XML::Parser is a good candidate here.

All you want is to go through the XML document once, and process each group of tags one by one. At any given time, you only need a handful of those tags and it content stored in memory.

Put your code database access code in handlers.

With XML::Parser, the parser only does the parsing (which is what you need), but the rest is up to you, including whether you want to store the XML file in a data structure, or how big a portion of it should be stored in the memory at a given time. As I said, in your case, all you want at a given time is the content of a handful tags.