Minutes of Face-to-face meeting 2004-12-2/3

The first face-to-face meeting of the Network was held at ESO, Garching, Dec. 2-3, 2004. The minutes of the meeting are give below.

Participants: KlausBanse, PascalBallester (part), VadimBurwitz, NicolaCaon, DavidGiaretta, PrebenGrosbol, PeterLinde, FabioPasian, DanielPonz, PhilippePrugniel, KlausReinsch, ThijsVanDerHulst, BillCotton (part/video), BiancaGarilli, JanneIgnatius, PalleMoller (part), KariVasko, DougTody

Thursday, Dec. 2, 2004 (Chair: PrebenGrosbol)

13:00-13:15 Welcome

• After a short welcome, the agenda was adopted with minor changes.
  • The first discussions on requirements and architecture were moved forward so at BillCotton could participate through the video link from NRAO.
  • The discussion on the SAMPO (ESO/Finnish) project was moved to the first day since JanneIgnatius and KariVasko had to leave around noon the last day.

13:15-13:45 Vision statement and Scope

• The Vision Statement was shortly discussed with the following considerations:
  • It was agreed that multiwavelength analysis should be explicitly be mentioned (DT)
  • DG suggested that 'discovery' should be stated as a driving factor
• The 10 individual scope items were then reviewed and agreed on. The following issues came up in the discussion:
  • on scope-2: There were general agreement that the environment should be bassed on free, open source concepts (ref. GPL) as far as possible. On the other hand, it should be relative easy to use commercial products when needed. P.Quinn shortly joint the meeting and confirmed the general policy to aim for free, open software.
  • on scope-3: DP and BG recently add a page on possible areas where a common environment not would be feasible. DT mentioned that the proposed architecture would solve some of these issues by leting an Integrator package only the required parts of the environment for a specific purpose.
  • PL mentioned that most of the items were not unique to astronomy. This is generally true except for e.g. Scope-5. For many parts like Scope-6, Scope-9 and Scope-10 one could involve e.g. computer science.

13:45-14:15 User scenarios

• The 5 user scenarios on the TWiki were briefly mentioned. As the work on architecture and design continues other scenarios may be needed such as:
  • A System Integrator scenario to show how different pieces of the system can be put together for a specific purpose (DT)
  • How a user doing calibrations would utilize it (DP)
  • The usage by Instrument engineers (DP)
  • Using simulators and theoretical models (PF)
  • Applying the system to pipeline reductions (DT,DP)

14:15-15:30 Requirements

• First the top level division was discussed. It was felt that the last item 'Visualization and testing' should be split into two separate sections.
• The discussion on the individual point started with category 1 on 'Installation and running'. A revised version will be places on the TWiki before the next phone meeting. The main points brought forward were:
  • Interactive usage of the system is important (KR)
  • The usage of may/must/shall should be carefully checked (PL)
  • The issue of Windows compatibility was mentions (PL) but the current wording was found satisfactory
  • Desktop usage must be explicitly be mentioned including control of updates and local data (DT,PG)
  • The term 'normal astronomers' should be replaced by astronomers without special computer science background (PL)
  • Software versioning must be supported (KR)
  • Standards for errors and logging should be mentioned (TvdH^?)
  • The relative importance of different requirements should be added
  • Testing of full system and/or individual task should be considered.
• Due to time limitation, the discussion was stopped and resumed next day

15:30-15:45 Coffee break

15:45-17:00 Architecture

• Since BC had to leave at 16:00, he mentioned the importance of having the architecture both deal with
  • light-weight, interactive reductions, and
  • heavy batch-type processing.
• After this, DT summarized the document on ArchitectureAndFrameworks as it only had been available for a few days. The following points were raised:
  • Whether different components could share memory or not. If they were deployed in the same container it would be possible although this normally not would be the case.
  • Where marshaling of parameters and data passed between components would be done e.g. by container or application layer
  • PG brought up the case where a user would like to interact fast with a model via a GUI. In this case, one would bind the GUI to the container used for the component implementing the model and thereby ensure a fast interaction.

17:00-17:30 *VO and Grid interface:

• The general relation and interface to both VO and Grid were discussed. It was clear at an exact understanding of what is included within VO/Grid compared with what we are doing is of high importance to ensure that the two areas are complementary and do not duplicate effort unnecessarily.
  • DT outlined the interface to VO as described in the architectural concept, that is it can both provide services to VO, by publishing applications as Web services, and use remote VO resources, by providing a VO client interface.
  • DG emphasized that not only VOTables could be used but also virtual tables (for accessing very large tables).
  • Issues on how to publish services to VO and data models remain (DT)
• The interface to Grid is clearer as it is mainly a service to be used by the Execution level framework of the architecture and therefore will not be seen by standard users.
• Exchange of information and compatibility with groups working on Grid Workflows (e.g. AstroGrid) should be pursued.

