I'm looking for further information (going beyond information given in e.g. GEO publications) on how databases like NCBI-GEO (http://www.ncbi.nlm.nih.gov/geo/) are set up. I'm curious how such a system can be designed in order to allow growing numbers of experiments, samples and expression data - without the need of redesigning a database layout, when a new type of data is available. Would be glad, if you could share resources, examples, detailed explanations or your own experience - anything welcome ;)