When querying for a specific record, the database engine needs to perform a table scan, comparable to a reader trying to find a specific word in a book by thumbing through its pages. This process can be accelerated by parallelizing the process (as is done with using Spider and PaQu) and/or by using indexes: just like a book’s index helps the reader to find the page with the desired word more quickly, a database index helps accelerating database queries. This becomes especially helpful, where large tables with millions of records like in the MultiDark database are involved.
However, database users don’t need to worry about indexes too much. First, the database server picks the most useful index itself, the user does not need to interfere here (and actually cannot do it anyway, since PaQu/Spider doesn’t support this). Second, the database is parallelized, so the benefit of indexing is not that great anymore.