Title: ESO Reflex A Graphical Workflow Engine for Data Reduction Richard Hook Euro VO Data Centres Alliance
1ESO ReflexA Graphical Workflow Engine for Data
ReductionRichard HookEuro VO Data Centres
Alliance Theory Grid Workshop,Garching, April
2008
2Overview
- The Sampo Project
- The ESO data reduction context
- ESO Reflex
- Examples
- Future plans
- Apology this talk is not really about grid
activities (and definitely not about theory)
3Sampo overview
- As part of Finlands joining fee for ESO a
contribution in kind of computer scientist
staff was made available and named Sampo. - Sampo started in January 2005 and ended in
January 2008 - The aim of the project was to assess the
requirements for ESO data reduction and analysis
software infrastructure in the medium term and
perform a series of pilot projects to assess
different options and produce useful tools - The project was managed at ESO Garching with the
team based in Helsinki, Finland
4The Sampo priorities
- Enabling and facilitating science-grade reduction
of ESO data from the La Silla Paranal Observatory
was identified as its primary goal of the Sampo
project - Long-standing request from the community
- Input from Instrument Scientists in Garching and
Chile - Sampo SAC
- Sampo has concentrated on developing ESO Reflex,
a graphical user interface to run ESO data
reduction recipes - Other sub-projects conducted by Sampo
- PyMidas a Python interface to Midas
- VODA a pilot project addressing the integration
of data analysis environments and the Virtual
Observatory (cancelled)
5The data reduction challenge
- ESO cannot reduce all of the data its telescopes
and instruments produce to a level where their
full scientific potential will be exploited - The responsibility for the quality of the
scientific reduction of the data can only rest
with the individual users - The users are, then, faced with the challenge of
a timely and accurate data reduction - Proprietary and/or archival data from several
instruments often need to be combined - As volume and complexity of data increase, a
detailed knowledge of the instrument, data format
and header content is essential to fully exploit
the data - Science-quality pipelines are increasingly needed
also for Quality Control - General-purpose tools like IRAF and ESO-MIDAS are
inadequate for the task and instrument specific
software, implementing carefully tuned algorithms
is essential
6Data reduction by individual users the context
- ESO provides pipeline recipes for all VLT/VLTI
instruments - They remove the instrumental signature and are
used for quality control at ESO and distributed
to the community - In some cases the data products are adequate for
scientific analysis, but this is generally not
yet the case - Offline tools (Gasgano and EsoRex) are available
to call the pipelines, but lack some of the
functionalities needed by the community - AND
- Many older general purpose reduction and analysis
systems remain in wide use (MIDAS/IRAF etc.) as
they contain valuable algorithms - Many instrument-specific packages have been
developed in the community (e.g., Euro3D tools,
VIPGI etc) - Greater use will be made of remote data resources
and the Virtual Observatory (VO)
7Introducing ESO Reflex a graphical data
reduction environment
- A data reduction system for the end user
requires - Modular recipes to provide access to intermediate
products - Interactive tools, defined or customized by the
user, to analyze intermediate and final data
products - A user-friendly, intuitive and flexible interface
- The ESO Reflex tool, addresses the interface
issue, with a focus on the use case of ESO data - Dedicated invoker for CPL-based recipes
- General invoker for Python scripts (hence PyRAF
PyMidas) - General invoker for IDL scripts
- Invoker for the command line
- Many other Taverna features for free
8ESO Reflex look and feel
ESO Reflex is based on Taverna, a popular open
source Java workflow engine
9Main features of ESO Reflex (over and above what
Taverna offers)
- In interactive mode the user can make changes to
input data and parameters during execution - Errors on during recipe execution are detected by
ESO Reflex and appropriate action can be taken - (Some) flow control looping, skipping,
conditional statements, etc. - Parallel execution full advantage on
multi-processor or multi-core machines - Customisability workflows are easily modified,
Python and IDL interfaces, system commands can be
invoked, easy access to VO and other web
services - FITS file handling (using Gasgano code) data
organization, tagging, selection
10ESO Reflex in action FORS2 MXU
11ESO Reflex in action a FORS Calibration
workflow.
12ESO Reflex and VO services
- Reflex/Taverna works very well with distributed
resources and web services - Reflex also supports the PLASTIC protocol for
passing information between client-side VO tools - As a small test project a workflow has been
developed that finds VO image data, passes it to
a remote processing server and uses PLASTIC tools
running locally - Reduction of local ESO data and access to remote
VO facilities within same environment
13(No Transcript)
14Potential shortcomings of ESO Reflex
- Yet another tool to learn, support etc.
- Not as powerful as a script
- Flow control limited (loops not easy)
- Not suitable for very complex workflows
-
15The bottom line on ESO Reflex
- The Sampo project has enabled ESO to explore
options for future data reduction systems - The ESO Reflex model is cost effective because it
capitalises on ESOs investment in CPL-based
algorithms - Offering ESO pipelines to the community in a more
flexible way - Stimulating more scientific feedback from the
community and, therefore, steering algorithm
development towards more science-grade
applications - It exploits existing systems to provide those
facilities that CPL is not designed to offer
(e.g., graphics, interactivity, etc.) - It avoids the need to develop from scratch a
multi-purpose, and expensive, full-fledged data
reduction environment
16The future of ESO Reflex and pipelines
- The Sampo project itself ended in January 2008
- ESO has examined the outcome of Sampo and
concluded that - ESO Reflex is a viable way of addressing some of
the mid-term needs of the user community - The continued development of science-grade
CPL-based data reduction recipes is vital - The development of interactive tools is also
needed
17Status and plans
- ESO Reflex V1.1 is currently available as a
beta-test version on request - Current version is built on Taverna 1.5 -
conversion to Taverna 2.0 will be needed, and
involves significant changes - API changes
- Use of Maven
- Some UI refactoring
- A public release is planned for
late-2008/early-2009