letkeman wrote:
> I'm writing and implementing a protocol to provide application distribution
> in a fault tolerant scalable and consistent manner across the internet.
For the molecular biology application domain, the disadvantage of
distributed solutions is the overhead to maintain up-to-date data and
application configurations at all the nodes. Molecular biology
applications frequently suffer from different data at the service nodes,
which means that a consistent distributed query will potentially result
in inconsistent or at least incomplete answers. Therefore, I would
suggest that you introduce quality control of data and applications into
your load-balancing paradigm if you consider to go into database
searching (which admittedly might not be a suitable task for a
screensaver). A potential drawback of the architecture you are proposing
might be the centralized tree-like structure, which is somehow
contradicting the fault tolerance.
Just for the records, some years ago the Biocomputing Basel group
(Biozentrum Basel) has written the Hierarchical Access System for
Sequence Libraries in Europe (HASSLE) as a tool to access sequence
databases remotely: Comput. Applic. Biosci. 10, 3135-3140 (1994). It is
no longer maintained but you might find some aspects of the load and
quality assessments interesting. As the original server as ceased to
exist and none of the group is working at Biozentrum Basel any longer a
colleague of mine as volunteered to make an archive available at
http://www.doelz.com/software/hassle (the code and its documentation are
still freely available, public domain).
Regards
Reinhard Doelz
up to 1996, Biocomputing Basel, Basel University
