Mind explaining the origin of your 32 TB figure? I must be missing something enormous, but as far as I can tell the SciMag database dump is 9.3 GB, the LibGen non-fiction dump is 3.2 GB, and the LibGen fiction dump is 757 MB. That's a pretty huge divergence.
The Libgen Desktop application (which requires only a copy of the database) would then use the DB metadata to make LibGen locally searchable, and would only retrieve the individual books/papers on request?
Source: http://gen.lib.rus.ec/dbdumps/