Astronomy Worspace Editor ------------------------- Anthony Rogers, Space Telescope Science Institute (rogers@stsci.edu) Norman Hill, Canadian Astronomy Data Centre (norm.hill@cdc.ca) Abstract: --------- o The VO has now completely standardized the data structures and methods to retreive astronomical data from diverse databases around the globe. This is a huge step forward and opens up numerous possiblities for improving research and discovery in Astronomy. However, the current web-based infrastructure to retreive and analyze astronomical data presents a new set of challenges to the VO: (1) Potential users around the globe may be unaware of all of the VO's capabilities and therefore never access it. (2) Users that are aware of the VO infrastructure may be overwhelmed with the current web-based tools to access it and quickly abandon it. (3) There does not seem to be an overall glue which binds the individual components of the VO together into an easy to use and reusable whole. Limitations of an existing Web-based research projects ------------------------------------------------------ o Locating data sources can be overwhelming to a new VO user. o Retreiving data from different client Web interface(s) at each institution can be time-consuming. o Sophisticated Web-based applications, in general, have limitations: + Embedded applets can be slow in loading and can not access data on the local hard drive. + Web based tools can suffer from poor feedback due to "reload, wait and flicker" syndrome. + The Web is inherintly "stateless", so application developers must always maintain of Session Level and User Level State. + Sophisticated work-flow type applications become clunky on the Web to both develop and use. (For example, try filing your taxes using the HR-Block web site. Now try filing it over multiple days and multiple browser restarts.) o Historically, many separate customized steps are required to complete a successful research project. This is by definition, a "Workflow". o Historically, there is no standard "Project Definition", just a collection of data files and custom scripts produced by the research team. + The project is not easily shared among colloborators, making it difficult to "farm out" pieces of the project and coordinate it's development. + The project is not available for critics. + The project components are not typically re-used outside of the original author(s). What is Needed -------------- o A common desktop "smart client" application that utilizes back-end VO Web Services with following charactersistics: + An easy to learn user interface to increase productivity and minimize frustration. + A Workflow "Project" Definition that can be shared among colleagues (and critics alike). + A development platform that promotes re-use of both standard VO tasks and customized user Analysis and Display tasks. + A centralized repository for all developed tasks which are easily loaded by others. + A rapid prototyping capability. + The ability to split a Workflow Project into independent, self-contained tasks to foster cooperative development on a project. + The ability to easily combine the independent tasks into a productive whole to produce results. What We Did ----------- o Searched the Web for existing an workflow editor. o Built new VO tasks within the workflow editor and linked them together to solve a "real world" project. o We used Colin Aspen's "real world" project idea: TwoColorDiagramPlotter described below. NOTES: o We looked at the commercial workflow product "Together" and quickly discarded it because it was too complicated and only provided documentation when purchased. (NOTE: This is a really bad Sales/Marketing Idea). o We decided on "Triana" (http://www.trianacode.org/): o Triana was easy to learn, written in Java (multi-platform), easily extended and best of all: Free! TwoColorDiagramPlotter Project Description (submitted by Colin Aspen) --------------------------------------------------------------------- It would be nice to have a service that looked at SDSS and 2MASS and from input filters and a region of sky produce a two color plot of the sources included. The plots could be totally SDSS (e.g. r'-i' vs. i'-z'), totally 2MASS (e.g. J-H vs. H-K) or a mixture (e.g. r'-i' vs. i'-K). Plot could use voplot in-line (?) This can be extended to color-magnitude diagrams also... Steps Involved -------------- o Downloaded Triana: o Built the following tasks in Triana (each written in Java): + ConeSearch (SDSS) : Searches SDSS for Catalog objects around specified RA, Dec and radius returning a VoTable. + ConeSearch (2MASS) : Searches SDSS for Catalog objects around specified RA, Dec and radius returning a VoTable. + ColumnAdder : Adds specified color columns to an existing VoTable. + Xmatch : Crossmatches 2 VoTables given a search radius and produces a single VoTable as the result. + VoPlot : Launches VoPlot and to plot the colors of the input VoTable. o Created the following Workflow "Graphs": (1) ConeSearch(SDSS)----->ColumnAdder------>VoPlot (2) ConeSearch(2MASS)---->ColumnAdder------>VoPlot ConeSarch(SDSS)------> (3) Xmatch----->ColumnAdder------>VoPlot ConeSarch(2MASS)-----> o NOTE: Tony built his workflow tasks using Triana under Windows XP. Norm built his workflow tasks using Triana under Linux. The final Workflow graphs were built on Triana under Windows XP. Results ------- o We were able to quickly build standard VO tasks in Triana. o We were able to drag/drop and link Triana tasks in the workflow editor to quickly produce real results. o We were able to launch external processes (not written in Java) within a Triana framwork. o We succesfully completed workflows (1) and (2) above. o We did not complete Workflow (3) successfully due to time limitations and annoying issues of running external processes under Windows. These can be resolved with time and persistance. Future Work ----------- o Cotinue to add the VO standard suite of Tasks in a Plug-N-Play fashion. o Continue to add other standard VoTable Analysis, Display and Plotting Tasks. o Colloborate with other astronomers on a large scale project to prove long term viability, collaboration and re-use. Conclusion ---------- o A workflow editor puts the "VO on the Desktop". o It is easy to adapt a workflow editor and insert standard VO Tasks. o Splitting up the workflow tasks among programmers increased the productivity. o It was trivial (and even "fun") integrating the tasks together at the end to visualize the results. o A Workflow Editor can provide a standard suite of VO tasks and custom tasks that can be used for: + Prototyping ideas quickly + Training Tool for the VO + A full blown astronomical research project split across multiple colloborators/programmers around the globe. + Rapid exploration of multiple data sets where a sequence of tasks are repeatedly executed with only small modifications between each execution.