for ENGAGE was to have an 'ideal world' goal for Mid-May '09 - the deadline for completion of work funded under the ENGAGE project - from which we can work back to person-months, staff availability etc. SO here goes!

I envisage a beta release of a 'scaleable' GUI, which can be easily configured to work at any required level of complexity with the underlying code (so for example, parameters would be at their default values if not selected for control by the GUI). The GUI would also allow configuration and selection/deselection of a series of currently available modules (different atmospheres, land and ocean models for example) and the ability for the model to run with the current standard 'GENIE-main' model glue but also with the BFG2 option. This would require effort from the Manchester team and would realise excellent the preliminary work already undertaken by their group. Part of this would be to BFG-wrap a number (1 or 2?) of modules not currently available to BFG-GENIE - although it would not be possible to comprehensively cover all current (non-BFG)GENIE-available models due to time constraints.

We would need to ensure that this GUI was future-proof - we already have a coherent XML configuration system, so it simply has to be able to dynamically expand to cope with more modules and parameter options. New model glue codes and approaches would presumably require new GUI code unavoidably.

The GUI would be released along with the GENIE source and presumably be included in the GENIE subversion repository for download.

Platform options:

A preliminary effort has produced 'Aladdin' which is a basic GUI built in Matlab, currently running only with the old non-XML configuration of GENIE. This was intended for release within the GENIELab Matlab Toolbox (http://source.ggy.bris.ac.uk/wiki/GENIE_GENIELab) - which includes Toolboxes for running GENIE ensemble experiments on the National Grid Service facilities, and integrates the OMII-supported GEODISE toolbox for optimisation and tuning. For this reason, the Matlab approach is attractive as Aladdin or its decendent will integrate with the Grid-Enables aspect of GENIE. Although Matlab is closed-source, commercial software, the Toolboxes can be open-source (i think) and freely available and can be run as an executables using freely available Matlab libraries (Sudipta - is this correct?) for those without access to the full software. GENIELab has been realised through the hard work of Andy Prcie at Soton, and Aladdin through the similarly heroic efforts of Sudipta Goswami at UEA. Sudipta is probably best placed to start a step-by-step plan of how the existing Aladdin might be developed into a beta release in 6 months time - what work needs doing etc.

Alternatively a platform-independent Java GUI is a possible option or possibly a web-based one (which would have to run locally as well as on remote servers, so would need to ship with its own web-server). My feeling is that moving away from the existing base of work with Matlab is a 'phase 2' goal and that we should build on the work Sudipta has already done for now, although this does depend on the specific requirements of the ENGAGE funding wrt open-source / commercial software and also the manpower available - Steve to confirm.

General comments

This is all very ambitios for what would most likely be a 6 month project, and I'm sure it will need simplifying somewhat. However this depends a bit on how / if we can split the work up into discrete chunks which can be tackled at different institutions. Someone will of course have to perform a technical project management / overview role and this might be best acheived by Gethin if he's happy with that.

BFG2 integration is valuable because other models and EMIC components are being wrapped in a compatible way, which would mean GENIE could integrate e.g. Hadley CENTRE atmosphere models for 'free'. This would mean that the GUI would have to be extra-clever to cope with reading the information for a newly acquired BFG-wrapped model to get the parameters which can be configured, then providing options to configure them. Not impossible, as BFG presumably needs to know about all of these anyway. Such non core GENIE modules would raise broader problems as well, such as integration with XML configs so this is a future goal which would not be realised within the short timeframe of this possible funding.

Most of the people currently involved have limited flexibility in their workloads and are fully funded over the coming months. However, it may be that they can juggle contracts around to fit some of this work in - people, can you confirm what availability you could muster between 1/12/08 and 15/05/09? At most I'd guess we could possibly get 1 full-time persons's worth of time out of the recipients of this email over the fundable period. Steve mentioned that OMII have a good track record in 'matchmaking' projects with people with the necessary skills so we might be able to bring some fresh blood in. Maybe Manchester have someone looking for a job with them who might work well on the BFG side of GUI intergration and module-wrapping?

In terms of short timescales vs lots of people and communications problems, the current GENIE team have existing strong working relationships and this would greatly strengthen the likely success of the project.

Apologies all for the rather rambling nature of all this. Getting it bashed out so that some discussion can get underway seemed important. I feel a bit like I've just volunteered you all for something without asking you first, so if you have better things to do with your lives / jobs over the next few months don't be afraid to decline this possibility.

Look forward to the discussion getting started

Best regards to all

Martin