S3/SimDB

Mon Feb 23 08:57:50 PST 2009

	Dear Miguel and Carlos,

	Allow me to share my experience about the possibility to use SimDB to  
develop theoretical services about micro-physics simulations. Note  
that, below, when I speak about micro-physics simulations, I mean  
simulations computing the structure of astrophysical objects with some  
micro-physics aspects : level excitation, lines, ....

	In the Euro-VO contect, we did a test of implementation of SimDB on  
PDR simulations (same kind of simulations than photo-ionization codes  
as CLOUDY). One of the objectives was to test the relevance of SimDM  
to describe such micro-physics simulations and to investigate the  
possibility to develop above the database some usefull services.

	My conclusions have been presented at Baltimore InterOp and are that  
even if SimDM does not permit a full description of such simulations  
(it is not its aim), it is possible to use it and it fullfills some of  
(?) the most important use-cases.

	I invite you to have a look on our prototype (the web-site is still  
in development and so, I require your indulgence) :
	http://pdr.obspm.fr/PDRDB/index.jsp
	The presentation in Baltimore can be found here :
	http://www.ivoa.net/internal/IVOA/InterOpOct2008Theory/IVOA08_Baltimore_Franck.pdf
	
	In this prototype, SimDM has been used to describe micro-physics  
simulations and as you will see, one can :
	a) browse and discover simulations on input parameters
	b) browse and find simulations reproducing observations : that is to  
say to help and solve inverse problems. (Up to now, such queries can  
be done on column densities but, technically, it is also possible to  
do queries on line intensities but it is not implemented yet. For  
example to find all simulations reproducing the CO line J=2->1 with an  
intensity above 1E-09 erg cm-2 s-1.)

	Those are the two most important use cases on the way users would  
like to discover simulations for such theoretical services (excepted  
maybe for spectra that we still need to add. I hope to investigate how  
to do it soon).

	In this implementation of SimDM, we did not use all the boxes of the  
DM. So even if complex, as Gerard said, few people will need to use  
all the DM. We did large use of some boxes (groups for example) that  
people doing large scale simulations would not use, and inversely, we  
did not use some of the classes that would be very relevant to them.

	We have difficulties to describe precisely (to be interoperable) some  
quantities as species or line intensities. But this is a not a lack of  
SimDM but instead a lack of vocabulary in the VO. I hope that efforts  
of standardisation of atomic and molecular physics as well as  
chemistry will help to develop a way to describe precisely transitions  
and chemical species.

	You mentionned the difficulty to use SimDM and implement it in a  
SimDB and you are perfectly correct. There are two kind of difficulties:
	1 - How to map our simulation results  with SimDM
	2 - How to implement SimDB
	
	Concerning the first point, it is indeed true that it is not trivial  
to know how to fill the various boxes of the SimDM with the quantities  
in micro-physics simulations. What is an ObjectCollection in micro- 
physics simulations ? Nevertheless, thanks to Gerard's help, we have  
been able to use it and the result is promissing. It would be too long  
to be described here and so I will write a document to share our  
experience.

	On the second point, the implementation should have been tricky  
because of the complexity of SimDM. Here, the Volute project developed  
by Gerard and Laurent Bourgès is quite usefull because it creates an  
"empty database" quite simply and the developer "just" has to fill the  
database with the outputs of the code. This simplify a lot the work to  
be done. Nevertheless, such implementation require a computer scientist.
	Based on our experience to use Volute, at Paris Observatory, teams  
are now developping two other SimDB based services : an MHD database,  
and a large scale simulations database. In both cases, we will use the  
Volute project.

	Now a word about spectra. Up to now, we did not included spectra in  
the database, even if they are computed during the simulation. We will  
soon investigate the possibility to use SSA and to link these spectra  
with the simulations in SimDB. I hope to do this for the next InterOp.
	
	If I can be of any help to help you and see if SimDM can describe  
your micro-physics simulations please tell me.

	Concerning S3, it seems to me that it does not have the same  
objectives than SimDB. SimDB is more a tool to describe broadly  
simulations so that end-users can browse in the VO to discover  
relevant simulations. S3 seems to me closer to the "old" SNAP with a  
discovery part and a data access part. Maybe it is closer to SimDAP  
than to SimDB. I will have a look again at your document.

	Best regards
	Franck


Le 23 févr. 09 à 02:57, Miguel Cerviño a écrit :

> Hello,
>
> First of all, thanks for the mails :) it looks that the list is  
> alive...
>
> Here there are my five cents:
>
> - Regarding SimDB, I must admit that I need to look at it in more  
> detail: I had been aware of SNAP, but I lost the change from SNAP to  
> SimDB/SimDAP and the first notice I had about the possible use of  
> SimDB on "microsimulations" was in Trieste interop. Any case, I  
> think that it is interesting to discuss the possible uses here, in  
> the mailing list. [Gerard has send some mails about SimDB, SNAP etc,  
> however not too much response has been in the list, and most of the  
> discussion has been done (mainly) in the Interop meetings, where not  
> everybody interested are able to attend, so, reading the mails in  
> the list, it is very difficult to know what's going on...]
>
> - I am still worried about simplicity: Lot of astronomers produce  
> model that would include in the VO (in fact, they put their models  
> in WWW pages), but they have no  knowledge on XML etc... so they  
> just look for the most simple solution (they want a simple solution,  
> otherwise they will not put their models in a VO way). I think that  
> it is needed to offer this kind of flexible solution, apart of other  
> solutions more detailed. Ej. In observations, a Cone Search, or a  
> Name Search, although simple, are very useful and powerful. Of  
> course, other types of searsch are interesting, and the development  
> of ADQL is needed, (but also, it does not means that all queries  
> must be done
> ONLY with ADQL, neither than ConeSearch or Name Search must  
> disappear).
> In this aspect, I am not sue that it is needed that S3 and SimDB  
> converge :)
>
> Cheers
>
> 	Miguel
>
>
>
>
>
>
>

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 1863 bytes
Desc: not available
URL: <http://www.ivoa.net/pipermail/theory/attachments/20090223/08f6a318/attachment-0001.bin>