Hi Maria, Maria A. Nieto-Santisteban wrote:
> it is interesting how reluctant you are to use a real dbms considering you > are already taking advantage of the dbms and the fact we (SDSS) already > did the job of creating a galaxy table for you. The SDSS CAS system is a great system. I was impressed how easily I could access the data for my study. I am _not_ reluctant towards a real DMBS. A DBMS is a great tool for assembling the raw data into an easily accessible format. My benchmark showed that the system performs well and was more an argument in favour of it than anything else. There are use cases though where a more simple system with the possibility to "run over the data doing complex things in a massively parallel way", may do a much better job. I believe the answer on which way to go really depends on what exactly we want to do with our data and a better understanding of the types of queries that will be performed. The galaxy clustering was an example, and I still think it would be good to explore other typical algorithms to get a better picture of what we are about to do with our data. There are a couple of things that a typical DBMS is not so good at. I had a list of items here initally - but I feel that this might trigger more discussions/misunderstandings and I don't want to waste anyone's time. I hope that these issues can be discussed without too much emotion. I am sorry that the thread took off in the wrong direction. Happy thanksgiving, - Leif. > > cheers > > Maria > > > > On Tue, 21 Nov 2006, Leif Hendrik Wilden wrote: > > >>Jacek wrote: >> >> >>>Thanks, but I am not sure if we want to put any energy into evaluating >>>ROOT at this point. For a variety of reasons LSST would MUCH prefer to >>>use a real database. We don't even have sufficient resources to look >>>into major dbms players like DB2 and/or Oracle (even though they are >>>very much interested in helping us) >> >>Hi Jacek, >> >>I do not think that ROOT can do the job for LSST and I was certainly not >>suggesting that it should be used. ROOT completely lacks all the >>features LSST will need to dynamically maintain the database at the base >>center. There may be some use cases for ROOT in different areas - but >>this needs a lot more thought and discussion. >> >>My intention was to start thinking of how to evaulate the performance of >>a "real database". With the results from the benchmark I can now roughly >>estimate how well the SDSS SQL server implementation actually works. It >>works better than I thought considering experiences with the MS SQL >>server I had a few years ago, but it is still significantly slower than >>a well implemented file based approach. >> >>Sorry if that did not become clear. >> >>- Leif. >>_______________________________________________ >>LSST-data mailing list >>[email protected] >>http://www.lsstmail.org/mailman/listinfo/lsst-data >> > > _______________________________________________ LSST-data mailing list [email protected] http://www.lsstmail.org/mailman/listinfo/lsst-data
