Report of my expectation from xml-binary working group

Hello,

I heard that this working group is newly formed, and is collecting
information
about uses cases of binary data in XML. I would like to report how I wish
to use binary data in my project.

We are designing a set of language to specify information according to which
a molecular dynamics program starts itself. The same language will also be
used to present the result of the simulation. We decided to use XML as the
platform to develop this language. We take CML as our reference a lot, but
we add a lot of ingredients most relevant to MD computations.

This approaches has been tested with small molecular systems. I think it
works fine, and my boss is happy about the object-oriented design of both
the package and the input/output language. However, when considering to
work on much much larger molecular system on our huge clusters (one of
them has ~1000 CPUs, one has ~800 CPUs, and one with > 2000 CPUs.),
we have to be serious about passing the information of more than 1 million
molecules between the CPUs. A supervising commitee suggests us to use
binary format for passing the computation data, such as the positions and
velocities of the molecules. This can be done in the object-oriented way
without a problem. I have done such a test system. However, it does not
seem natural to include such information in XML.

I think, not only for MD type of computation, many scientific computations
can be benefited from using XML type of data format, especially when they
are performed on grid computers. There are huge amount of numerical data
that should be passed around, while other information is necessary to make
these data semantically robust. Personally I look forward to the rapid
prgress
of the XML binary characterization project. This will very probably be the
preferable solution for our purpose.

Best regards,
Kuo Kan Liang
ICR, Kyoto Univ.

Received on Sunday, 20 June 2004 22:25:38 UTC