I'm using the eXist xmldb for storing metadata/test setups for product tests. Each document contains some identificatory data, the test results, and several interpretations of the data.
Currently I store all of my documents in a single collection, and it has been working fine. However, the size of the collection is now about 20.000 documents, and I have a feeling that response time is increasing. Within a few months, the system is to be put in full use across the company, and I'd guess that in a year there will be about 10 million documents. Is this a problem? Should I split the data over several collections? That would force the queries to be more complex, but maybe it will still be faster?
(As a side note, "exist" is probably one of the worst product names I've encountered. Trying to search for it is a mess.)