theme | transition | width | height | maxScale | controls | progress | slideNumber | enableMenu | css | defaultTemplate | |
---|---|---|---|---|---|---|---|---|---|---|---|
css/robot-lung.css |
none |
1920 |
1080 |
4 |
true |
true |
true |
true |
|
[[tpl-ukaea-slide]] |
- Motivation
- Background
- Schedule
- This workshop is aiming to give you an insight into what UDA is, how it works and how it can be used in IMAS to map experimental data
- Experimental data in IMAS format is needed to validate ITER software and can help creating multi-machine databases for AI/ML
- There will be hands-on session to cover how to set up a UDA server, develop JSON mappings and create a mapping plugin for your device
- At the end of this workshop you should have the knowledge of the basic building blocks needed to develop mappings for any machine
- Future training and documentation can build on this knowledge for machine specific cases
- UDA — Universal Data Access (formerly known as IDAM) is a client-server technology originally created for the MAST experiment
- Designed to separate data access and format from transferring that data across the network to a user
- Dynamically loaded plugin architecture allows for data access plugins to be written without changing UDA core code
- Consists C/C++ client library which talks to a C/C++ server running behind xinetd
- Several higher level language wrappers available, the most popular of which is pyuda — the Python wrapper
- UDA has been used for several years to provide mapped data access in IMAS
- Previous efforts have built upon existing work (EXP2ITM) and relied on somewhat shaky infrastructure stacks to enable this
- With release of AL5 and introduction of IMAS URIs UDA can now be used to read IMAS data files remotely
- Building on the improved UDA backend and plugins for AL5 the JSON mapping plugin is designed to provide a better way to write and manage the mappings for multiple machines
Name | Type | Description |
---|---|---|
UDA Overview | Presentation | Detailing how UDA works to provide context for it's use in IMAS |
IMAS UDA | Presentation | Detailing how UDA is used in IMAS remote data access and experimental data mapping |
UDA Installation | Presentation | Covering installing and configuring a UDA server, plugin installation, and basic UDA usage |
Name | Type | Description |
---|---|---|
Data Mappings | Presentation | Overview of data mapping with UDA and the new JSON plugin |
Data Mappings | Hands-On | Developing a new mapping file |
MASTU Mapping | Presentation | Data mapping with UDA for MAST-U |
Data Access Plugin | Hands-On | Developing a new machine specific plugin |
What we assume from attendees:
- Basic knowledge on IMAS and particularly its data model
- Interest in using UDA to map experimental data into IMAS
- Some basic C/C++ & Python knowledge
- Familiarity with Linux systems and command line usage
- Please ask questions during the presentations — any that can't be answered quickly can be discussed later or noted for expansion upon in future training materials
- All hands-on work is designed to be done on the provided ITER training VMs
- The material should work on other machines but may require the installation of some prerequisites and we probably can't support this during the workshop
- Adam & I will be available for on-site questions and will try and answer on-line questions when we can. Stephen will be provided on-line support to provide additional support.
- Please use Google docs for all questions and discussions
- WiFi -> choose "Guest wireless" and then "Login with your ITER account"
- How-to connect to SDCC: https://confluence.iter.org/display/IMP/ITER+Computing+Cluster
- Regular ITER bus lines to Aix (A, B, C) are departing at 17:45 from the parking lot in entrance C
- Work site visit Tuesday afternoon at 15:45
- Workshop dinner at the restaurant L'Orangerie in Hotel Aquabella (Aix-en-Provence) at 19:30
- Updated agenda and links to Google docs and more information can be found on the Inidico page https://indico.iter.org/event/81/