17:30-18:00 SAMPO (ESO/Finnish project)

• JI described the state of the SAMPO project:
  • It will officially start on 2005-01-01
  • Our OPTICON working group is mentioned in the ESO Council document
  • There will be a kick0off meeting in Jan. 2005
  • Six people have been hired for the project which will last for 3 years
  • The main objectives are to proto-type a future environment for data analysis and estimate the cost of a full production system
• KV went through the profiles of the project members

Friday, Dec. 3, 2004 (Chair: PrebenGrosbol)

09:00-09:15 Split into parallel sessions

• The working group separated into two groups and continued the discussions

09:15-10:30 Requirements (Chair: PrebenGrosbol)

• The discussion on individual requirements was continued:
  • Checks on parameters (2.2) should be considered (FP)
  • A general requirement on using existing standards whenever possible should be added (PL)
  • There a long discussion on if one should recommend a single scripting language (2.3) or allow many. It was agreed that there are practical advantages to using a single scripting language for most development initially, but that the Execution framework should permit any scripting language to be used.
  • Reference to comments and re-execution should be split in (2.6)
  • Quality flags should be added to (3.1/3.9) (FP)
  • Organization of data is important (3.5) (KR)
  • Explicit data types should be moved to level 3 in (3.6)
  • The wording in 3.8 should be revised
  • History records of how data were processed should be associated to the data themselves
  • It should also be possible to define where new data are stored
  • Explicit data formats (4.7) should be moved to level 3
  • Access to reference data must be possible (DP)
  • It should be easy to do simple tasks
  • Some mentioning of performance should be made

09:15-10:30 Architecture (Chair: DavidGiaretta)

• The architectural concept was discussed in more detail and a block diagram of a possible system produced to clarify some details of the architecture. The discussion will be summarized on TWiki pages before the next phone meeting. The architecture document was updated following the meeting to reflect these discussions (DT).

10:30-10:45 Coffee break

11:15-12:30 Summary of parallel sessions

• The two groups re-joined and the chairs of the parallel sessions summarized the discussions

11:15-12:30 Framework and interfaces

• The discussion on VO was continued
  • There are several NVO/IVOA meetings in the following weeks and we should make sure that our considerations are known (DT)
  • Scenarios for Data Access Layer services should be considered (DT)
• We should verify what Grid actually will be able to provide; Grid technology is still immature
• Interactions between our application layer and Grid workflows were mentioned

12:30-13:30 Lunch

13:30-13:45 Summary

• BillCotton joined the meeting through video link from NRAO. A short summary of the morning discussions was given.

13:45-14:00 Main issuses for next 6-9 months

• After the current level 2 requirements are consolidated, they have to be detailed with level 3 requirements
• Different groups (e.g. SAMPO, Starlink) who intend to contribute to proto-typing parts of the architecture have to be coordinated so that most important issues are exposed and too much duplication is avoided.
• A model for parameter passing between tasks have to be detailed
• Assuming that there is general agreement on the architecture (by end-January 2005), a high-level design must be started.
  • Major parts or subsystems have to be identified
  • All required interface must be listed
  • A minimum container with associated interface must be outlined
• A list of important features and technologies which should to investigated by proto-typing
• A proposal for a suitable demonstration of proto-types of simple

14:00-15:00 Conclusions

• The two main objectives of the meeting were achieved, namely:
  1. High-level requirements were formulated and discussed. Although minor revisions and re-wordings are expected, there was broad agreement on them. This opens the way for detailing them in the coming 6 month period.
  2. An architectural concept was proposed and discussed in detail. All participants supported it fully as a sound basis for a high-level design. It is important that all development groups in the individual countries can study the proposal and give comments. All participants were asked to approach their communities and encourage them to give their comments by end-January 2005 so that considerations on high-level designs can start shortly thereafter.
• It is expected that significant progress in both high-level design and detailed requirements (level 3) can be made during the spring of 2005. Actually it is required in order to provide input to projects like SAMPO, ALMA etc. This would suggest that a face-to-face meeting in the time frame of June 2005 would be fruitful. A decision and tentative dates should be discussed at the Jan. phone meeting. Due to cost, it is likely to have it at ESO.
• It is expected that a face-to-face meeting will be held just before the ADASS 2005 near Madrid Oct.3-5. as it is expected that there will be demonstrations of proto-types of basic architectural features of the environment.

Assigned to	Due date	Description	State	Notify
DavidGiaretta	Mon, 20 Dec 2004	make TWiki page describing architecture discussions including block diagram Done at meeting	closed		edit
PrebenGrosbol	Mon, 20 Dec 2004	revise requirements according to discussions	closed		edit
PrebenGrosbol	Mon, 10 Jan 2005	contact AVO to discuss our foreseen interface to VO	closed		edit
DougTody	Mon, 20 Dec 2004	contact NVO and IVOA to discuss our foreseen interface to VO	closed		edit
PrebenGrosbol	Tue, 18 Jan 2005	collect comments on architecture	closed		edit
PrebenGrosbol	Tue, 18 Jan 2005	contact Japan to get possible Japanese associated member	closed		edit
PrebenGrosbol	Mon, 10 Jan 2005	inform SAMPO on our architecture and requirement discussions	closed		edit

-- PrebenGrosbol - 14 Dec 2004
to top