A study into the usage of OODB for data analysis at a HEP experiment
Session: C (talk)
Speaker: Akinori, Kimura, Niigata University
Keywords: analysis, C++, data bases, ODBMS's, object-oriented methods
A study into the usage of OODB for data analysis
at a HEP experiment
Ikarashi 2-no-tyo 8050, Niigata, Niigata, Japan 950-21
Takashi Sasaki, Akihiro Shibata and Yoshinobu Takaiwa
KEK, National Laboratory for High Energy Physics
Oho 1-1, Tsukuba, Ibaraki, Japan 305
According to the growth of experimental scale of HEP, software used
for data analysis becomes more and more complicated and huge in
size. However, development of highly reliable software in a limited
period of time is not easy. Also as period of experiments become longer,
maintenance of software become very serious problem. In
order to solve these problems, the object oriented approach is a only
known established way in software engineering at this moment. Recently
many applications such as a GUI are made based object oriented techniques,
and also some applications for future HEP experiments are developed
An extreme difficulty in OOP is persistence of an object. Also due to
a long life of HEP software, evolution of schema must be supported.
Many of experiments are thinking to use OODB to solve these problems,
however, there are very few examples which already measure performance
of OODB. Serious usage and performance measurement of OODB are necessary
to study feasibility of this approach. The comparison of performance between
existing FORTRAN based software and newly developed object oriented one
is very good lesson for this purpose.
We here report the development of data analysis software for the AMY
experiment at TRISTAN using object oriented database. Originally, software
for AMY are coded using FORTRAN and the data are stored into the TRISTAN Bank
System (TBS) format. We converted the DST level reducted data
from the TBS format into the OODB. Several object database models are
examined to stored the objects into the OODB. Because the choice of database
model are thought to be important for better performance. The performance
with different database models are compared to explore better scheme.
We use the Objectivity/DB for this study which is also adapted in RD44
and RD45 at CERN.
In this paper, we will summarize the detail of database modeling and also
the result of performance measurements. The performance comparison is
done between FORTRAN version and newly developed C++ version of software
with different database models to analyze DST level data. We also mention
reliability of the OODB from the point of schema evolution.
Through this study, We designed and implemented the software fully
based on object oriented techniques.